Structural Dynamics For Engineers, 2nd Edition
Structural Dynamics For Engineers, 2nd Edition
Structural Dynamics For Engineers, 2nd Edition
Structures
Practical problems and their solutions
G.A. Rombach
University of Hamburg-Harburg
Index 351
vi
vii
viii
ix
xi
xii
xiii
xiv
Chapter 1
General
A few years ago, powerful computers were needed to conduct FE analyses, and only
experts and big consulting offices were able to perform this analysis. Nowadays, the
designing of an entire building can be handled by a simple PC. Graphical input makes
it easy to generate 3D FE meshes with several thousand nodes. Computer programs
can design concrete, steel or wooden structures that have linear or nonlinear material
behaviour, under static or dynamic loading. There no longer seems to be any limitations.
Nonetheless, this development has led to an increasing number of cases where FEM has
been misused.
As daily experience shows, results from computer calculations are often trusted with
blind faith. Users assume that expensive software for design of structures must be free
from any error. A graphical pre-processor and a user-friendly input of systems and load-
ings may suggest that a computer program has a high degree of technical competence and
reliability. Nevertheless, as practical experience shows, this confidence can only be
justified to a very limited degree. Almost no software is free from errors. Therefore a
critical approach is appropriate, as program errors may also occur in software that
has been in use for a long time and which may not have been recognised to date.
It should always be kept in mind that FEM is only a numerical tool based on numerous
assumptions and simplifications. This must be considered when using software for design
of structures. Otherwise, the result of numerical calculations may turn out to be totally
wrong. For explanation purposes, the following is a very simple example: a plate element
only provides a numerical model of a real slab. It is assumed to have a linear strain
distribution over its depth under pure bending. There are no stresses at the midplane.
With such a plate element, one will never be able to estimate the normal forces of a
simple supported rectangular slab due to temperature changes or shrinkage, even if the
supports are fully restrained in horizontal directions.
The modelling, the discretisation, of real reinforced concrete structures is the focal point of
this book. The fundamental aspects are illustrated by practical examples of concrete struc-
tures. This book does not look into the fundamental basis of FEM, as numerous publica-
tions are already available (see, for example, Bathe, 1982; Hughes, 1987; Zienkiewicz and
Taylor, 1989). The so-called state of the art of the FE techniques will not be discussed, as
there seems to be a great gap between ongoing research and the day-to-day problems that a
practicing structural engineer has to face. An engineer has neither the time to make highly
sophisticated numerical models nor the experimental data to verify his analysis. He is not
even interested in the ‘correct’ results. His goal is simply to estimate the required amount of
reinforcement and its accurate arrangement (the ‘dimensioning’ of a structure), in order to
build a safe and economical structure. The calculation of the member and internal forces
and moments is only a required step to reach this goal.
The examples shown in this book are calculated using standard software, used in day-to-
day practice, and not with one of the advanced general-purpose FE packages such as
ABAQUS, ADINA or ANSYS, which offer a great variety of different elements and
material models. Hence, the reader can easily verify the given examples using his or
her own software. A further reason for the strong relation to practical design is that a
user of a software package is not usually familiar with its theoretical background, and
hence cannot modify it. It does not help the user to know, for example, that a reduced
integrated three-noded shell element may give better results than a full integrated six-
noded isoparametric element. He is just using the ‘black box’. The user, however, is
supposed to have sufficient knowledge to see and solve the problems that may occur
in a FE analysis. This is where this book is intended to provide help.
It is surprising that, in structural engineering, the use of FEM causes numerous problems,
especially as this numerical method was first used by structural engineers. The world’s first
electronic programmable calculator was built by a structural engineer named Konrad Zuse
(Zuse, 1984) in May 1941. He was tired of repeating calculation procedures when designing
structures. Zuse also developed the first algorithmic programming language ‘Plankalkuel’.
In other fields of engineering, such as, for example, in the automobile or aircraft industries,
the numerical FE analysis of highly complex problems, such as the crash behaviour of a
car, the optimisation of aerodynamics or the processes in an engine, have become a
day-to-day practice. The reason for this discrepancy is that these sorts of costly and
complicated computer calculations are only economical for mass products. In contrast,
a building is usually a unique structure whose costs depend on several factors, and not
just the cost of building material. The numerical modelling of the complex behaviour
of the composite material ‘reinforced concrete’ causes far greater problems than the
elasto-plastic bi-linear behaviour of metals.
The uncritical or erroneous use of FE software can lead to serious damage, as the
collapse of the Sleipner A platform (Figure 1.1) impressively demonstrates (Holand,
1997; Jakobsen and Rosendahl, 1994). This so-called Condeep-Platform had a total
height of 110 m. The four towers rested on 24 cylindrical cells, each having a diameter
of 24 m. On 23 August, 1991, the concrete platform collapsed completely during its
lowering and sank down to the seabed. The actual financial damage was estimated at
about US$250 million. The cause of the total collapse was found to be the serious
inaccuracies in the global FE analysis of the structure and faulty reinforcement arrange-
ment in the connection area of the cell walls. The element mesh used for the calculations
was too coarse to determine the actual forces in the members. This disaster raises a
critical question, whether this sort of accident could still be allowed to happen nowadays.
The essential causes of this case of damage and the consequences of the numerical
analysis are discussed in detail in Section 5.1.1.
In this respect, it is the author’s opinion that the modelling of a structure as a whole with
the help of shell elements, for example, multi-storey buildings or bridges, is rarely mean-
ingful, although this approach is often encouraged by software companies. A structural
engineer must always be able to understand the behaviour of any complex structure and
to idealise it, so that the flow of forces can easily be understood and calculated. Complex
FE calculations can then be used to lower any excessive safety margins of simple models
and produce a far more economical structure. However, complex FE models must never
be used to replace either the design engineer or any of the engineer’s missing expertise.
Costly, sophisticated analyses do not always lead to more realistic results. Furthermore,
the amount of a FE analysis should be considered with respect to the degree of accuracy
that is actually needed. The results of any calculation can only be as accurate as the
underlying assumptions of its numerical model. One should always keep in mind that
there can always be considerable variation in the actual loading of a structure and in
the properties of its materials.
Continuous
structure
(slab) Stresses sx
my
y
z
Midplane
Finite Elements
xi xi
yi yi
xi
xi
yi yi
Nodes
Nodes
global stiffness matrix [K], from which the unknown nodal displacements {u} can be
calculated.
[K] . {u}={F}
where:
The main task is to find form functions that can approximate the behaviour of a special
structural element and satisfy the compatibility condition. For simple elements such as
the two-noded truss element shown in Figure 1.3, the relation between the nodal
u2 u5
y
u6
u1 u4
u3 x
l
EA : axial stiffness
EI: bending stiffness
stresses
V ec s N s M t
M
N
E .A 0 0
.
-E A 0 0
l l
12EIx 6EIx - 12EI x 6EIx
0 0
F1 l3 l2 l3 l2 u1
F2 6EIx 4EIx - 6EI x 2EIx u2
0 0
F3 l2 l l2 l u3
= .
F4 -E A 0 E .A 0 0 u4
l 0 l
F5 u5
F6 u6
0 - 12EI x - 6EIx
0
12EIx - 6EI x
l3 l2 l3 l2
4-noded
element 8-noded element
s = –2 s = –2
s s
r s=0 r s=0
s = –1 s = –1
r = –1 r=0 r = +1 r = –1 r=0 r = +1
u(r, s) u(r, s)
1 1
0.8 0.8
u1 = 1
u1 = 1
0.6
0.6
0.4
0.4
0.2
0.2 1 0 1
0 –0.2 0
0 s –1 s
–1 –1
0 –1 0 1
1
r r
forces {F}e and the nodal displacements of each element {u}e can be found by means of
equilibrium conditions. Shape functions are not needed for this simple approach.
For more complex elements such as, for example, plate or shell elements, virtual work or
virtual displacement principles are used. Figure 1.4 shows the displacements within a four-
noded and an eight-noded isoparametric element for a unit deflection of u1 ¼ 1 at node
number 1. It can clearly be seen that both elements have different displacement distribu-
tions w(x, y) and thus different strains (" w0 (x, y)), even if the nodal values are equal.
The basics of the FE technique set out in the preceding text therefore give the following
important conclusions:
1.2.2.2 Loading
An FE model is based on nodal forces and nodal displacements (Figure 1.3). This is still
valid when the software allows for an arbitrary load arrangement to be possible. In this
case, the computer program calculates the equivalent nodal forces. Depending on the size
of the elements, this may lead to a considerable extension of the actual loaded area. This
is important, for example, in the analysis of slabs under concentrated loads, such as
wheel loads on bridge decks.
A single-span beam modelled by only three elements is used to demonstrate the problem
of equivalent nodal forces (Figure 1.5). The beam on the left is loaded by a uniform load
of q ¼ 10 kN/m. The software estimates the equivalent nodal forces to F ¼ 50 kN for the
inner nodes. Thus, the member forces for a beam loaded by two single forces instead of a
uniform load are calculated. This results in incorrect shear forces. A beam with a single
load in midspan (Figure 1.5 right) gives the same member forces. Thus, for this simple
example, the FE analysis cannot distinguish between a uniform load and a concentrated
load at midspan.
15 m 15 m
displacement displacement
75 kN
50 kN
3×5m
25 kN =0 25 kN FE-nodal forces
Furthermore, loads on fixed nodes are mostly neglected in the design and hence in the
results. The total support force is only the sum of the loads on the unrestrained nodes.
This has to be considered when support forces from FE analyses are used for the loading
of other members. Loads on columns and walls, which may be fully restrained in the
numerical model, are neglected by some structural software.
The nodal forces result from the chosen element with respect to its shape functions and
not from ‘engineering’ experience. Figure 1.6 shows the nodal forces for a four-noded
and an eight-noded plate element under constant vertical loading. For the four-noded
element, the load is distributed equally to the nodes, whereas the eight-noded element
shows uplifting forces at the edge nodes.
Figure 1.6 Nodal forces for a four-noded and eight-noded isoparametric plate element under
uniform vertical loading
1/3 1/3
1.2.2.4 Discretisation
The numerical modelling of a real structure, the ‘discretisation’ (subdividing a structure
into a finite number of different elements), is where most of the errors are made
(Figure 1.7). Here, among other matters, special note should be made of the following
factors.
Size of elements
Some years ago, the number of elements was limited due to the capabilities of both
computer hardware and software, whereas nowadays a sufficiently fine discretisation
of an entire building can be done without any major difficulties. Furthermore, an element
mesh can be produced very quickly by graphical pre-processors. Nevertheless, even
automatic mesh generation should not be used in an uncaring or uncritical manner;
engineering knowledge is still required. An inadequate modelling of apparently
irrelevant details, such as, for example, small cantilever slabs of a bigger plate (Section
4.12.2) or openings in a flat slab near columns, can lead to faulty calculations and an
unsafe design. A sufficiently fine element mesh should be used in regions of high
deformation or stress gradients.
10
Real actions
Real structure
real actions
Design
Equivalent structural loads Beam
system and loading
Pin support
Nodal loads
Finite Element
model
nodes
element
Calc. of nodal
displacements,
stresses and
member forces
Calc. of required
reinforcement
(incl. min. As)
detailing
construction
real structure
that the various element types cannot be simply joined together, even if this is not
hindered by the computer software (see Section 3.2.3).
Support
The numerical modelling of the supports of any structure should be carried out with
great care, as this is where fundamental mistakes can be made. This aspect of a design
will be explained for a beam in Section 2.9, for a deep beam in Section 3.2 and for a
flat slab in Chapter 4.
Singularities
Singularities, or infinite stresses and internal forces, occur in slabs or shear walls under
highly concentrated point-loads. It should be kept in mind that this problem only occurs
in a numerical model caused by simplifications and assumptions of element behaviour.
For example, the assumption of a linear strain distribution in a slab is not valid in the
region of pin supports. A real structure does not show any infinite internal forces. In
regions of high compression, the concrete may ‘yield’. Tensile stresses may cause the
formation of cracks. The high forces shown by an FE analysis in singularity regions
do not happen in a real structure, and thus need not to be considered in the design,
except by some additional reinforcement. Nevertheless, the user should know about
these problems and how the results of the numerical analysis (e.g. bending moments
and shear forces) should be interpreted.
11
Kinematic systems
In general, a software program gives out warnings or stops the analysis when a
structure becomes kinematic. In these cases, the system of equations and thus the
nodal displacements have no unique solution. These warnings can, however, be easily
overlooked due to the great quantities of data produced, which can only be checked
by graphical control. In addition, some software packages may automatically fix all
kinematic degrees of freedom. If the distribution of forces is reasonable, these mistakes
will not be noticed, and the structure will be designed for actions and erroneous member
forces of a kinematic structure.
REFERENCES
Bathe K-J (1982) Finite Element Procedures in Engineering Analysis. Prentice Hall,
Englewood Cliffs.
CEN (European Committee for Standardization) (2004) Eurocode 2, Part 1: Design of
concrete structures – General rules and rules for buildings. December.
Holand I (1997) SINTEF Report No. 17: Sleipner A GBS Loss. Main Report, Trond-
heim.
Hughes TJR (1987) The Finite Element Method. Prentice Hall, Englewood Cliffs.
Jakobsen B and Rosendahl F (1994) The Sleipner Accident. Structural Engineering
International 3: 190–193.
Zienkiewicz OC and Taylor RL (1989) The Finite Element Method. London.
Zuse K (1984) Der Computer – Mein Lebenswerk. Springer, Heidelberg.
12
Chapter 2
Truss and beam structures
The internal forces and moments in trusses and beams can be determined by means of
FEM as well as by load transfer methods. Either method produces the same model
problems when considering actual structures.
Truss systems are not only used for analysing beams and columns. Nowadays, they
are also used as an equivalent system for the structural analysis of flat shell structures
such as shear walls (see Section 2.5) or T-beam bridges (see Section 2.8). The main
reason for this simplification of spatial shell structures by simple beams is that the
amount of calculations needed for beam or truss systems is considerably less than that
needed for 3D shell design. This is especially true for the calculation of reinforcement
requirements.
This chapter starts off by examining various detailed problems related to the correct
modelling of the so-called ‘discontinuity regions’. Discontinuity regions are those
where the essential assumption of truss elements – the Bernoulli hypothesis, which
states that the section remains approximately plane before and after loading – is not
valid. Examples considered are those of beam column intersections, frame corners,
beams with abrupt or smooth change of cross-sections or openings, halving joints and
inclined haunches.
Even if the action effects in these discontinuity regions cannot be calculated accurately, it
is important to model the different stiffnesses of a structure. The main difference in the
various models used for frame corners are shown using the examples of a frame bridge
with shallow foundations and the transverse design of a hollow box girder bridge. The
modelling of a foundation slab bedded on ground (Section 2.4.1) and a bridge column
supported on piles (Section 2.4.3) will also be discussed.
Following the discussion of these detailed problems, we look at the design of whole
structures. The calculation of coupled concrete shear walls with large openings, which
are used as bracing elements in high-rise buildings, is shown in Section 2.5. Then, the
modelling of a complex bracing system of a high-rise building, consisting of core
elements of different shapes and shear walls, is discussed, followed by the analysis of a
hollow box girder and a T-beam bridge by means of a grillage (plane grid) system.
This chapter concludes by looking at some problems in the calculation of reinforcement
requirements in beams and material nonlinear analysis.
13
In design calculations, ‘exact’ values are often not needed. In general, the purpose of the
structural analysis of a frame is not to calculate the maximum member forces at the beam
column junction of the centrelines, but rather at the inner face of the corner for bending
or at a distance of 0.5d to 1.0d for shear, depending on the design code (d ¼ effective
depth of a cross-section). However, the exact modelling of the stiffness of the frame
corner is important, as these can have a large effect on both the internal forces and
the deformation of a member or structure.
In a real structure, the frame corners generally behave like a stiff diaphragm. Therefore,
in these areas, the nodes of the truss model cannot move independently of each other (see
Figure 2.1). The simplest way of taking this condition into account is through a stiff
coupling of the corner nodes. An alternative is to introduce an additional stiff inclined
truss element. However, this may cause numerical problems because of the great stiffness
differences of the system.
The following two examples will show the influence of the numerical modelling of beam
column intersections. The first one is a portal frame bridge with shallow foundations,
Detail
Deformation
neglecting the rigid
corner region
14
which is widely used for road underpasses. The second example deals with the transverse
design of a hollow box girder bridge.
15
g asymmetric earth pressure (increased active earth pressure on one wall and active
earth pressure on the other)
g live load on the backfill (left/right)
g traffic loads on the bridge
g traction and braking forces of the traffic
g temperature variations, for example (for concrete bridges in Germany, see EC1),
TM,neg ¼ Te,min T0 ¼ 27 K T0, TM,pos ¼ Te,max T0 ¼ þ37 K T0
g differential temperature of the horizontal member TM,pos ¼ þ15 K and
TM,neg ¼ 8 K (for concrete bridges in Germany).
Furthermore, it must be remembered that, due to the short length of the foundation
beam at the inner face of the frame (see Figure 2.3), member forces, as calculated by
Real structure
70
6.20 m
9m
80
3m 90 50
Truss model
1 21
Detail 201 301
framework corner
Detail framework corner
A and q is missing
6.25 m
Double
cross-section
Rigid Real Loading on
coupling loading truss model
9.90 m
101 221 113 133 321 121
80 Node number
Springs
3.45 m 95
16
the computer program, cannot be used for the purpose of design in this region. This is
without doubt a discontinuity region to which the Bernoulli hypothesis does not
apply. Nevertheless, the entire foundation should be modelled, since not only the
member forces of the structure but also the soil pressure distribution and the settlements
must be determined.
The truss elements will be located at the centreline of the beam and columns (Figure 2.3).
But, in this case, the cross-section of the corner region will not be modelled
correctly. However, the additional dead load can be neglected in general. The elastic
bedding of the foundations can be modelled with individual springs or with
special boundary elements. In this example, the bedding modulus is taken to be
ks ¼ 10 MN/m2.
The walls are loaded by a trapezoidal earth pressure distribution. It should be kept in
mind that the height of the truss model is smaller than the height of the real structure
(i.e. the height of the column is 6.25 m, whereas the height of the real structure is
6.20 þ 0.80 ¼ 7 m; see Figure 2.3 for details). Therefore, additional horizontal loads
must be applied on the corner nodes. The same applies to the vertical loads on the
horizontal member.
Figure 2.4 shows the calculated bending moments, the shear forces and the displacement
of the structure under dead load only. These were calculated both with and without
coupling of the corner nodes to illustrate the effect of this measure.
With this system, as expected, the influence of the nodal coupling is very small (see
Figure 2.4). This further applies to the relevant design forces from an unfavourable
combination of relevant actions. The differences are estimated at less than 3% for the
bending moment in the critical sections, and approximately 6% for the vertical
displacement at midspan.
The design of a bridge is usually done separately for the transverse and longitudinal
directions. In the following example, we are looking at only the transverse behaviour.
To do this, a 1-m-wide section is taken of the bridge and modelled by truss
elements. The variable depth of the beams and the inclination of the axis of gravity
are taken into account (see Figure 2.7). There are pin supports under the webs,
which are fully restrained in the vertical direction. This is a rough simplification of
the real behaviour (see Figure 2.75). Distortions of the cross-section caused by
unsymmetrical loads are neglected. The system is also a frame structure. Of further
interest is the modelling of the corners, the junction between the inclined webs and the
deck slab.
17
Figure 2.4 Shear forces, bending moments and deflections with and without coupling of the
corner nodes for the load case ‘dead load’
No. with
coupling
268 278
(rel. section) 478 487 (rel. section)
Bending moment
–224 (kNm/m) 234
175 –182
(rel. section) (rel. section)
86
–86
–38 –20
Vertical deflections
(mm)
4.9 5.2
16.1 15.7
The behaviour of the structure will be examined under two different theoretical unit
loadings, a linear loading of q ¼ 10 kN/m at (a) the outer edge of the cantilever slab
and (b) at midspan of the top slab.
For the load on the cantilever slab, the bending moments and displacements are only
slightly (by 5%) influenced by the modelling of the corners (with/without coupling)
(see Figure 2.9). For the line load acting at midspan of the top slab, the coupling of
18
Figure 2.5 Standard and deviator segment, Second Stage Expressway, Bangkok (Rombach, 1997)
the corners causes the bending moments and the deflection at midspan to vary by
approximately 10% and 50%, respectively. The latter value definitely cannot be
neglected. The difference between the moment and displacement shows that the coupling
of the corner nodes mainly reduces the span width. The increased restraint of the top slab
does not have a significant effect on the member forces. The bending moment Mf of a
fully restrained single-span beam under concentrated load at midspan is proportional
Figure 2.6 ‘Standard’ span, Second Stage Expressway, Bangkok (Rombach, 1997)
19
Figure 2.7 Cross-section and resulting truss model of a hollow box girder bridge (standard
segment, Second Stage Expressway, Bangkok)
20 1370 cm
40
140 150 20
35
240
20 cm
85 250 75 550 cm
Junction
1 2 3 4 12 13 14 15 16 17 18 26 27 28 29
1 5 11 19 25
2 3 4 5 6 7 10 11 12 13 14 15 16 17 18 20 23 24
25 26 27 28
6 7 8
8
9
9 10 19 20
21 22
22 23 24
21
31 51
31 32 50
50
3233 49
49
33 34 48
48
34 3536 47
37 38 39 40 41 42 43 44 45 4647
36 37 38 39 40 41 42 43 44 45
to the span length to the power of 1 (Mf ¼ Fvl/8), whereas the deflection at midspan f is
proportional to the span width to the power of 3 ( f ¼ Fvl3/[192 . E . I]). However, it must
be pointed out here that the influence of the node coupling depends largely on the
geometry of the system. The width of the webs of segmental hollow box girder bridges
with external prestressing is very small in relation to in-situ constructions with bonded
tendons. The latter usually has a web thickness of more than 50 cm due to the space
required for the tendons and the couplers. Here, the influence of the coupling of the
corner nodes on the shear forces and moments can be much larger.
With shell or membrane models (diaphragms), the behaviour of the ‘elastic’ structure can
be analysed more precisely. This will be shown in the Chapters 3 and 4. However, the
4 5 6 1 3
2
5 6 7
4
9
8
20
Figure 2.9 Bending moment distribution and deflections due to a single force at the free edge of
the cantilever slab (top) and at midspan of the deck (bottom)
q = 10 kN/m –33.5
Bending moments [kNm]
–20.5
–12.6
(–13.8)
–17.4 –15.7
(–18.3)
–10.8 ( .. ) system
(–10.0)
with coupling
–5.5 7.43
q = 10 kN/m –1.45
Deflections [mm] (–1.4)
3.6
(3.3)
6.2 ( .. ) system
(5.6) with coupling
7.3
(7.7)
Deflections [mm]
–0.15
q = 10 kN/m (–0.1)
0.44
(0.30)
results of such calculations are presented here in order to show the complex stress
distribution in corner regions.
Figure 2.11 shows the membrane forces in a web-deck slab-junction for a load of
10 kN/m acting at the free edge of the cantilever beam (FE mesh; see Figure 2.10).
21
One recognises a very complex force distribution, which certainly cannot be modelled
by beam or truss elements based on the linear strain theory (the Bernouilli hypothesis).
Outside this discontinuity region, the strains and stresses are linear over the depth of
the section. The discontinuity region extends up to a distance h from the continuity.
In Table 2.1, the bending moments and the displacements in the relevant sections of the
structure are compared with each other for both load cases. Good agreement can be seen
in the bending moments. However, there are relatively big differences in the midspan
displacements.
Besides the structural analysis, good detailing of the reinforcement in corner regions is
very important with respect to both the load-bearing capacity and the serviceability of
the structure (see Figures 2.8 and 2.12). Many experimental investigations were
conducted and several theoretical design models have been evaluated to design such
sections.
Figure 2.11 Main membrane forces and distribution of the horizontal resp. vertical membrane
force nx (flange) and ny (web) over the section depth; loading: q ¼ 10 kN/m at the free end of the
cantilever slab
1357
857
-1344.
-787
–638
q = 10 kN/m
Shown section
–459
Z X
22
5 12 3
5 3
4 4
Bending moments: kNm/m Section 1-1 Section 2-2 Section 3-3 Section 4-4
For frame structures with inclined haunches, the variable depth and the inclination of the
axis of gravity should be modelled in addition to the coupling of the corner nodes.
23
Figure 2.12 Frame corner with high opening bending moment; strut-and-tie model and resulting
arrangement of reinforcement (according to Schlaich et al., 1987)
C
h
T
C Z
T
h
As will be demonstrated later, the inclination of the axis of gravity generally only has a
small influence on the action effects, where the system is not horizontally restrained.
Nevertheless, this inclination should not be neglected, as it is very important with respect
to the design for shear (Figures 2.15a–c). A straight beam axis with constant depth of
elements will cause mistakes as the slope of the compression strut with respect to the
tension chord is neglected. Therefore, the resulting change in the design shear resistance
VRd is not considered (see Figure 2.15c).
VRd ¼ VRd;s þ Vccd þ Vtd (Eurocode 2, 2004: Part 1, Equation 6.1) ð2:1Þ
Centreline
Nodes
24
1:10
50 cm 93 cm
14.0 m 14.0 m
h = 50–93 cm
Figure 2.15 Various models for an inclined haunched beam and resulting internal forces and
moments
Fsd Fsd
Neutral axis
MEd
MEd Fcd a
Vccd Vccd
Truss system with inclined axis
Fcd
Loads and internal forces
Vtd Fsd
Fsd
Fsd Fsd
FCd FCd
Truss system with straight axis
Neutral axis
25
where:
VRd,s is the design value of the shear force that can be sustained by the yielding
shear reinforcement.
Vccd is the design value of the shear component of the force in the compression
area, in the case of an inclined compression chord.
Vtd is the design value of the shear component of the force in the tensile
reinforcement, in the case of an inclined tensile chord.
The support conditions have to be checked when the inclination of the axis of gravity is
considered in the FE model. For a straight single-span girder under uniformly distrib-
uted loads, a horizontal restraint of the supports does not change the member forces
and moments in the structure. This is not true for inclined truss elements where the
support nodes are fixed in the horizontal direction. The strutted frame system is modelled
in the following example. Figure 2.16 shows the member forces and deflections for a fully
restrained single-span beam of variable depth (inner span of a bridge). The span length is
50 m. In order to show the influence of the variable depth of cross-section, a relatively
large depth of h ¼ 4.0 m has been chosen at the supports, whereas in midspan the
depth is only h ¼ 1.0 m. Two different support conditions are considered:
As can be seen from the distribution of the member forces, a strutted frame system results
if both supports are fully restrained. This causes high normal forces, even for the slender
beam in the example. In Model (b), the support bending moment for the fully restrained
system are reduced by 26%, the shear forces by 19% and the displacement at midspan by
approximately 10% with respect to Model (a).
It should be noted that, per definition, normal forces are always in the direction of the
beam axis and shear forces are perpendicular to it. That is why the shear forces at the
supports are not equal to the vertical support force of 750 kN.
The manual calculation of the internal forces of a beam with variable depth can be made
with the help of Figure 2.17. The diagram shows the relationship of the bending moment
of a beam with constant depth to the one of linearly increasing depth, depending on the
length of the inclined haunch and the depth of the beam ha and hh. For the previous
system, the following bending moments are calculated manually:
Table 2.2 lists the member forces at the supports and midspan for the various models.
The results of a shell analysis are also given for comparison.
The internal forces and the displacements of the truss model are confirmed by the shell
model. There is also good agreement with the manually calculated results. As can be seen
26
Figure 2.16 Member forces and displacements of a single-span haunched beam for different
support conditions
1.0 m
4.0 m
12.5 m 25 m 12.5 m
50 m
Bending moment
Fully restrained beam Beam restrained in vertical
direction and against rotation
–8240 kNm
kNm
–6520
1150 kNm
Shear force
–375 kN
607 kN
–375 kN
kN
–747
Normal force
–1213 kN
kN
–1176 kN
–1258
–45 kN
–90 kN
Deflections
26.3 mm
29.3 mm
27
Figure 2.17 Factors for the calculation of the support and midspan bending moment of a
haunched fully restrained beam (linear increase of depth)
as aF
hh
1.4 ha
ha/hh = 4.0
a a
ha/hh = 3.0 l
1.3
Mhaunch = as,F* Mwithout haunch
ha/hh = 2.0
1.2 Example
Support bending
input parameters:
moment
ha/hh = 4.0/1.0
1.1 ha/hh = 1.25
a/l = 12.5/50 = 0.25
bending moments of a fully restrained
1.0 a/l beam with constant depth:
0 0.1 0.2 0.3 0.4
Midspan
moment
q .l 2 30.50 2
ha/hh = 1.25 Ms= – 12 = – 12 = –6250 kNm
0.7 q .l 2 30.50 2
MF= – 24 = 24 = –3125 kNm
Analytical Truss Truss analysis Truss analysis Shell analysis Shell analysis
analysis analysis inclined axis inclined axis beam system fully system not
(Figure straight beam fully not restrained in restrained at restrained in
2.17) axis restrained horizontal the supports horizontal direction
direction at the supports
Support
N 0 0 1258 90 1051 0 kN
V 750 750 607 747 750 750 kN
M 8250 8225 6520 8240 6372 8181 kNm
Midspan
N 0 0 1176 0 1150 0 kN
V 0 0 0 0 0 0 kN
M 1156 1150 1100 1150 1150 1195 kNm
Displacement in midspan
w 29.2 29.2 26.3 29.3 26.7 29.1 mm
28
from Figure 2.16, the horizontal restraint of the beam has a great influence on the
internal forces.
To summarise: if the system is not restrained in the horizontal direction, and if there are
no normal forces, the inclination of the axis of the truss elements can be neglected with
regard to the member forces. However, for a shear design, the elements should be given a
variable depth.
h1
h2 h
h1
h2
h h h h
h h
h
b
h1
h1
h2
b
h2
29
Figure 2.19 Horizontal membrane forces ( ¼ normal stress width of beam) in the region of a
halving joint (shell analysis)
Membrane force
in horizontal direction
176 259
122 142
System
–190 –176 –137 Load q = 1 kN/m
60 80
Shown section
2.5 m
9.5 m
124
187 Width of beam b = 0.22 cm
8 138
The location of the beam axis changes suddenly at a halving joint. This can easily be
modelled by coupling the nodes in the joint similar to a frame corner. In the region of
openings, two separate beams for the compression and tension chords should be used
(Figure 2.20). The normal and bending stiffness of these elements should be fixed
taking account of possible cracking in the tension zone or the ‘yielding’ in regions
with high compressive stresses and the resulting reduction of stiffness.
Truss system
30
q = 10 kN/m
System 1 80 cm
1250 cm
Width of beam b = 22 cm
System 2
40 cm
1m
20 cm
System 3
40 cm
1m 20 cm
Only one half of the structure is shown, as the system and its loading is symmetric about
the midspan. The following three systems have been analysed (Figure 2.21)
g System 1: beam without opening
g System 2: beam with an opening (1.0 0.4 m) at midspan at the lower side of the
cross-section
g System 3: beam with an opening (1.0 0.4 m) at midspan on the centreline of the
cross-section.
Figure 2.22 shows the bending moment distribution of the various models, both with and
without considering such openings in the numerical model.
There is no change in the bending moment distribution of system number 2, as there are no
normal forces acting on the beam. Therefore, the bending moment distribution does not
depend on the opening with respect to the model. Nevertheless, a sudden change in strain
can be observed near the opening (see Figure 2.23), which has a small influence on the defor-
mation of the given structural system. Due to the assumption of the numerical analysis (elastic
material behaviour, Bernoulli hypothesis), the strain in the concrete section is equal to:
c M
"c ¼ ¼ h=2
Ec Ec I
where:
M is the bending moment at the face of the opening
h is the overall depth of the cross-section left or right of the face of the opening
I is the second moment of area left or right of the face of the opening.
31
symmetry axis
System 1: without opening midspan
Element axis = neutral axis
191
System 2a: with opening at lower edge, straight element axis
Element axis = neutral axis
191
System 2b: with opening at lower edge, element axis = neutral axis
Element axis = neutral axis
191
System 3a: with centric opening, straight element axis
Element axis
191
N = –320 kN
5.5
N = +320 kN
189
In the case of a centric opening, the bending moment of the unweakened cross-section is
replaced by a compression and a tension force. This different behaviour has to be
modelled by two separate beams. No normal forces are calculated when this opening
is neglected, and the beam axis is kept straight. This means that the stiffness is not
modelled correctly, resulting in a doubling of the midspan displacement (Table 2.3).
32
Figure 2.23 Stress distribution near the edge of the opening – beam system
1 1
For comparison purposes, the internal forces and the deformations of the beam are
calculated by a shell model. Figure 2.26 shows the membrane forces and the distribution
over the depth of the beam in various sections. In most sections, a linear strain distribu-
tion can be seen. Here, the Bernoulli hypothesis is valid. Large differences are only
obtained in the vicinity of the opening (see Figure 2.24). Thus, a beam system, which
is based on a linear strain distribution, will always lead to incorrect results to a greater
or lesser extent.
Design of the discontinuity regions can be done by strut-and-tie models (Figures 2.25 and
2.27). The results of a linear elastic shell model can be used to evaluate the load paths.
It must be noted that strut-and-tie models are only valid for the ultimate limit state
design. A fully cracked structure is assumed. Therefore, these models cannot give any
33
information with regard to the serviceability of the structure (cracking). However, it can
be useful to reduce the permissible stresses in the reinforcement in order to reduce the
crack width in the D-region.
It should be noted that the shear force in the opening will be carried mainly by the
compression member. This has to be considered in the design of the opening region.
The influence of the different FE models on the member forces in beams with openings
was not significant in the earlier-mentioned example. However, this only results from the
statically determinate structure and the assumption of a fully elastic material behaviour.
Figure 2.25 Strut-and-tie model for a halving joint (Schlaich and Schäfer, 1998)
+M B D B +M
C2 T1 T2 » 0.5T1
C1
T3 » 0.8T2 T3 = C2 tan a
T1 C1 » 0.8T2
T2 a C2
Reinforcement lb,net Reinforcement
34
Figure 2.26 Main membrane forces in various sections (system and loading see Figure 2.21)
Axis of symmetry
midspan
6042
One of the main questions is the stiffness of the fully tensioned chord in the tension zone
of the opening. Due to cracking of concrete, the stiffness is significantly reduced with
regard to the elastic value.
This problem is clearly demonstrated in the following example for a fully restrained two-
span beam under uniformly distributed loads (Figure 2.28). The beam has an opening of
20 50 cm close to the intermediate support. Three different models for the opening
region are used. In the first system, the opening is neglected, whereas in the second,
the tension and compression chord in the opening is modelled by two separate beams
35
that are rigid, coupled with the centre nodes of the undisturbed beam (Figure 2.20). In
the third system, hinge couplings (no bending moments) are used to take account of
the lack of bending stiffness in a fully tensioned member.
There are only very small differences for Models 1 and 2 (Figure 2.28). But, Model 3
results in totally different member forces. The left span tends to become a cantilever
beam. The bending moment over the intermediate support becomes positive.
Figure 2.28 Member forces of a two-span beam for three different models used for the opening
–147
10 kN/m
opening 20/50 cm
System 50
5.0 m 5.0 m
20
Model 1 Model 2 Model 3
–76.5
30.8
20.8 (21.1)
(20.7) 33.0 49.2
85.0
–15.0
–50.0 –50.0
(–49.8) Model 2 (–50.3)
–70.9
36
Figure 2.29 Deformation of girder in the opening region (plane shell analysis)
20 kN/m
Detail
It should be pointed out, however, that Model 3 is an extreme case, presented just to
demonstrate the importance of the correct manner of modelling the stiffness in the
region of the opening. Parametric studies with different element stiffness have to be
conducted to find the most realistic distribution of member forces for this system.
Figure 2.29 shows the deformation of the girder in the opening region. Great shear
deformations can be observed, which can hardly be modelled by beam elements. The
horizontal stresses are almost linear in the opening region (Figure 2.30).
37
Footing
The friction between the foundation and the ground can be modelled by horizontal
springs. Linear as well as nonlinear spring characteristics can be used. A nonlinear
analysis is required if certain load arrangements show tension stresses in the ground,
which cannot occur in reality due to uplifting of the foundation beam.
The results of the model (a–c) are identical if the length or the spacing of the elements is
sufficiently small. In practice, the advantage of using special interface elements instead of
individual springs is that the normal stiffness does not depend on the length of the
element for a given soil modulus. For individual springs, the stiffness depends on the
spacing of the elements (Figure 2.31a). Continuous bedded elements are used in elastic
supported foundation slabs, where the effects of the bedding can be directly introduced
into the element stiffness matrices (see Section 4.10).
All of these methods, except the continuum model (d), are based on a linear relationship
between the local force and the local deformation of the soil (foundation or stiffness
modulus method). This method does not consider the shear stiffness of the soil. There-
fore, in general, the displacements and associated reactions of the soil and the structure
are not compatible (Figure 2.32). The error caused by this simplification can often be
38
Figure 2.32 Deformation of soil and structure – foundation modulus and constraint modulus
method
Settlement Settlement
neglected in practice. However, when designing a shallow foundation, one has to keep in
mind that a uniformly distributed load on an elastic bedded beam or slab does not result
in any member forces (see Figure 2.32). In such a case, it is recommended that studies
with 2D or 3D continuum models are made (e.g. constrained modulus method). But,
this usually requires a much greater effort. Please note that the considerable uncertainties
of soil behaviour cannot be overcome by using more refined numerical models.
When using distinct spring elements, the reaction of the ground is introduced into the
structures by single forces. Even at the free end of a foundation beam, a force is esti-
mated, which results in a shear force at the end of the beam (Figure 2.33). Therefore,
the calculated shear forces are only correct in the middle of each beam element. However,
for design purposes, the distribution of the internal forces can be smoothed out.
The discretisation, the number of elements per length, has considerable influence on the
member forces. This aspect will therefore be discussed in greater detail. As an example,
an elastic supported beam with rectangular cross-section (b/h ¼ 1.0/0.8 m) and a length
of 5.0 m is analysed under a uniformly distributed load and a single load at midspan.
First, a very coarse element mesh is used, having only two beam and three spring
elements (Figure 2.33). The spring stiffness is calculated for a constant influence width
of 2.50 m/2 ¼ 1.25 m for the outer springs and 2.50 m for the inner springs. As demon-
strated in Figure 2.33, even for a system under a uniformly distributed load, significant
bending moments and shear forces are estimated. As mentioned previously, due to the
assumption of the elastic modulus method, no bending moments and shear forces
should be estimated under uniform loads. In principle, the calculated member forces
correspond to a two-span continuous beam where the intermediate support has settled
downwards. With a refined element mesh, the results are more reasonable (see
Figure 2.34a). Here, the member forces and the displacement in midspan are plotted
against the number of truss and spring elements.
At first, one would suppose that more elements are required for a beam with a concen-
trated load than for one under a continuous load. This is not the case, as the results of a
parametric study demonstrate (see Figure 2.34b). Here, the member forces are plotted for
39
Figure 2.33 Beam supported on three elastic springs under uniformly distributed load
1m
Foundation beam
Cross-section
b/h =1.0/0.8 m
2.50 m 2.50 m
Bending moment (in kNm)
–1.7
14.8
14.8
24.3 25.7
–25.7 –24.3
Deformation
2–10 elements against the analytical values (see Figure 2.35). With regard to the bending
moment and displacement at midspan, six elements are sufficient. For the shear forces, a
more refined element mesh is required. The midspan deflection for both load cases is not
very sensitive to the number of elements.
As the previous calculations have shown, the length of beam elements with respect to the
distance of the springs is limited by the following requirements.
g The deformation of the structure and the resulting soil reactions must be modelled
with a sufficient degree of accuracy. The element length is limited by the shape of
40
Figure 2.34 Midspan deflection, shear forces and bending moments of an elastic supported beam
with increasing number of spring and beam elements
V M
[kN] [kNm]
Deflection at midspan
vz (max = 0.40 mm)
5.1 15
2.6 7.5
Shear force V
Bending
moment M
Number of
0
elements
2 4 6 8 10
(a) Uniformly distributed loads
Deflection at midspan
vz (max = 4.4 mm)
100
Finite element versus analytical
Bending moment M
Shear force V
results in %
F = 1000 kN
50
Number of
0 elements
2 4 6 8 10
(b) Concentrated load at midspan
the structural deflection curve and the form functions of the used truss and
interface elements. If the foundation has a small bending stiffness, as in the
preceding example, the deflection curve can be described by using only a few
elements.
It is generally recommended that the length of a beam element l depend on the so-called
‘characteristic length’ L. The following calculation provides the characteristic length for
41
Figure 2.35 Deformations and member forces of an infinite elastic bedded foundation beam under
concentrated load
F
1.0
0.8
Deflection x
0.6
Shear force
0.4
Bending moment
0.2
x/L
0.0
1 2 3 4 5
–0.2
Deformation y(x):
F · L3
Ec · Ic · y(x) = · e–x/L · [cos(x/L) + sin(x/L)]
8
Bending moment:
F · L –x/L
M(x) = –Ec · Ic · y n (x) = ·e · [cos(x/L) – sin(x/L)]
4
Shear force:
V(x) = –Ec · Ic · y m(x) = –0.5 · F · e–x/L · cos(x/L)
where:
g The distribution of the member forces must be modelled with a sufficient degree of
accuracy. The correlation between the element length and the shear force
42
distribution results from the element type used in the numerical analysis. In the
case of no internal element loads, they are based on the assumption of a linear
shear force distribution within each individual element. Refinement of the element
mesh results in a more accurate description of the actual maximum shear force
under the concentrated load.
Elastic bedded beams have large bending moments under concentrated loads. These high
values are usually not relevant for the design, as point-loads do not exist in reality. All
forces act on a distinct load area. A load distribution with an angle of 458 can be assumed
up to the centreline of the beam or slab (see Section 4.10). In the example, this results in a
load area of greater than 0.80 m, which would reduce the maximum bending moment by
20%. The loaded area has to be modelled by at least two elements to describe the shear
force distribution with a sufficient degree of accuracy.
In the case of a linear elastic analysis based on the bedding modulus method, one has to
remember that tension forces cannot actually occur in the ground due to uplifting of the
foundation beam. The bedding of the beam has to be neglected in those regions where the
numerical calculation shows any uplifting of the foundation beam. In such a case, load
case combinations are not permissible, and this will significantly increase the amount of
calculation effort.
50
5.0 m
43
Figure 2.37 Member forces and settlements of a foundation beam with linear elastic and nonlinear
behaviour of concrete (max load q ¼ 1000 kN/m)
320
without tension
520
462
453
Vertical deflections (in mm)
–2.0
3.1
4.7
6.2
7.5
Longitudinal reinforcement (in cm2)
14.1
23.3
19.6
The member forces, the settlements and the required reinforcement are shown in
Figure 2.37 for a linear elastic as well as the nonlinear material behaviour of the
reinforced concrete beam. For nonlinear calculations, the amount of reinforcement
has been fixed by the results of the linear analysis. A linear elastic analysis results in a
maximum bending moment under the wall of m ¼ 520 kNm/m and a maximum displace-
ment of w ¼ 4.7 mm. However, the maximum bending moment is reduced by 20–25%,
and settlements are increased by 32%, when a nonlinear behaviour of the reinforced
concrete beam and its resulting stiffness reduction in the region of the maximum bending
moments are considered. Figure 2.37 (right) shows the results of the nonlinear analysis,
44
Figure 2.38 Bending moment and settlements with increasing load (nonlinear concrete material
model)
Settlement m in kNm/m
in mm v in kN/m
Bending moment
10 at midspan
500
q = 500 kN/m
q =1000 kN/m
q =1500 kN/m
q =1750 kN/m
both with and without taking the tension-stiffening effect (see Section 2.11.3) into
account. It can be seen that neglecting the tension-stiffening effect, which gives the
minimum stiffness, results in smaller bending moments. Therefore, this effect has to be
considered in the design of foundation beams and slabs.
Figure 2.38 shows the distribution of the bending moment at midspan and the settle-
ments of the strip foundation under an increasing load. At the beginning, there is a
linear relation between the load and the vertical displacements and bending moment,
respectively. The beam is uncracked, and its behaviour is fully elastic. At an approximate
load of q ¼ 1400 kN/m, the elastic load-bearing capacity of the critical section at the
midspan is reached, starting the formation of cracks. A plastic hinge develops. In this
area, the bending moment can only increase slightly. Any further increase in load-
bearing capacity can only be possible for a load redistribution to the less stressed sections
and a concentration of the soil pressure under the wall. This results in an excessive
increase of the settlements of the beam. It should be considered in the design that a
load increase with a safety factor does not affect the maximum bending moments
but will lead to considerable increase in the settlements. The shear force at the relevant
45
section near to the wall does not depend on the stiffness of the system, but from the
equilibrium condition.
One must distinguish between bored and driven piles. Due to their great diameter and
reinforcement, bored piles can carry normal forces and bending moments. Driven
piles have small bending stiffness and, due to their slenderness, can only carry vertical
loads.
In the following example, only bored piles will be discussed, as the vertical normal forces
in driven piles can easily be estimated from the equilibrium conditions (see Figure 2.44).
Rigid piles can be modelled by linear elastic supported truss elements. The bedding
modulus ks and the stiffness of the horizontal springs may vary along the length of the
pile and its circumference. According to Timm and Baldauf (1988), the distribution of
ks along the length of the pile respectively the exponent n should be chosen as follows
(see Figure 2.39):
Figure 2.39 Bored pile – numerical model and distribution of bedding modulus ks for a horizontal
force at the pile head
M
V H
n=0
z
Pile
Horizontal
Truss n = 0.5
springs
elements
n=1
n=2 ks
Vertical
spring ks(z) = ks(d ).(z/d )n
46
If there are no results available from actual pile tests, the bedding modulus ks may be
estimated by the following expression from DIN 4014 (1990):
ks ¼ Es/d
where:
ks is the bedding modulus
Es is the stiffness modulus of the ground
d is the diameter of the pile d 4 1.0 m.
The stiffness modulus for non-cohesive soils varies between Es ¼ 100–200 MN/m2 for
gravel and Es ¼ 10–100 MN/m2 for sand. The horizontal support in the upper region
of the pile by the ground should only be included in the design if it can always be
guaranteed during the whole lifetime of the foundation.
It is rather difficult to choose the adequate soil properties. This problem will not be
discussed further here. It should only be noted that, in every design, the distribution
of the member forces and the displacements are considerably influenced by the behaviour
of the soil.
The modelling problems will be further considered for a bridge column on a pile founda-
tion (Schornbachtalbrücke; see Becker, 1994) (Figures 2.41–2.43). Figure 2.40 shows the
dimensions of the structure. The bridge column is founded on 14 reinforced piles, each
having a diameter of d ¼ 61 cm. In order to simplify the numerical model, the pile
inclination and the enlarged footing are neglected.
Loading
Only a horizontal braking force of H ¼ 870 kN, acting at the top of the bridge column in
the horizontal ( y-) direction, is considered. This results in a bending moment at the
bottom of the pile cap of:
M ¼ 870 kN . 15.8 m ¼ 13 750 kNm
Manual analysis
The normal forces in the different piles can be calculated from equilibrium conditions if
one neglects the deformation of the structure (rigid pile cap) and the bending stiffness of
the piles. This results in normal pile forces of F2 ¼ 585 kN for the outer row and
F2 ¼ 195 kN for the inner row (Figure 2.44). No bending moments are estimated in
this approach.
Truss system
The piles are modelled with truss elements that are supported horizontally by spring
elements (see Figure 2.46). A linear distribution of the bedding modulus along the
length of the pile with ks ¼ 0 MN/m2 at the head and ks ¼ 100 MN/m2 at the pile toe is
assumed. The interaction between the individual piles and the friction between the
piles and the ground is neglected. The vertical settlement of the pile toe is modelled by
linear elastic springs. For simplicity, a constant cross-section of the column is used in
the following, since only the pile foundation is of interest in this example.
47
3.6 m 1.6 m
16:1
6:1
3m
50:1
50:1
11 m
1.8 m
bored piles
d = 61 cm
5:1
15 m
The pile cap cannot be modelled by truss elements, as this is a typical discontinuity
region (see Figure 2.45). Therefore, the nodes of the pile’s heads are fixed to the
lowest node of the column base. The pile cap is, therefore, modelled as an infinite
rigid body. The disadvantage of this model is that the member forces of the pile cap
are not calculated. Further investigations are required if the bending deformations of
the pile cap cannot be neglected.
48
49
The stiffness of the vertical spring has a great influence on the bending moments (Figure
2.48) and the horizontal deformation (Figure 2.49) of the piles. A fixed vertical support
reduces the greatest bending moment by a factor of 2 as compared with that from an elastic
support with C ¼ 400 MN/m. The main reason for this big difference is the rotation of the
pile cap. The greater the inclination of the infinite stiff pile cap due to settlements of the piles,
the greater is the rotation of the pile heads and the resulting bending moments.
F1 = 3.0F2
10.F1.2.25 m + 4.F2.0.75 m = M = 13 750 kNm
F1 = 13 750/23.5 = 585 kN
1.80 m
F2 = 585/3 = 195 kN
M = 13 750 kNm
M = 13 750 kNm
H = 870 kN
–F1 –F2 F2 F1
2 piles –F1 –F2 F2
5 piles F1
in a row
1.50 m each
1.50 m each
50
Fv
1.80 m
Fc
Ft
5 5
A
pile column
6 6
7 7
8 8
Section A–A Section B–B
9 9
10 10
Coupling
11 11
of nodes
100 200 300 400 100 120 140 160 180 Section C–C
Y X Y X
Vertical spring
Z Z
51
Vertical load Q
srg
Q(s)
Qr(s)
Qs(s)
sg = 0.1D
Qrg Qsg Qg
Qr load carried by friction
Qs load carried by pile toe resistance
D pile diameter
Figure 2.48 Bending moment distribution in the pile (load: Hy ¼ 870 kN at column head)
68
48
37
V V V
M M M
H H H
–2
–3
–15m
52
Figure 2.49 Horizontal deformation of the pile (load: Hy ¼ 870 kN at column head)
V V V
M M M
H H H
–15 m
When comparing the bending moment distributions, one has to remember that a pile
usually has a uniform reinforcement arrangement around its circumference. Therefore,
the () sign of the bending moments does not matter.
The distribution of the normal forces is not shown in this example, as it is only slightly
influenced by the stiffness of the vertical spring. The normal forces in the piles can easily
be calculated manually from the bending moment at the base of the bridge column
(Figure 2.44).
The main difference between a manual and a numerical analysis is that, in the
latter, bending moments and normal forces are estimated. Bending moments may
cause a high increase of the pile reinforcement due to the small lever arm of the internal
forces.
In this example, the displacement and rotation of the pile cap does not significantly
increase the member forces of the column. The horizontal deformation of the column
head is less than 18 mm (C ¼ 400 MN/m) or 8 mm (fully restrained), assuming an elastic
material behaviour.
53
Figure 2.50 Bending moment distribution in the pile with resp. without horizontally fixed pile cap
(load: Hy ¼ 870 kN at column head)
+8 195
68
Pile cap fixed in
horizontal direction
V
M
–22
Pile cap not H
restrained
V
M
C = 400 MN/m
C = 400 MN/m
Please note that long-term settlements of the soil underneath the pile cap may reduce the
vertical and horizontal bedding of the pile cap. Thus, a bedding of the pile cap by the
ground should be handled with great care.
Inclined piles
If the inclination of the outer row of piles is considered in the numerical model as in
reality (Figure 2.40), the maximum bending moments are estimated at the pile heads
(Figure 2.51). Again, there are large differences in bending moments between those of
fully restrained and flexible supported pile toes.
54
Figure 2.51 Bending moment distribution in the edge piles (load: Hy ¼ 870 kN at column head)
C = 400 MN/m2
4
10
Mz
Pile Column
Hy
23 24
To summarise, the vertical restraint, the elastic bedding of the piles and their distribution
along its length and any horizontal restraint of the pile cap have significant influences on
the member forces in the structural system. All these parameters depend on the soil char-
acteristics, which in reality scatter within a great range. Therefore, parametric studies
may be required to estimate the correct relevant design forces.
As the restraint of the column due to the foundation shows less scatter, it may be useful
to estimate the required design forces of the column by a separate model (i.e. a truss that
is partially restrained at its lower end). The resulting forces can then be used as loads for
the foundation.
A linear increase of the bedding modulus over the length of the piles has been assumed in
the preceding example. In general, the restraint of the soil against horizontal deformation
is not proportional to ground reactions. Therefore, such nonlinear behaviour has to be
considered in case of large horizontal loads.
Please note that one must check that the calculated soil pressure is not positive (tension)
or greater than permissible (passive earth pressure).
55
In the case of large bending moments, a nonlinear design of the piles (including the
reduction of stiffness due to cracking) can result in a significant reduction of the
maximum member forces. However, it is doubtful whether such a refined model is
more accurate than a linear elastic one, as the basic input parameter, the soil stiffness,
can show a great scatter in reality.
In a frame model, the structural system is modelled by straight truss elements that are
located at the centreline of the cross-section of the individual members. Modifications
are required for massive structural parts, such as the columns shown in Figure 2.53.
In addition, it should be noted that truss models are only valid for slender shear walls
(beam system) due to the underlying assumption of a linear strain distribution.
Further investigations may be required for areas near the supports at the base of the
structure.
The member forces of the horizontal ‘beams’ and vertical ‘columns’ are required for the
design of the structure. In this case, one must consider the deformation behaviour of the
whole structure. The different parts of the structure are modelled by straight truss
elements. Special attention has to be given to the joints between the horizontal beams
and the vertical ‘columns’. The horizontal beams are clamped at the inner face of the
vertical ‘columns’ and not at their centrelines. If this is not considered, the span length
would be much too large, and the calculated stiffness of the structure would be smaller
than in reality. There are various possibilities to model the real behaviour of the structure
(Figure 2.53).
56
Figure 2.52 Bracing elements of a high-rise building – frames and coupled shear walls
beam
wall
wall
57
Figure 2.53 Models for a coupled shear wall system with large openings
h1 l2 h1
EIR
EI = ¥
(l2 + h1 )3
EI* = EIR
l 23
b
(l2 + h1 )
EA* = l EAR
2
EIR
c
Rigid coupling
of inertia of the new beam is calculated as though it has the same bending stiffness
as the real beam, but with a shorter span length. It must be noted that the
deformation of the horizontal beam is not calculated correctly.
g Model c – coupling of the nodes at the joints: Nodes in the joints are coupled
together (see Figure 2.53c). This method has been previously explained in Section
2.1.
g Model d – 2D shell model (diaphragm): Shell models will be discussed in Chapter
3. Considerable effort is needed to evaluate such a model. It should be noted that
the nonlinear material behaviour in different regions of the structure can hardly be
modelled. Furthermore, the software cannot usually estimate reinforcement
requirements. This is the reason why membrane models are not widely used in
practice.
58
Figure 2.54 Models for a coupled shear wall with large openings
EI* EI = ¥ EIR EI = ¥
EI* EIR
EI* EIR
EI* EIR
The major difference of these models will be examined for the system shown in Figure 2.55.
The walls have a thickness of t ¼ 20 cm. They are made of concrete-grade C35/45. The
structure is loaded with a uniform horizontal load of q ¼ 10 kN/m. The deflection pattern,
which has been calculated with a shell model (diaphragm), is shown in Figure 2.56. A large
shear and bending deformation in the upper horizontal beams can be observed.
Table 2.4 lists the member forces in a few sections of the various models. Figure 2.57
shows the member forces and the deformation of the shear wall calculated with a truss
model and coupling of nodes at the joints. Models b and c give nearly the same results.
Variations of the axial stiffness (E . A) only slightly affect the member forces ( 5%). As
expected, the unmodified frame system results in much greater displacements, greater
bending moments and less normal forces than the ones of a 2D shell system.
The resulting member forces of a shell model are provided in column 7 of Table 2.4.
There is a good agreement in the results of the truss system. Therefore, the considerable
effort required for a plane shell analysis does not seem to be justified for this particular
structure.
The calculations shown earlier are based on a linear elastic material behaviour. The
maximum tensile stresses in the horizontal beam are greater than the mean tensile
strength of the concrete fctm (max M ¼ 41.2 kNm, section E–E, ct ¼ M/W ¼
41.2/8.33 ¼ þ4.9 MPa), resulting in cracking of the horizontal girders and a reduced
stiffness. This aspect must be considered in design if more precise data for the deforma-
tion of the structure is required. However, for the preceding example, the influence of the
stiffness of the horizontal beams on the member forces is small.
59
E F
33.75 m
0.50 m
C D
C D
A A B B
It should be noted here that the accuracy of a truss system in the lower region depends on
the support conditions. A nonlinear strain distribution is likely to occur. If more precise
information is needed, then a shell model (diaphragm) may be helpful.
In the preceding example, special beams are used to connect the two vertical elements
(‘columns’) together. Considerable construction effort would be needed to build such
beams. Therefore, they may be avoided where possible, and the existing slab may be
used only as a coupling element. However, this results in a complicated 3D shell
system that requires much calculation effort. To avoid this, a truss system may again
be used. First of all, the cross-section properties of the equivalent, fictitious horizontal
beams are needed. The effective width beff can be taken from diagrams provided by
Wong and Coull (1980). The figures presented by them are only valid for simple regular
systems, which are rarely built in practice. If more accurate values for the stiffness of the
horizontal beams are required, then it is better to use a 3D shell model with a simplified
equivalent structural system.
60
Nowadays, whole buildings can be modelled by 3D shell elements (see Chapter 6).
However, this tremendous effort is not justified for ordinary structures. A truss
system, where the whole structure is reduced to just the main load-bearing elements,
the so-called ‘cores’, is sufficient for most cases in practice. In the real system, the bracing
elements are connected together by slabs, which are assumed to be rigid in their
midplane. Also, the bending stiffness of the slabs is neglected.
Simple regular systems with a congruent deformation pattern of the bracing elements
can easily be designed with the so-called method of fictitious bars (equivalent beam
method; see Figures 2.58, 2.60 and 2.61) (Beck and Schäfer, 1969). This simple
method is often used to check numerical results, even for complex bracing systems. If
61
Table 2.4 Member forces of a shear wall with large openings (Figure 2.55 and Figure 2.56)
Sections (see Without Model 3 with Model 3 with Model 4 Model 5 shell
Figure 2.53) modifi- modified I* modified I*, A* coupling of analysis
cations (Figure 2.53b) (Figure 2.53b) nodes
(Figure 2.53c)
there are differences between the analytical and the numerical analysis, it is always a
question of which are the correct member forces. Therefore, in the following text, the
differences between manual and numerical models are discussed.
In reality, often complicated bracing systems are built like, for example, girders that have
different cross-sections over their length or columns that are not continuous throughout
the height of the building. In such cases, a manual calculation using the method of
fictitious bars is not possible. A numerical analysis will then be required for a 3D truss
model of the bracing system.
In addition, a numerical model can be useful where nonlinear material behaviour has to
be considered, or when a dynamic analysis is required (i.e. earthquake effects).
Even in this day-to-day task, engineering knowledge is needed when using design soft-
ware. There are several problems that may arise when using a truss model, such as
62
Figure 2.57 Member forces and deformation of a shear wall with large openings, nodes at the
column/beam joints coupled (see Figure 2.53c)
9 × 3.75 = 33.75 m
q =10 kN/m
63
Cores
Fictitious member
Slabs
based on the linear elastic behaviour of the concrete material. However, the reduction
of stiffness caused by crack formation in concrete (Stage II) has to be considered. It
is well known that the torsional stiffness of a cracked member is significantly lower
than the elastic value. Therefore, the torsional stiffness GIT may be neglected for a
concrete structure in Stage II condition. Rough data are given in ‘Heft 240’ of the
German Association for Concrete Design (Grasser et al., 1991). There, the following
values are listed:
Very often, the bracing elements have large openings (e.g. for the doors to the elevator
shaft), which may reduce the shear and torsional stiffness significantly. For beams
with open cross-sections or for hollow box girders with openings, the influence of the
floor slab has to be considered when calculating the stiffness parameters. An engineering
judgement is required here.
64
6m 6m 6m 6m 6m 6m 6m 6m 6m 6m 6m
7.5 m
26 m
9m
7.5 m
66 m
Further investigations are required if the bracing elements are located close to each
other. In the fictitious truss model, it is assumed that the slab can only transfer
normal forces and that it has no bending stiffness. This assumption is not valid if the
distance between the individual bracing elements is small. In such a case, the cores
may be modelled as one single stiff cross-section. An alternative is to introduce a stiff
coupling of the nodes. This will be explained in Section 2.6.3.
65
y 3
SS3
=Sm3
Sm1
Centre of gravity
1 of fictitious bar
Centre of torsion
SS1 of fictitious bar
YS
YM
zS zM
2 SS2
Sm2
z
W
Ssi = centre of gravity of bracing element i
Smi = centre of torsion of bracing element i
y 3
Centre of gravity
of fictitious bar
Centre of torsion
SS1 of fictitious bar
YS
YM
zS zM
SS2
z
W HZ ey
66
where:
Hy,M; HZM are the resultant horizontal forces, related to the centre of
torsion.
Mx,M is the resultant torsional moment, related to the centre of torsion.
The member forces of the individual bracing elements are estimated on a cantilever
beam loaded by horizontal forces Hy and Hz. Torsional effects of the individual
elements are neglected.
67
Figure 2.62 Rotation of a beam with U-section under pure torsional moment
Rotation around the axis of gravity Rotation around the axis of torsion
not correct correct
MT
J=
GIT
Axis of torsion
Axis of torsion
y1 Centre of y1
gravity
z1 z1
y 3
SS3 = Sm3
Sm1
SS2
2
Sm2
68
All nodes in a
level have same
Side view vertical deflection
The following different types of nodal coupling are used in practice (see Figure 2.65):
1. Bending stiff slab
vx ¼ vx0 þ ’y0 . (z z0) ’z0 . ( y y0)
vy ¼ vy0 þ ’x0 . (z z0) þ ’z0 . (x x0)
vz ¼ vz0 þ ’x0 . ( y y0) ’y0 . (x x0)
2. No bending stiffness, where the rotation of the nodes at the same level is not
identical
vx ¼ vx0 þ ’z0 . ( y y0)
vy ¼ vy0 þ ’z0 . (x x0)
3. No bending stiffness, where all nodes at the same level have the same rotation
where:
vx0, vy0, vz0 is the displacement of the reference node
’x0, ’y0, ’z0 is the rotation of the reference node.
Model 2 is used for comparison only. It does not reflect the real behaviour of the slab.
Whether Model 1 or 3 should be used depends on the behaviour of the actual system
with respect to the stiffness of the slab and the distance of the bracing elements. A
stiff coupling (additional to coupling without bending stiffness: rotation ’x ¼ ’x0,
’y ¼ ’y0, ’z ¼ ’z0) may cause restraints in the system, as can be seen from Figure 2.64.
69
Figure 2.65 Displacement and rotation of the bracing elements in case of a stiff slab
Plan view
(x4, y4)
y
4 Reference node
x (x0, y0) (x1, y1)
1
(x3, y3) 3
2
(x2, y2)
(x4 ,
y4 )
4 y
(x3 ,
y3 ) 3 x
Vy y
x Ref
e
(x0 , rence n
y0 ) ode
Vx
(x1 ,
vx = vx0 – jz0(y – y0) y1 )
vy = vy0 + jz0(x – x0) 2 1
jz = jz0
(x2 ,
y2 )
When the deformation and the rotation of all nodes in the same level are identical,
vertical loads are distributed according to the normal stiffness of an individual bracing
element. This does not reflect the actual behaviour of the structure. In reality, the vertical
loads are carried by the bracing element, which is close to the load.
The actual deformation behaviour of a slab is generally similar to a slab without bending
stiffness but which is infinitely stiff in its normal plane. In such a case, the rotation of all
elements around the vertical axis ’z must be identical at one level. Therefore, Model 3
should be used.
The axis of gravity of the different truss elements is coupled together, resulting in
torsional moments where the axis of torsion does not coincide with the axis of gravity
(e.g. cross-section with L- or U-shape). In this case, torsional moments and rotations
are estimated.
70
Plan view
40 m
qx = 0.442 kN/m 2
4.0
4 y 1
10 m
x
4.00
4.0
3 2
4.0 2.0
2.0
Elevation
3.0 3.0 3.0 3.0 3.0 3.0 3.0 3.0
qx = 4.42 kN/m
24 m
301
Truss model
401 901 201
101
4
925
Z 2
Y 1
X
member with U-shape. The slabs are modelled as mentioned before, using special
restraints with and without coupling of the nodal rotation ’z (’z is the rotation
around the vertical axis).
For the sake of simplicity of further calculations, it is assumed that all bracing elements
are fully fixed by the foundation slab. If the structure has a stiff box as a basement, it can
be assumed that the trusses are fully restrained at ground level, thus resulting in smaller
forces and displacements. A flexible foundation can be modelled by using spring
elements at the base of the beams.
71
40 m
4.0
qx = 0.442 kN/m 2
4
y 1
10.0 m
x
4.0
2
4.0
Manual
analysis
Z
4 2 1
3
–335 –725 –215
–19 –5 4.42 kN/m
Truss analysis Mt
vx,i = vx0 +..., My
Vy,i = vy0 + ...
4 1
3 2
Mx
My 0 –518 (+155%) –430 (–59%) –330 (+153%)
Mx –18 0 79 –62
4.42 kN/m
Truss analysis
vx,i = vx0 +...,
vy,i = vy0 + ... Mt = 0 My
jz,i = jz0
Mx
3 4 2 1
The distribution of the bending moments, from the method of fictitious bars, are
shown in Figure 2.67 for a uniformly distributed horizontal load of q ¼ 0.442 kN/m2
(total horizontal force is H ¼ 106.1 kN) in the x-direction.
72
Table 2.5 Support reactions (horizontal force and restraint bending moment) of the bracing
elements calculated with the method of fictitious bars
P
Element no. 1 2 3 4
The distribution of the total load between the different bracing elements is given in
Table 2.5. The bending moments at the supports are listed in Table 2.6.
Figure 2.67 shows the bending moment distribution from manual analysis (fictitious bar
method), and the values of the two different truss models. It can be seen that the coupling
of the rotation around the vertical axis ’z has a considerable effect on the member forces.
The bending moments in the bracing elements numbers 1 and 3 are more than 1.5 times
larger than for the system without coupling.
These large differences are caused by the torsional moments of element number 2, which
results from the distance of the axis of gravity and the axis of torsion (Figure 2.68).
Figure 2.68 Deformation of bracing element number 2 for different locations of centre of torsion
under uniform load
ex Mt = Fx · ey · Fy · ex
Fy
4m
Centre of gravity
Fx
4m
Centre of torsion
Centre of
y torsion
qx = 10 kN/m
Tip of beam
Concrete grade C20/25
73
Table 2.6 Bending moment My at the supports of the bracing elements (in kNm)
Element no. 1 2 3 4
Without coupling the rotation around the vertical axis, the transverse bending moments
Mx are small. In the case of coupling of the rotation ’z, the torsional moments of bracing
element number 2 are nearly zero. On the other hand, the transverse bending moment Mx
increases.
The small differences between the results of the equivalent beam method and the
numerical truss model are caused by different assumptions for each model. Within the
equivalent beam method, the bending stiffness of the bracing elements is considered,
whereas the torsional stiffness of the individual members is neglected. This simplification
is required in order to obtain a simple analytical solution. On the other hand, a numerical
calculation will always be based on the linear elastic material behaviour. An elastic
bending and a torsional stiffness are used in this example (Stage I – uncracked).
The assumption of linear elastic material behaviour should be checked in the case of large
bending and torsion action effects. As already mentioned in Section 2.6.1, it may be
necessary to neglect the torsional stiffness GIT if any section of the beam is cracked.
The results of a 3D shell analysis are used to verify the truss models. The whole structure
had been modelled with approximately 4410 shell elements (element size 1 1 m) (see
Figure 2.69). The columns in the real building are neglected.
Table 2.7 summarises the main results of the shell analysis. The load on bracing element
number 1 is increased by approximately 30%, and that of element number 2 is reduced
by approximately 13% relative to the values from the equivalent beam method. Overall,
there is a relatively good agreement with the results of the simple manual model.
In Figure 2.70, the deformation of the structure is plotted. The displacements are
increased by a factor of 10 000 thus resulting in a big differential deformation between
the bracing elements numbers 1 and 2.
In reality, the maximum displacements are approximately 1.1 mm in the x-direction and
0.2 mm in the y-direction only (concrete grade C35/40, linear elastic material behaviour).
The good agreement in the results of the three different models demonstrates that a
considerable effort to calculate the ‘correct’ member forces, for example, with a 3D
shell model, is generally not required. However, the time required for the 3D shell
74
analysis is more than 100 times than that needed for the simple truss model. More time is
needed for the discretisation, the verification of the results and for summarising the main
output. In addition, a computer program can generally not calculate the correct arrange-
ment of the reinforcement bars for a shell model, as will be demonstrated in Chapter 3.
One advantage of a shell model is that, not only the bracing elements, but also the slabs
can be modelled. Thus, the deformation and load-bearing behaviour of the structure is
modelled with higher accuracy. However, the calculated member forces of the slabs
cannot be used for the design, as the columns are neglected and the size of the plate
elements is too big.
In Table 2.8, the bending and torsional moments, shear forces at the base of the columns
and the deflection and rotation for a uniformly distributed load of q ¼ 0.442 kN/m2 in the
x-direction are listed for different numerical models.
75
Figure 2.70 Deformed structure (increased by a factor of 10 000) loaded in the x-direction
(q ¼ 0.442 kN/m2)
The various analyses of the given bracing system can be summarised as follows.
g The calculated bending moments of all truss models are greater than that of the
‘real’ structure (shell model number 6).
g There is a significant difference between the member forces of the various truss
models and the more realistic shell model number 6.
g The results are very sensitive to the modelling of the slab. Rigid nodal coupling
with ’z,i ¼ ’z0 should be used.
g Wrapping torsion can be neglected with regard to the simplifications of the truss
model.
g The support reactions are not very sensitive to the location of the beam axis as the
global deformation behaviour dominates.
In the case of manual analysis, the stability of the structure is checked with Equation 5.18
of Eurocode 2: Part 1 (2004), shown in the following text. Only if Fv,Ed becomes bigger
76
q = 1 kN/m2
40.0 m
y
S 4 = M4
10.0
4.0 S3 = M3
S2
1.11
M2
0.96
Columns: b/h = 40/40 cm Height of the building htot = 24 m
Slab thickness: h = 20 cm
Wall thickness: t = 30 cm
Models
77
Core no. 1
78
Model Shear force Vx: Bending moment Torsional moment Deflection Rotation
kN My: kNm MT: kNm ux: mm ’z
Truss model
2. No w.t. – centre of gravity* 318/265 4069/4357 7/197 0.8/1.9 0.36/1.8
3. With w.t. – centre of torsion* 346/404 4083/4864 42/302 0.8/1.0 0.34/0.57
4. With w.t. – centre of gravity* 352/412 4233/5025 42/308 0.8/1.1 0.35/0.60
Core no. 2
Model Shear force Vx: Bending moment Torsional moment Deflection Rotation
kN My: kNm MT: kNm ux: mm ’z
Truss model
2. No w.t. – centre of gravity* 182/194 2447/1741 3.6/54 3.1/4.2 0.36/1.8
3. With w.t. – centre of torsion* 140/46 2490/1171 3.0/24 3.5/4.0 0.34/1.8
4. With w.t. – centre of gravity* 134/44 2251/1029 3.0/21 3.1/3.5 0.35/1.5
Shell model
5. Shell model without columns 246 2217 5.8 2.6 0.28
Model Shear force Vx: Bending moment Torsional moment Deflection Rotation
kN My: kNm MT: kNm ux: mm ’z
Core no. 4
Model Shear force Vx: Bending moment Torsional moment Deflection Rotation
kN My: kNm MT: kNm ux: mm ’z
79
Finite-element Design of Concrete Structures
than the specific value, a more refined analysis, including second-order effects, is
required. The same holds true in the case of an unsymmetrical arrangement of various
bracing elements.
Stability parameter
P
ns Ecd Ic
Fv;Ed 4 k1 (Eurocode 2, 2004: Part 1, Equation 5.18)
ns þ 1:6 L2
where:
Fv,Ed is the total vertical loads on braced and bracing members
ns is the number of storeys
L is the total height of the building above the level of moment restraint
Ecd is the design value of the modulus of elasticity of concrete (Ecd ¼ Ecm/ cE)
Ic is the second moment of area of uncracked bracing member(s).
The stability of a building depends on the bending stiffness of the bracing system EcdIc,
the height of the building L and the total vertical load Fv,Ed. The latter value is the
maximum load under serviceability conditions, including the reduction of the live
loads allowed by Eurocode 2. The elastic modulus Ecd is given in the codes. Thus,
only the moments of inertia Icy and Icz are unknown.
In the case of regular bracing systems, the required cross-section parameter is the total
sum of the moment of inertia of the individual bracing elements in the relevant direction.
In non-regular bracing systems, the equivalent moment of inertia Ic may be estimated by
a comparison of the maximum horizontal deformation of a cantilever beam with the
results from the truss model loaded by an arbitrary horizontal force (Figure 2.71).
The calculation can be based on linear elastic material behaviour, as long as the
maximum tensile stress in the concrete is less than fctk;0.05/ c.
The torsional moments are estimated by assuming two bearings located at one support
axis. In such a case, it is recommended that each bearing be considered separately in the
numerical model, including the transverse spacing between them as well as their distance
from the centre of gravity of the beam. This results in a 3D numerical model instead of a
plane grillage system. The additional work is justified in this case, as the greatest
80
Figure 2.71 Estimation of the equivalent moment of inertia Iequiv. of a bracing system
Ec, Iequiv
l
Fh . l 3 Fh . l 3
f1 = f2 = Ic,equiv =
3 . Ec. Ic 3 . E c . f2
Figure 2.72 Structural models for design of a hollow box girder bridge
Beam
Uniform loads Wheel loads
part.restraint
Supports + Wheel
loads Top view
Member forces: Distortion of cross-section
– bending moment M
– shear force V
– torsional moment T
– support forces F
Support reactions as
loads on truss system
81
Beam model
1 21 41
Z X
Y
Support
Cross-section
Z
Y
Neutral axis
Support
Cross-section of
beam element
Z
X Support
Y
components of forces and deflections in each bearing are needed, and not only
the maximum total support reactions. As an alternative, the rotation around the
longitudinal axis may be restrained; however, this requires manual estimation of the
relevant bearing forces from the vertical load and torsional moments. A 3D model is
also required if a superstructure has an unsymmetrical cross-section.
The diaphragms at the support axis are modelled by stiff beam elements, or more
efficiently, by coupling of the nodes at the support. As an alternative, torsional restraints
at the supports can be considered.
Various load cases have to be considered in the design of a bridge structure. They have to
be combined in the most unfavourable manner. The relevant positioning of the traffic
loads, for example, axle loads, can be considered in two different ways. The first one
can ‘drive’ the traffic loads by the computer over the bridge in all different lanes. This
results in an enormous number of load cases and a major computational effort. In
addition, one has to know in advance which different parts of the structure should be
loaded to get the greatest member forces (see Figure 2.74). Therefore, numerical integra-
tion of influence lines for each node and each force can be used as a reasonable alternative.
82
Figure 2.74 Load arrangement by means of an influence line (train load UIC 71) max/min bending
moment at axis 3
max Ms3
The beam model is based on a rigid cross-section (Figure 2.77). Distortions caused by
unsymmetrical actions have to be considered separately by means of a frame system
(Figure 2.75). This system may be supported either in the vertical direction or in the
direction of the webs (see Section 2.1.2). For this model, the relevant member forces in
the webs and in the transverse direction are estimated for a longitudinal uniformly
distributed load.
Figure 2.75 3D shell model of a hollow box girder bridge (only one half of the structure is shown)
rt
ppo
Su
Line load over
the web
Cross-section in midspan
Undeformed
Z X
an
Z X Deformed Midsp
Y
Y
83
Figure 2.76 Distortion of a hollow box girder bridge at midspan; beam model (left) and real
deformations (right, bottom)
a
Deformed structure – elastic support
A slab system may be used to calculate the bending moments and shear forces caused by
single forces (e.g. wheel loads) and non-symmetric loads on the top slab (see Section
4.11.3). This slab is partially restrained at the webs (see Figure 2.77). The stiffness of
the equivalent bending springs can be estimated by comparison of the bending
moment of a beam and a plate structure under a uniform line load. As an alternative,
bending moments can be estimated from charts or influence lines, such as those of
Homberg (Homberg, 1973; Homberg and Ropers, 1965). Sometimes, an interpolation
of the values of different support conditions (fully or partially restrained at the web
supports) and the location of the single loads is required. However, such an analysis
may be more time consuming than an FE analysis. Further information on the analysis
of plate structures is given in Chapter 4.
A grillage system is still widely used, for example, in prestressed T-beam bridges. For
such structures, the results of a 2D plate or 3D shell analysis (see Figure 2.75) may be
‘nice looking’, but are generally of little use for a design where the resulting member
forces are required rather than the accurate stresses and membrane forces. Also, a
considerable amount of computation time is needed in order to consider all the relevant
load cases.
84
The example of a T-beam bridge is shown in Figure 2.78. This structure has been
constructed by prefabricated T-beams with an additional cast-in-situ concrete top
slab. However, before we consider the modelling of real structures, we will examine
the essential special features of a grillage system for a simply supported rectangular slab.
The solid two-way slab is represented by a 2D grillage system where the longitudinal and
transverse beams are connected at the nodes. The loading is always perpendicular to the
midplane of the slab. The following example only considers uniform loading.
85
Figure 2.78 Don Muang Tollway, Bangkok (Mühle and Kroppen, 1997)
20 in-situ concrete
8
12
23
9
36
42
Cross
beam
Elastomic
2.57 2.57 2.57 2.57 2.57 2.57 2.57 2.57 2.57 bearings
After the discretisation of the structure, one has to determine the equivalent vertical line
loads on the beams in both directions. This can be done by assuming that the displace-
ment of the beams at the nodes, where they are connected together, should be the same
(Figure 2.79). For midspan of a simply supported beam structure under uniformly
distributed loading q (q ¼ q1 þ q2 ¼ total load), this results in:
5 q1 l14 5 q1 l14
f1 ¼ f1 ¼
384 Ec Ic 384 Ec Ic
5 q2 l24 q2 l24
f2 ¼ f2 ðx ¼ 0:1l2 Þ 0:0041
384 Ec Ic Ec Ic
86
However, this load distribution only applies to one location. It follows from the deflected
condition (see the right-hand side of Figure 2.79), as well as the previous calculations,
that the equivalent load on the beams decreases from midspan to the supports,
respectively, in the direction of the shorter span length. For simplicity, this non-
uniformity is usually neglected in the analysis of a grillage system.
The accuracy of the analysis usually increases with the number of trusses in both
directions. In the following example, the influence of the number of beams on the
member forces and the displacements of a simply supported rectangular slab with an
aspect ratio of lx/ly ¼ 1.5 is examined (Figure 2.80). The structure is loaded by a
uniformly distributed load of q ¼ 10 kN/m2.
a) junction of the beams at midspan b) junction of the beams near the support
q2 q2
l1 l1
q1 q1
l2 l2
Girder 1 Girder 2 Girder 1 Girder 2
f
f
87
15 m
For the purpose of comparison, the displacement at midspan and the relevant
bending moments in both directions are calculated either by tables or by a 2D FE
model (shell).
Czerny (1999) gives the following values (concrete grade C30/35, slab thickness
h ¼ 30 cm):
Figure 2.81 shows the bending moments, the shear force and the displacements of the
grillage system with three girders in each direction. As would be expected, the load is
mainly transferred in the direction of the shorter span width. The bending moment
and shear force distribution show an unsteady distribution, which is caused by the
discontinuous interconnection of the beam elements.
The convergence of the grillage model is shown in Figure 2.82. Here, the ratios of the
results of the grillage system to that of the slab are given for an increasing number of
beams in both directions.
A system with one beam in both directions results in 43% (mxm) resp. 61% (mym) greater
bending moments compared to the slab. This large error can be traced back to the fact
that the 2D spatial load dispersion of a solid slab is not modelled by a girder grillage.
Even for this simple system, at least 7 7 beams are required in order to achieve suffi-
cient accuracy in the member forces. Figure 2.82 shows that, with more than 9 9
beams, the calculated results are lower than the correct values (factor <1.0), which
may lead to an unsafe design.
By means of a truss system, the forces in members can be calculated for slabs with
reduced twisting stiffness or orthotropic slabs. Slabs that have a reduced twisting
stiffness can only carry the loads in two orthogonal directions. The twisting bending
moment mxy cannot be sustained in the edges where two simple line supports meet.
Some examples of systems with reduced twisting stiffness are
g precast concrete slabs without additional cast in situ concrete cover, if a joint is
located closer than 0.3l from the corners (DIN 1045-1, 2008)
g slabs having large openings in the region of the edges
g slabs where the edges are not restrained against uplifting.
88
Figure 2.81 Bending moments, shear forces and displacements of the beams (3 3 division)
Bending moments
in kNm
372 97
118
280 543
312
248 161
251
479
134
Shear forces
in kN
–66
–32
156
–48 157
–26
73
115 322
Deflections
in mm
9.5 15.5
4.9 9.3
The reduction of twisting stiffness in the girder model is considered by reducing the
torsional stiffness of the beam elements. Figure 2.83 shows the results for the extreme
value IT ¼ 0 in relation to that of the elastic value IT for an increasing number of
beams in each direction. The bending moments at midspan increase to more than
80%, if the torsional stiffness is neglected. This factor is significantly higher than the
value given in ‘Heft 240’ of the German Concrete Association (Grasser et al., 1991).
89
Figure 2.82 Bending moment and deflection at midspan of a grillage system against the correct
plate values with increasing number of beams
1.9 my
1.8
mx
1.7
1.6 Bending moment m ym
Girder/plate ratio
1.5
1.4 Bending moment m xm
1.3
1.2 Deflection at midspan
1.1
1.0
0.9
1 3 5 7 9
Number of girders in each direction
Figure 2.83 Comparison of the bending moments and displacement at midspan with the plate
values (stiffness against twisting moments) for an increasing number of beams in both directions
with IT ¼ 0
1.9
Bending moment m xm
1.8
1.7
1.6 Bending moment mym
Girder/plate ratio
1.5
my
1.4 Deflection
mx
at midspan
1.3
1.2
1.1
1.0
0.9
1 3 5 7 9
Number of girders in each direction
90
The procedure for a girder analysis is demonstrated for the double T-beam bridge shown
in Figures 2.84 and 2.85. The numerical model is illustrated in Figure 2.87. The longi-
tudinal girders are represented by straight beam elements. Their axis is located in the
centre of gravity of the T-cross-section. Where the longitudinal girders have an
unsymmetrical cross-section, which is very often the case for the outer girders (see
Figure 2.85), the principal axis is not purely in the vertical or horizontal directions.
This results in a two-axial bending for the beam, which can only be considered by a
3D numerical model. In order to avoid this additional effort, the inclination of the
principal axes is generally neglected. This simplification is justified, since in most cases
the effect of the inclination of the main axis on the member forces and moments is
very small. Furthermore, the shear centre is placed in the centre of gravity of the
cross-section. The effective width of the flanges has to be eventually taken into account.
With these simplifications, the structural analysis of the bridge by using a flat truss
system (grillage) is possible.
For simplicity, a pin support in the centre of gravity of the main girders and a restraint
against torsion caused by the cross beams and the two separate bearings in each axis at
the supports are assumed in the system shown in Figure 2.86. Therefore, the distance
between the bearings and the distance of the support level to the centre of gravity has
to be considered when estimating the relevant bearing forces.
Cross-section
15
35
48
48
15
48
48
221
33 m 44 m 44 m 44 m 44 m 33 m
A B C D E F G
91
6.45 m
70
A B C D E F G
33 m 44 m 44 m 44 m 44 m 33 m
Z X
3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 Longitudinal
0.00
4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 beams
2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231
21 22 23 24 25 26 27 28 29 30
Y X
Ax
is
C
92
In case of torsion, the longitudinal girders rotate around their centre of torsion, which
in case of a symmetric T-beam is located at the junction of the centreline of the flange
and the web. It follows from this that a load on the flanges results in tension forces in
model A and compression forces in model B in the transverse beams (Figure 2.89).
However, since the torsional stiffness and thus the restraint against rotation of a
T-beam are small, the difference between the two models can be usually neglected. It
should be noted that, in case of model B, the end forces of the transverse beams are
introduced into the T-beam eccentrically, which results in torsional moments in the
longitudinal girder.
Cross-section
Model A
Longitudinal
Longitudinal girder 2
girder 1 Transverse girder (location)
Model B
93
Centre of gravity
Truss system
a) Transverse beam in centre of gravity of longitudinal girder
Equivalent beam
Equivalent beam
Rigid connection
Transverse beam
Longitudinal beams
94
Figure 2.89 Deformation of the longitudinal and transverse beams under constant line loading
Real system
Single or line load Single or line load
beff beff
Grillage system
a) Transverse beam at centre of gravity of the longitudinal girders
95
Figure 2.90 Member forces and deformation of a fully restrained beam due to vertical
displacement or rotation of one support
12 Ec I
VA ¼ VB ¼ w
l3
Thus, the equivalent moment of inertia is proportional to the span length powered by a
factor of 3.
If the bending behaviour of the deck slab is the dominant feature, that is, the longitu-
dinal girders rotate under external loading the unit rotation of the support ’ has to
96
Figure 2.91 Member forces and deformation of the equivalent structural systems due to a
displacement of the supports of w ¼ 0.10 m
Longitudinal Longitudinal
girder 1 girder 2
6.30 m
System
–3837 kNm h*=0.487 m
Bending
moment –4647 kNm
Deflections
0.1 m 0.05 m
Transverse beam with variable depth Transverse beam with constant depth
5.20 m 5.20 m
System 0.35 m
0.48 m
1.1m
Bending h*=0.482 m
moment –3835 kNm
Shear force
f –1475 kN –1475 kN
Deflections
0.1 m 0.05 m 0.1 m 0.05 m
97
q = 10 kN/m
Load case 1
q = 10 kN/m
Load case 2
apply. In this case, the equivalent moment of inertia is proportional to the span length
powered by a factor of 2
12 Ec I
VA ¼ VB ¼ w
l2
In case of inclined haunches, as in the chosen example, the support forces due to the unit
displacement w ¼ 1 and the unit rotation ’ ¼ 1 may be estimated numerically by a
plane truss system. As the bending moments are independent from the bending stiffness
of the equivalent system (support bending moment of a fully restrained beam under
uniformly distributed load M ¼ ql2/12), model A would result in correct shear forces
but with incorrect bending moments (Figure 2.91). Therefore, in such cases, the model
B is recommended, as the member forces in the longitudinal and transverse girders are
needed for design.
From the figures in Table 2.9, it can be seen that the results of both models are quite
similar except the transverse bending moment Mz, which is, from its definition, equal
to zero in the case of a plane truss system.
98
Figure 2.93 Member forces and deflections between axis C and D – 3D truss system [mm, kNm]
Bending moment My
–1741
–1707
–1269
–1297
760
570
01 .
–272
–292
134
148
4
193
95
–205
-
498
494
148
147
164
.
97.1
112
115
–165
264
164
163
Torsional moment MT
–74
–54
–419
419
54
–68
–36
37
69
Shear force Vz
–195
–259
261
197
–23
–36
37
23
1.6
. .. .
Deflections
1.2
. . .
–0.5
0.4
System Cross-section
33 m 44 m 44 m 44 m 44 m 33 m
Load case 1
A B C D E F G
Load case 2
99
Figure 2.94 Member forces and deflections between axis C and D – 2D truss system [mm, kNm]
Bending moment My
Load case 1
–1786
–1752
–1333
Load case 2
–1304
593
789
–284
–305
135
152
203
Torsional moment MT
–58
–401
402
64
10.51 .
–71
–65
65
71
Shear force Vz
–195
258
260
196
–35
–23
36
24
Deflections
1.24
.
1.6
. . . . . . . . . . .
0.81
.
–0.6
0.38
System Cross-section
33 m 44 m 44 m 44 m 44 m 33 m
Load case 1
A B C D E F G
Load case 2
100
Table 2.9 Forces in members and displacements of the relevant sections [mm, kNm]
The sawtooth shape of the member forces (Figures 2.93 and 2.94) results from the local
loading of the transverse beams. For dimensioning purposes, these values can be
smoothed.
The different parts of the load transfer should be first analysed. This is done for the
transverse beam in the middle of the span C–D. Figure 2.95 shows the member forces
and displacements of this beam.
Using the numerically determined deformation pattern of the transverse beam, the two
different parts of the shear force caused by either a vertical shift or a rotation of
the supports can be calculated by means of the expressions provided in Figure 2.90.
The resulting bending moment distributions are shown in Figure 2.95. For the two
load cases, the shear forces due to vertical shift and those due to rotation have similar
ranges and different signs. For this system, the equivalent stiffness of the transverse
beam consists of both parts of equal size.
As noted in the preceding text, the equivalent stiffness of the transverse beams
depends on different parameters and cannot be determined exactly. Therefore, the
influence of the stiffness of the transverse beams should be examined in more detail.
For this, the sectional height is varied from h ¼ 0.0 (no transverse beams) up to
h ¼ 0.60 m. The study of parameters is made by a plane truss system (grillage).
Figure 2.96 shows the bending moments at the supports Ms and at midspan of span
C–D Mf of longitudinal girder number 1 for the different depths of the cross-sections.
In Figure 2.97, the vertical support forces at axis C are plotted. This figure shows the
significant influence of the loading. An increase in the section depth only results in
minor changes in the bending moment, whereas its reduction causes considerable
increase in the member forces of the loaded girder.
101
Figure 2.95 Member forces and deflection of the transverse beam at midspan of span C–D
m t = 33.4 kNm/m
Load case 1 q = 10 kN/m Load case 2 q = 10 kN/m
Results of the truss analysis
–6.6
Bending moment –2.5
7.5 [kNm/m] 24.3
55 50 47 38
–48 –23
–41
–56
A B C D E F G
Grillage system
102
Figure 2.96 Influence of the section depth h of the transverse beams on the bending moments of
the loaded longitudinal girder
M in % A B C D E F G
60 h
Mf – Load case 1 mt = 33.4 kNm/m
40 Load case 2 q = 10 kN/m
Ms – Load case 1
20
0.6 0.8 1.0 1.2 h
0 in m
0.2 0.4
–20
–40
The peak values of the torsional moment decreases with the number and spacing of
the transverse beams. This is demonstrated in the T-beam bridge shown in Figures 2.98
and 2.99. An eccentric constant line load of q ¼ 10 kN/m at a distance of a ¼ 2.0 m from
the centreline of the outer girder (A) is applied. Figures 2.100 and 2.101 show the
member forces and displacements of the longitudinal girders of an inner span between
axes 3–4 for 10 and 5 transverse beams in each span. In the case of the 10 transverse
Figure 2.97 Influence of the section depth h of the transverse beams to the support forces at axis C
Fv in % A B C D E F G
120
Fv
100 Support force in axis C
Load case 1 q = 10 kN/m
80
Load case 2
60 h
mt = 33.4 kNm/m
40 Load case 2 q = 10 kN/m
Load case 1
20
0.6 0.8 1.0 1.2 h
0 in m
0.2 0.4
–20
–40
103
Cross-section of superstructure
Cross-section of 210
10 12
precast beams
153
1 2 3 4 5 6 7
beams per span, the greatest torsional moment is max MT,10 ¼ 51 kNm, whereas half the
number of transverse beams gives a value of almost double, that is, max MT,5 ¼ 92 kNm.
A further increase in the number of transverse beams would further reduce the calculated
artificial torsional moments. Consequently, the calculated torsional moments are only
the result of an inaccuracy of the numerical model. The considerable influence of the
distance between the transverse beams can generally be traced back to the fact that
the structure mainly carries the load on the cantilever slab by transverse bending of
104
the slab and not by torsion of the longitudinal girders. Next, the torsional stiffness of a
cracked member is very small.
It should be noted here that the design of the bridge not only has to consider its final
condition but also its construction stages. The bridge shown in the earlier figures is a
composite structure, consisting of precast T-beams and a cast-in-situ concrete slab.
During construction, the precast beams have to carry the whole loads, whereas the
cast-in-situ slab is nearly unstressed. Due to creep and shrinkage, load redistribution
between the precast beams and the cast-in-situ slab takes place. In addition, the
design in the transverse direction has to consider that the full height of the flange of
the precast girders cannot be taken into account. The joint in the flanges between the
precast longitudinal beams is not reinforced and is therefore not able to carry any tensile
forces.
A support on the lower side of a beam can be modelled by means of an additional node,
which is fixed to the node above in the centreline (see Figure 2.103).
105
Figure 2.100 Forces in members of the longitudinal girder between axes 3 and 4 due to an
eccentric line load of q ¼ 10 kN/m – 10 transverse beams per span
–399 –377
10 kN/m
Shear force V in kN
A
–73
72
.
Torsional moment MT in kNm
51
–51
Deflections vz in mm 2.5
106
Figure 2.101 Forces in members of the longitudinal girder between axes 3 and 4 due to an
eccentric line load of q ¼ 10 kN/m – 5 transverse beams per span
–431 –407
10 kN/m
Shear force V in kN
A
–88
86
92
–92
Deflections vz in mm 2.5
107
Undeformed
Deformed
Deflection
Small horizontal support forces occur due to pure bending if the fixations are located in
the centre axis, as there are no stresses or strains in the neutral axis. However, significant
horizontal forces Fh are obtained if the beam is fixed at the lower side in the horizontal
and vertical directions. For a single-span beam with a rectangular cross-section under
constant loading q, the restraint force Fh results in:
q l2
Fh ¼ ðh ¼ depth of the beamÞ
8h
Figure 2.103 Member forces of a beam with support in the centreline and at the lower side of the
beam
Fh = 0 q
Bending moment
q·l
Fv =
2
ql 2
Mmax =
8
q
2
ql
Fh =
8h
Bending moment ql 2
Mmin = –
q·l 16
Fv =
2 ql 2
Mmax =
16
108
The bending moment at the supports is equal to Ms ¼ Fh . h/2 ¼ q . l2/16 and the
midspan moment is reduced by 50% to MF ¼ q . l2/16 (see Figure 2.103).
However, the strong reduction of the midspan moment should be considered with great
care. The horizontal restraint forces Fh and thus the significant changes of the bending
moments of the beam can only happen in the case of an infinitely stiff support, which
in reality does not exist. Furthermore, the restraint effect is reduced by the nonlinear
strain distribution in the support region and the time-dependent deformations of
concrete (creep and shrinkage).
The following example should serve as clarification. A single-span beam with a span
length of l ¼ 40 m under a uniform load of q ¼ 45 kN/m is examined. The beam has a
rectangular cross-section with b/h ¼ 0.42/1.9 m.
q l 2 45 402
Fh ¼ ¼ ¼ 4740 kN
8h 8 1:9
The deformations at the lower side of the beams are as follows (Ec ¼ 30 000 MPa):
This reduces the restraint force Fh due to the uniform load q by 2/3.
109
38 –2
–100
–100
2.4 m
28
–300
–300
100
0 28
76 38
76
–2
0
0 458 458 458 458 0
6 × 2.75 m = 16.50 m
38 –2
–100
–100
–100
–300
0
2.4 m
28
–300
0 28 38
76
–2 76
0
Please note that the horizontal fixation may be overlooked when the analysis is checked
by a system plot only.
The following only refers to the ultimate limit state design (ULS), as the design
for serviceability can be very different according to the requirements of the relevant
codes.
110
Figure 2.105 Actions, strains and internal forces (resistance) of a reinforced concrete section
Actions Resistance
c2 –fcd
s2
a =k jx
x = h
As2 –F2d
MEd –Fcd
z = h
d
h
NEd
As1
s1
Fsld
c1
Strain Stresses and stress
resultants
g checking of the structural safety: Balance between internal and external forces
without exceeding the permissible stresses and strains of the materials
g estimation of the required reinforcement and its distribution in the beam under
consideration of the ultimate and the serviceability limit state
g estimation of the strains and stresses.
The nonlinear behaviour of concrete and steel as building materials has to be considered
in the design (Figure 2.106). In addition, concrete is not allowed to carry any
tensile forces in the ultimate limit state design. The tensile strength of concrete is only
used in the design of the serviceability limit state (crack width, restraint forces and
displacements) if it results in an unfavourable effect.
The design is based on the straight strain distribution over the depth of the cross-section
(Bernoulli’s hypothesis) in Stage I (uncracked) as well as in Stage II (cracked) conditions.
Figure 2.106 Stress–strain diagram of concrete and steel according to EC2 part 1
Eurocode 2, 2004
εc (<0) εs
111
Figure 2.107 Possible strain distributions under ultimate limit state according to Eurocode 2
Eurocode 2, 2004
(1 – c2 /cu2 )h
or
(1 – c3 /cu3 )h D B
As2
C
d
h
Ap p p(0)
As1 A
c
p, s
ud c2 cu2
0
(c3) (cu3)
A – steel tension strain limit
B – concrete compression strain limit
C – pure compression strain limit
The nonlinear material behaviour with respect to the stress–strain relation is defined
in the relevant codes (see Figure 2.106). It should always be kept in mind that these
curves are only approximations of the real material behaviour for use in design. The
parabola–rectangle stress–strain relation of concrete is only a simplification of
the stress distribution in the concrete compression zone, and cannot be used for other
design purposes. In reality, significant deviations may occur. The permissable strains
in the concrete (Figure 2.107) have to be considered in the ultimate limit design in
addition to the defined stress–strain relation of concrete.
For a more realistic estimation of the stiffness of a member, the tension-stiffening effect
has to be taken into account (see Section 2.4.2). Otherwise, the stiffness of the member in
the relevant section is underestimated. The tension-stiffening effect can be considered by
modifying the stress–stain relation of reinforcing steel in Stage II, for example, according
to Figure 2.108. From this figure, it can be seen that the tension-stiffening effect only has
a significant influence on the stiffness of a member when the steel strain has not reached
the point of yielding.
The dimensioning, that is, the estimation of the required reinforcement, can only be done
by iteration. This is true even for simple rectangular cross-sections under uni-axial
bending. For cross-sections of arbitrary shapes, such dimensioning is an optimisation
problem. The designer must default a reasonable distribution of the reinforcing bars in
the cross-section and give some information regarding the iteration process (i.e. how the
calculated reinforcement in the different locations has to be increased or decreased).
In addition to the reinforcement required for the ultimate limit state, the minimum
reinforcement and the ‘shift rule’ (horizontal displacement of the envelope line of the
112
Figure 2.108 Modified stress–strain diagram of steel with regard to the tension-stiffening effect
CEB-FIP Model Code 90
s
ft
fy
t(sr2 – sr1)
1.3sr
sr
s
sr1 sr2 smy sy smu su
total tensile force) have to be considered. The latter is done by increasing the local
normal force in the longitudinal tensile reinforcement by:
cot cot
Fsd ¼ VEd
2
where:
The shear and compressive stresses depend on the minimum width of the beam in the
tensile region. For polygonal cross-sections of an arbitrary shape, the relevant section
for the shear design has to be defined by the user of the computer program.
In addition, the influence of point-loads close to the supports, the variation of the cross-
section’s depth (inclined haunches), and the influence of an indirect support has to be
taken into account.
In case of torsion, a reduction in the resulting tensile force in the compression zone of the
cross-section can be applied. It should be noted that torsion reinforcement is only
required in case of equilibrium torsion, where the equilibrium of the structure depends
on the torsional stiffness of the structural members. As the torsion stiffness of a concrete
113
ad
an
m /t
=b
ü
sB
A
A
dk dk M
VEd T Ed
Rigid plate at the end
for load transfer MT
bk bk
member decreases significantly in case of cracking (Stage II), the minimum reinforcement
is sufficient for compatibility torsion. The calculation of the internal forces is usually
based on linear elastic material behaviour. Therefore, a computer program will always
determine an amount of torsional reinforcement. This problem can be overcome by
setting the torsional stiffness of the members to very small values.
Even if a nonlinear analysis can be easily done by means of the available software, it
should be kept in mind that such a calculation is still very time consuming and needs
114
A material nonlinear structural analysis can be useful or necessary for the following
reasons.
With material linear elastic analysis, it is supposed that a structure fails as soon as, in
any cross-section, the load-carrying capacity is reached. This applies only for statically
determined members. Statically undetermined structures for the most part show, by
redistribution of forces, considerable additional load-bearing reserves that can only be
considered in a materially nonlinear analysis. It should be pointed out that, in the
materially nonlinear analysis of concrete structures, the estimation of the member
forces and the design for the limit state is not sufficient proof of the safety of a structure.
One has to check that a brittle failure of a member cannot occur. This is done by limiting
the plastic rotation of a section. A brittle failure may happen when the concrete reaches
its limit strains before the reinforcement yields. The limited rotation of a concrete section
limits the theoretical value of force redistribution of an elastic member. The verification
for the rotation capacity of a concrete section is given in the codes.
The algorithms used in the calculation of the load-carrying capacity, based on the
nonlinear stiffness of any massive concrete beam or column under biaxial bending
with normal forces, have been known for a long time and are implemented in numerous
programs (Busjaeger and Quast, 1990). The required stress–strain relations for concrete
and steel can be taken from the codes (Figure 2.110). The simple parabola–rectangle
diagram for concrete should be used only for calculations in the ultimate limit state.
For the estimation of realistic deflections, as well as for nonlinear analysis, a more
realistic stress–strain relation, as given in Figure 2.110, is needed.
115
Figure 2.110 Stress–strain-curves according to EC2 (Eurocode 2, 2004) for concrete, cold-worked
reinforcing steel and prestressing steel
σc (<0)
40.0 fcm = fck + 8 N/mm2
–fc
Serviceability limit state
σc = [1 – (1 – εc/εc2)n] · fck nonlinear analysis
fck = 30.0
Ultimate limit
σc = [1 – (1 – εc/εc2)n] · fcd state
fcd = α · fck/γc
20.0
0.4fc
Concrete
10.0 fck = 30 N/m2
σs σp
ftk fpk
fyk fp0.1k
fyd Design
Reinforcing Prestressing
steel steel
εs εp
0.2% fyd /Es εud εuk 0.1% εuk
The stress–strain diagram of concrete can be estimated from the following equation.
c k 2
¼ (Eurocode 2, 2004; Part 1-1, Equation 3.14)
fcm 1 þ ðk 2Þ
where:
¼ "c/"c1
k ¼ 1.05 . Ecm . |"c1|/ fcm
Figure 2.111 shows the –" curve for a concrete-grade C30/37 for deformation and
nonlinear analysis. The meaning of the various compressive concrete strengths fcm, fcR
and fcm/ c is explained later.
Most programs can handle any polygonal stress–strain relation. The problem of a
material nonlinear analysis for beams and columns, therefore, does not exist in the
more-or-less complicated material description of the concrete.
116
σc (<0)
40.0 fcm = fck + 8 N/mm2
–fc with fcm
(for deformation analylsis)
fck = 30.0
In the ultimate limit state, the concrete tensile strength may not be considered. This
assumption considerably underestimates the stiffness of a beam. Hence, a realistic
estimation of the load-deflection behaviour of a structure is not possible with this
simplification. Realistic values can be obtained only if the tension-stiffening effect, the
stiffness of the concrete in tension, is considered. There is no generally accepted
procedure on how this should be done. In the following, two very different approaches
that lead to partially different results will be discussed.
According to Eurocode 2-1 (2004), the steel strains can be reduced by a certain level
(Figure 2.108). Another attempt considers a concrete tensile strength dependent on the
concrete strain at the level of rebar (Quast, 1981). It is not yet clear which attempt
gives more realistic results. It seems dependent on the structure and the loading.
117
50
40
Elastic modulus in GPa
20
10 Normal concrete
Lightweight concrete
0
0 10 20 30 40 50 60 70 80 90 100
Compressive strength fck in N/mm2
With a linear analysis of beams and trusses, there becomes only one material constant,
the modulus of elasticity of concrete Ec, which is needed. Nevertheless, this value in
reality is not fixed. It depends on numerous factors, such as the amount of cement or
the hardness of the gravel.
Figure 2.112 shows the measured modulus of elasticity Ec for concrete specimens in
relation to the characteristic compressive strength fck. The big scatter even in the elastic
state can be seen clearly.
Furthermore, Ec is not constant due to the nonlinear material behaviour of the concrete
even in uncracked conditions. Therefore, even a complex nonlinear analysis of a concrete
member can only give an approximation of the real deformations.
The modulus of elasticity Ec can be determined from the various stress–strain curves of
concrete. To avoid this effort, which involves high inaccuracy (see Figure 2.112), mean
values of Ecm are given in Eurocode 2 (2004) for different concrete grades. Ecm is the
secant modulus of elasticity between c ¼ 0 and c 0.4fcm (Figure 2.111).
Ecm ¼ 22 . ( fcm/10)0.3 ( fcm in MPa, Ecm in GPa)
(Eurocode 2, 2004: Part 1-1, Table 3.1)
Nevertheless, the secant modulus Ecm is often needed to estimate the deformation of a
member, because with an existing concrete stress c, the accompanying strain "c is
needed. The tangent elastic modulus Ec0m is to be used if the strain increment "c is
needed for a given stress increment c.
The differences between the tangent and the secant moduli are smaller than 15%
for normal concrete (Table 2.10). With regard to the scatter of material parameters, in
118
Table 2.10 Secant modulus of elasticity of concrete in GPa#Data taken from Eurocode 2, 2004
fck 20 25 30 35 40 45 50 55
fcm 28 33 38 43 48 53 58 63
Ecm 29.962 31.476 32.837 34.077 35.220 36.283 37.278 38.214
Ecm from –" curve 24.191 26.186 28.019 29.726 31.386 33.087 34.769 36.344
reality (Figure 2.112), these differences and any ‘accurate’ analysis for Ec seem to be
more or less of a theoretical nature.
The stress–strain curve of concrete is linear for compressive stresses smaller than 0.4fc,
as can be seen from Figure 2.113. Hence, it is possible in the serviceability state to
attach a triangular-shaped compression zone. With this simplification, the deflections
of a structure in the cracked state can be easily determined manually (see Section 2.11.2).
The deformation of a concrete structure increases with time due to the time-dependent
behaviour of concrete (creep and shrinkage). Creep deflections can easily be estimated
by means of a modified modulus of elasticity.
Ecm
Ec;eff ¼
1:0 þ ’ðtÞ
Nevertheless, this simplification is only valid if the distribution of the member forces
does not change. Shrinkage leads to curvature and deflections due to the bond between
reinforcing bars and concrete.
Creep parameters and shrinkage strains can be taken from Eurocode 2 (2004) or
from DAfStb 525 of the German committee on concrete structures (DAfStb, 2003).
Figure 2.113 Stress–strain diagram of concrete grade C30/37 – secant and tangent modulus of
elasticity
σc/fcm
0.50
σc (<0)
–fc
0.45 en
t
ng
Ta
0.40
nt
ca
Se
0.4fc
0.35
Detail
0.30 εc (<0)
Secant
0.25
Tangent
0.20
fck = 30 N/m2
0.15
0.10
arc tan Ec0m
0.05
arc tan Ecm εc × 10–4
0.00
0.0 –1.0 –2.0 –3.0 –4.0 –5.0 –6.0
119
Nevertheless, the big scatter of the creep factor should be kept in mind, with the
coefficient of variation lying at approximately 30%. Hence, the arithmetic values can
considerably deviate from the deflections appearing in a real structure.
The preceding explanations should merely make clear the difficulties with the definition
of the modulus of elasticity Ec, even in the uncracked, linear elastic state. In view of the
large variations in the material parameters in the construction practice, and in particular
of the creep factors as well as the very large influence of the cracking of concrete (State II)
on the stiffness of a member, complex or ‘exact’ analysis seems to be rarely required. For
the exact estimation of the real deflections, all influencing factors must be known.
However, this is only the case for the post-calculation of existing structures and if
these structures were examined before thoroughly.
The advantage in disregarding the tension-stiffening effect particularly lies in the fact
that the deflections of a reinforced concrete beam or column with rectangular cross-
section can be estimated in the cracked state rather easily by hand. Therefore, it is
possible to control the results of a numerical analysis. Furthermore, the calculated deflec-
tions are upper bound values.
For the following calculations, a triangular-shaped concrete pressure zone (see Figure
2.114) is assumed, and the tension-stiffening of the concrete is neglected. Furthermore,
only beams under pure bending are treated. The external normal force NEd is positive
for tension.
d ∆ls +NEd
h z
zs
+
εs Fs
σctm
120
As Es
l ¼ ; e ¼
bd Ec
The height of the compression zone is independent of the loading. The lever arm of the
internal forces in the cracked state is: zII ¼ d x/3.
Remark: The equation zII ¼ (1 0.6 . mEds) . d is valid only for "s ¼ 5% and for the
parabola–rectangle stress–strain curve, that is, in the ultimate limit state.
MEd
MEd ¼ Fs zII ¼ As Es "sd zII ¼) ¼ "sd
As Es zII
For the deflection calculation, the second derivation of the deformation w00 (x) of a
member is needed. This value is also called curvature 1/r.
1 "s
¼
rII ðd xÞ
121
r r
Fc
1 x 1 MEd
d–x d–x z +NEd
zs
εs Fs
εs
The height of the compression zone x and also the lever arm of the internal forces z only
depend on the constant factor (el). Thus, the distribution of the curvature is similar to
the distribution of the external bending moment if the height of the cross-section as well
as the reinforcement is constant through the whole member.
Now, the deformation of a member can be determined easily by double integration of the
curvature (Figure 2.116). Further details are available in Section 2.11.3.1.
ðð ðð ð ð
1 1 dx ¼ M M dx
w¼ w00 dx ¼ dx ¼ M
r r EI
∆li ∆li +1
Undeformed
beam
wi
wi +1
wi′
i wi +2
w i′ +1
i+1
Deflected
i+2 beam
122
w0iþ1 w0i 1 0 1
w00i ¼ ¼ ¼) wiþ1 ¼ li þ w0i
li ri ri
w wi
w0iþ1 ¼ iþ1 ¼) wiþ1 ¼ w0iþ1 li þ wi
li
If the deflections in a few sections are needed, it is easier to determine the required values
by means of the moment–area method (principle of conservation of energy). More details
are given in Section 2.11.3.1.
The length of the neutral axis is enlarged when the section cracks (Figure 2.114). This
effect must be considered in case of restrained forces.
The basics of such an analysis will be explained on a simple tensile member shown in
Figure 2.117. In the elastic, uncracked State I, the reinforcing bars show the same
(small) strain like the concrete if one neglects creep effects. Due to the different elastic
moduli of concrete and steel, the tensile stresses in the reinforcement gives: s ¼ e . c
(where e ¼ Es/Ec).
If one increases the tensile stresses, the first crack will open in the cross-section that
shows the smallest concrete tensile strength fct,av. In the crack, the whole of the external
force is carried only by the reinforcing bars. This results in the maximum steel strain and
stress. By contrast, the concrete tension is zero. Due to the bond between rebars and
concrete, shear forces are transferred from the reinforcement to the concrete on both
sides of the crack. Consequently, the tensile stresses increase in the concrete until they
reach the local concrete tensile strength fct,av again, and a new crack opens. This process
goes on with increase in the external normal force until so many cracks have opened that
the concrete tensile stresses cannot reach the tensile strength, due to the small bond or
transition length. This state is called the final cracking state. The concrete carries a
part of the external tensile force between the cracks and increases the stiffness of the
member with regard to the pure State II.
123
Figure 2.117 Distribution of tensile and bond stresses between two cracks
Crack
Rebar
ss(x)
Steel stresses
les
results can be verified by tests. Nonlinear calculations are useful for failure analysis
where all input parameters (loads, real material parameters, etc.) as well as the crack
pattern are known.
Hence, in practice, for nonlinear analysis, easy calculation models are required. Two very
different methods are available and in use. According to Eurocode 2 (2004), the steel
strain in the rebars for the pure State II can be reduced to consider the tension-stiffening
effect. This results in an increase of the stiffness of the structure. The size of the reduction
depends on the strain difference in the reinforcement "s ¼ "sr2 "sr1 shortly before and
immediately after the cracking, and is fixed in the code (CEB-FIB Model Code, 1990)
(Figure 2.118). The strain difference "s is big for slightly reinforced cross-sections,
because the tensile force in the concrete before cracking must be carried by the small
amount of reinforcement after the crack opens. This results in big steel strains "sr2. In
contrast to the procedure described in the following, the strain change "s is nearly
independent of the steel strain in the final cracked state (Figure 2.118).
124
Figure 2.118 Modified stress–strain curve for concrete steel to consider the tension-stiffening
effect
CEB-FIB Model Code, 1990
σs
ft
σsm ft – fy
σs2 = (εs – εsy)
σs2 εsu – εsy
fy
fy fy
σsm = εs + (εsy – 1)εsmy
εsy εsy
βt(εsr2 – εsr1)
1.3σsr 0.3σsr
σsr σsm = (εs – εsr1) + σsr
ε*sr – εsr1
fy σsr
σs2 = εs σsm =
εsy εsr1
εs
εsr1 εsr2 εsmy εsy εsmu εsu
εsy
ε*sr = 1.3σsr – – εsy + εsmy
fy
The equations to estimate the modified steel strains from DAfStb (2003) are listed in the
following text.
where:
"sm is the mean steel strain.
"uk is the ultimate steel strain.
"s1 is the steel strain in the uncracked state.
"s2 is the steel strain in the crack in the fully cracked state.
"sr1 is the steel strain in the uncracked state under the loads, where fctm is
reached.
"sr2 is the steel strain in the crack under the loads, where fctm is reached.
t is the parameter to consider the influence of the load duration or a repeated
load on the mean strain ( ¼ 0.40 for a short duration; ¼ 0.25 for a constant load
or for frequent load changes).
s is the tension stress in the tension reinforcement that is calculated on the basis
of a cracked cross-section (tension in the crack).
125
Stresses in MPa
–3.0
+ –2.0
–1.0
fct,R
2.0
Tension Compression
sr is the tension stress in the tension reinforcement that is calculated on the
basis of a cracked cross-section for the loads that lead to the first cracking.
d is the factor for the consideration of the ductility of the reinforcement ( ¼ 0.8
for high-ductile steel; ¼ 0.6 for normal-ductile steel).
Figure 2.120 Reduction of the concrete tensile strength as a function of the concrete strain in the
most strained steel fibre
I Region II III
1.0
0.8
Linear
Factor α
0.6
Parabolic
0.4
0.2
0.0
εcr Maximum concrete strain in the reinforcement layer εu
126
between the concrete and the reinforcement is strongly disturbed and, therefore,
according to Quast (1981), the concrete stiffening effect should be neglected. In the
region between these limit values, the maximum concrete tensile stress ctm is reduced
as a function of the concrete strain of the most stretched steel fibre with the value
(Figure 2.120).
This procedure shows a better correspondence with test results than a reduction of the
steel strain, as the calculation of 38 tests of concrete members (Quast, 1981) has shown.
In the calculation, the mean concrete tensile strength fctm will not be used but a tensile
strength fct,R. Quast (1981) suggests using the following values:
The deflections can be calculated by hand with the help of the curvatures of a member if
one assumes a constant modulus of elasticity Ec in the compression region and the para-
bolic-shaped curve of the concrete stresses in the tensile region is neglected (Figure
2.121). With the latter assumption, the increase of the member stiffness from State I
to State II is overestimated slightly (see Section 2.11.2). The equilibrium of the forces
and moments results in:
Force balance:
Fh ¼ 0 Fs þ Fct jFc j ¼ NEd ¼) "s Es As þ ctm b ðh xÞ
x
¼ "co Ec b þ NEd
2
Geometrical condition:
"s "co "co x
¼ ! "s ¼ ðd xÞ or "co ¼ "s
d x x x ðd xÞ
d ∆ls +NEd
h z
Fct
zs
+
εs Fs
σctm
127
Moment balance:
hx 2x
Fs z þ Fct þ ¼ MEds
2 3
Thus three equations are available to estimate the three unknowns, "s, "co and x. The
mean concrete tensile strength cmt depends on "s. The equations can easily be solved
by means of mathematical programs. Alternatively, the height of the compression
zone x and, with it, the curvature 1/rII, can be estimated iteratively. As a start value,
one chooses the height x without tension-stiffening.
Hence, an exact determination of the concrete tensile strength is only possible by tests in
the real structure. Any analytical estimation of fctm, for example, by means of the
equations given in the codes, will have big uncertainties. Hence, boundary value
considerations should be carried out if the concrete tensile strength has an essential
role on the distribution of the member forces and the deformation.
In Eurocode 2 (2004: Part 1-1 Table 3.1), the following mean and quantile values are
given:
ð2=3Þ
Mean value: fctm ¼ 0:30 fck for concrete 4 C50/60
As can be seen from the values in Table 2.11, the mean concrete tensile strength fctm for
normal concrete is approximately 10% of the typical concrete compressive strength fck.
fck 20 25 30 35 40 45 50 55
fctm 2.2 2.6 2.9 3.2 3.5 3.8 4.1 4.3
fctk;0.05 1.5 1.8 2.0 2.2 2.5 2.7 2.9 3.0
fctk;0.95 2.9 3.3 3.8 4.2 4.6 4.9 5.3 5.6
128
50 cm Concrete: C30/37
qd = 3.0 kN/m
the nonlinear analysis of reinforced concrete members that is absolutely necessary if one
wants to interpret and understand the results of numerical analysis. Manual procedures,
with which the deflections of a beam can be simply determined in the cracked state with
sufficient accuracy, are explained.
A 10-m-high cantilever beam that is loaded by a uniform horizontal load qd ¼ 3.0 kN/m
(Figure 2.122) is treated. The self-weight of the concrete member and the time-dependent
deflections due to creep of concrete are neglected. As the maximum concrete compressive
stresses under the given load are considerably smaller than 0.4fck, a linear stress–strain
curve is used for the concrete (triangular-shaped concrete compression zone). With
this simplification, manual calculations of the deflections are possible, and the arithmetic
procedures are easier to understand.
1. Elastic analysis
Horizontal deflection of the cantilever tip:
qd l 4 3:0 104
f ¼ ¼ ¼ 25:4 mm ð f ¼ 22:6 mm with Ec0m Þ
8 Ecm Ic 8 28 309 0:54 =12
2. Nonlinear analysis without tension-stiffening effect and with mean material
parameters
For simplicity, the curvature (1/r) at the support section will be determined only in the
following. The remaining values may be calculated accordingly.
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Height of the pressure zone: x ¼ d ½2 þ e l e l e l
where:
As 12:6
l ¼ ¼ ¼ 0:0055
b d 50 46
129
Concrete compressive stress: co ¼ "co . Ecm ¼ 0.00045 . 28 310 ¼ 12.7 N/mm2
With the known curvature, the maximum deflection of the beam can be easily calculated
by means of the principle of virtual work (Figure 2.123).
ðl
f ¼ M 1 dx ¼ 1 10 0:00405 10 103 ¼ 101:3 mm
0 r 4
If one considers that the concrete tensile stresses from a height of approx. 3.60 m are
smaller than the mean tensile strength (ct 4 fctm), and thus the concrete is uncracked
in this region in theory, the maximum deflection of the beam decreases to approx.
f ¼ 88 mm.
Figure 2.123 Structure, bending moments and curvature
M 1/r ‘1’ M
qd = 3.0 kN/m
10 m
130
The internal lever arm zII of a cracked cross-section with a triangular-shaped concrete
pressure distribution is independent of the loads.
Steel tensile stress and strain with the cracking moment Mcr in the State II (cracked):
Fsr 0:1438
sr2 ¼ ¼ ¼ 113:5 MPa
As 12:6 104
sr2 113:5
"sr2 ¼ ¼ ¼ 0:00057
Es 200 000
Steel tensile stress and strains with the cracking moment Mcr in the State I (uncracked):
Mcr 0:0604
sr1 ¼ e zs ¼ 7:07 4 0:21 ¼ 17:2 MPa
Ic 0:5 =12
sr1 17:2
"sr1 ¼ ¼ ¼ 0:00009
Es 200 000
The resulting stress–strain curves in the reinforcement are shown in Figure 2.124. In this
example, the reduction of the steel strain is low, about 0.1 mm/m, which results in a
change of steel stresses of around s ¼ 24 N/mm2. Therefore, the tension-stiffening
effect has little influence on the deformation of the beam.
131
600
fyR = 550
500
450
Stresses in σs MPa
400
0.0001
350
With tension stiffening
300
250
200
Without tension stiffening
150
100
σsr2 = 113
50
Strain εs
0
0 0.001 0.002 0.003 0.004 0.005
εsyR = 0.00275
The numerical integration of the curvature over the length of the beam results in a
maximum horizontal deflection of the cantilever beam of f ¼ 83 mm.
s ¼ 190 MN/m2
Within this method, a relatively high concrete tensile strength of ct ¼ 1.1 MN/m2 is used
in the support region. Hence, it is also not surprising that the maximum deflection of the
cantilever is only f ¼ 48 mm.
The considerably higher stiffness of the method of Quast compared with the procedure
using the reduction of the mean steel strain also becomes clear with the moment-
curvature curves (Figure 2.125).
Furthermore, the big change in curvature 1/r can be seen in Figure 2.125 when State I
remains. The method of Quast avoids this discontinuity. A reduction of the mean steel
132
160
120
With TS
100
Quast
80 Without TS
60
40
20
0
0 1 2 3 4
Curvature 1/r in mm/m
strain has only a low influence on the deflections, and with it on the curvature. In
Table 2.12, the main results of four models are listed.
The deformations of a structure are usually estimated with the mean values of the
material parameters fcm, fctm and Ecm. Hereby, one assumes that local damage or a
locally bad-quality material (e.g. by bad compaction of concrete) has little effect on
the member forces and the deformations.
133
40 fcm = 38 N/mm2
30
fcm/γc = 25.3 N/mm2
σc: MPa
10
C30/37
εc
0
0 –1 –2 –3
εs1 = –0.0023 εc1u = –0.0035
The required safety standards are reached through a decrease of the characteristic
material parameters and an increase of the characteristic loads. An analysis with these
design values would overestimate the deformations of a structure.
Hence, for a nonlinear analysis, the ‘real’ material parameters and stress–strain relations
for the building materials are needed to estimate the member forces and to design the
structure in the ultimate limit state with a sufficient safety margin. Two very different
approaches are given in Eurocode 2, Part 2 (2005) and DIN 1045-1 (2008).
With the global safety factor method, the ultimate resistance of a section Rd and the
member forces are estimated with the mean material parameters divided with a partial
safety coefficient R. The partial safety coefficient for permanent and frequent design
situations and fatigue analysis is R ¼ 1.27 (EC2) and 1.3 (DIN). For exceptional situa-
tions, R ¼ 1.1 has to be used.
1
Rd ¼ f ;f ;f ;f ;f
R cR yR tR p01;R pR
ftR ¼ 1.08fyR for reinforcement steel with high ductility (DIN 1045-1, 2008:
Equation 19)
ftR ¼ 1.05fyR for reinforcement steel with normal ductility (DIN 1045-1, 2008:
Equation 20)
134
fcR ¼ 0.85 . . fck for concrete 4 C50/60 (DIN 1045-1, 2008: Equation 23)
fcR ¼ 0:85 fck =c0 for concrete 5 C55/67 (DIN 1045-1, 2008: Equation 24)
The factor is equal to 0.85 for normal concrete. It considers the lower fatigue strength
of the concrete for dynamic loads as well as the difference between the cylinder and
uni-axial compressive strength. Hence, with short-term loads, for example, in the
ultimate limit state, bigger values of 0.85 4 4 1.0 can be used.
As can be seen from Table 2.13, the value of the concrete strength fcR deviates from the
mean value fcm, divided by the safety coefficient c ¼ 1.5 by up to 30%. Figure 2.126
shows the stress–strain relations for the different approaches.
It should be noted that the modulus of elasticity of the reinforcing steel Es should not
be modified by a safety coefficient (EsR ¼ Esm/ R ¼ 153.846 MPa; Figure 2.127). This
would result in a considerable reduction of the structural stiffness and significantly
higher deformations with regard to the mean values if the steel has not reached its
yielding strength. The yielding stresses of the reinforcing steel of both methods are
nearly identical, with fyk/ s ¼ 435 N/mm2 or fyR/ R ¼ 423 N/mm2.
With the reduction of the modulus of elasticity of the reinforcing steel Es comes an
increase in the deflection of the cantilever tip in the preceding example (with
qd ¼ Q . qk ¼ 1.5 . 3 ¼ 4.5 kN/m) by 24% to f ¼ 132 mm and f ¼ 164 mm (procedure
with reduction of the steel strains). With Quast’s method, a maximum cantilever
deflection is estimated as f ¼ 100 mm. The big increase in the deflection can lead to
135
σs: MPa
Member forces
fyR = 1.1 · fyk = 550 N/mm2 ftR = 1.08fyR = 594 N/mm2
fyk
= 435 N/mm2 500
γs Design ftR
fyR = 457 N/mm2
= 423 N/mm2 γR
γR
300
200
Es
=
Es
100
15
=
4
20
GP
0
a
GP
Strain εs in mm/m
a
0
0.0 1.0 2.0 3.0 4.0
2.2 2.7
The question of which procedure gives more accurate, realistic results is still an issue of
research. Nevertheless, for practical purposes, it is important to know that both methods
can give very different results, even for simple systems. This clearly demonstrates that the
nonlinear analysis of reinforced concrete structures is still very complicated, even with
the software available today. In this respect, it is clear why nonlinear methods are not
permitted for the design of the superstructures of concrete bridges in Germany.
Hence, if nonlinear procedures are used, it is advised to carry out boundary value
analysis and to check the sensitivity of the results for different input parameters.
A complex probabilistic analysis is rarely justified for a real concrete structure due to the
big effort and the missing statistical data. Thus, simple approaches for nonlinear analysis
are still needed.
REFERENCES
Beck H and Schäfer H (1969) Die Berechnung von Hochhäusern durch Zusammenfas-
sung aller aussteifenden Bauteile zu einem Balken. Der Bauingenieur 44: 80–87.
Becker M (1994) Erneuerung der Theodor-Heuss-Brücke in Heidelberg. Beton- und
Stahlbetonbau 89: 154–156.
Busjaeger D and Quast U (1990) Programmgesteuerte Berechnung beliebiger Massiv-
bauquerschnitte unter zweiachsiger Biegung mit Längskraft. Deutscher Ausschuss für
Stahlbeton, Heft 415, Berlin.
136
137
Chapter 3
Shear walls and deep beams
Shear walls and deep beams are thin 2D flat spatial structures that are loaded by forces
parallel to the midplane of the membrane (Figure 3.1). The stresses and strains are
uniformly distributed over the thickness. However, one must make a distinction between
shear walls and deep beams (Figure 3.2). Shear walls are continuously supported plane
members loaded by normal forces, where the maximum width of the cross-section is
greater than four times its minimum width. If this is not the case, a member is treated
as a column. Deep beams are plane spatial members whose height is greater than half
of their effective span width leff. Furthermore, these beams are not continuously
supported. In contrast to ordinary beams, shear walls and deep beams usually have a
nonlinear strain distribution over their depth (see Figure 3.3). Shear deformation
cannot be neglected.
The distinction between deep and slender beams is not only necessary for the calculation
of the internal forces, but also for the reinforcement arrangement. In ordinary beams, a
minimum shear reinforcement (stirrups) is needed, whereas deep beams only need the
minimum surface reinforcement.
This section only discusses the FE analysis of deep beams, as they are more often used in
practice. Shear walls can be designed in the usual manner similar to a column.
In accordance with Section 5.1.1 of EC2 (Eurocode 2, 2004), the internal forces in the
ultimate limit state can be calculated by using the following methods
Of these methods, the one that uses FE models based on linear elastic material behaviour
is most commonly used in practice. Comments on the evaluation of strut-and-tie models
(plastic analysis) are given at the end of this chapter.
The member forces of calculations based on linear elastic material behaviour can be used
for the design in the ultimate limit state and the serviceability limit state. The effects of
the redistribution of internal forces due to cracking of concrete or its ‘yielding’ under
139
y + (y/y) . dy
qy yx + (yx/y) . dy
qy dx dy t
xy + (xy/x) . dx
t x
dy
dy
qx dx dy t x + (x/x) . dx
xy
y
x
yx
t
dx y
dx
t
h Shear wall h Deep beam
lx l eff
Figure 3.3 Single-span deep beam – main membrane forces (left) and horizontal stresses in
midspan (right) for various heights h
h/l = 1.0
Stress distribution
acc. to Navier (linear)
h
h/l = 0.75
h/l = 0.5
h/l = 5.0
140
high pressures are neglected. Nevertheless, the nonlinear behaviour should be considered
in both the design and the detailing of the reinforcement. For example, the longitudinal
reinforcement of a single-span deep beam should be located at the bottom face of the
member, and should not be distributed according to the tensile stresses, which are
estimated by linear elastic analysis (Section 3.3).
The next section will provide the principles of an FE calculation for deep beams by a very
simple example of a single-span deep beam.
The results of the FE calculation are the node deflections vx, vy, the membrane forces nx, ny,
nxy and the stresses x, y, xy within the elements (see Figures 3.1 and 3.5). The membrane
forces are obtained by multiplying the normal stress x or y by the thickness of the wall.
The objective of the design of a concrete member is to estimate its required reinforce-
ment. For beams and slabs, this can be automatically produced by well-known algo-
rithms that are implemented in structural software. In the case of walls, a computer
program is usually unable to estimate these reinforcement requirements or its correct
distribution, as it is based on an element per element design. This problem will be further
discussed in Section 3.3. The reinforcement requirements of the single-span deep beam,
therefore, should be estimated by numerical integration of the horizontal tensile forces nx
over the midspan depth. For the structure in this example (load case 1), this results in a
tensile force of Fs ¼ 31 kN (see Figure 3.5). The lever arm of the compressive and tensile
141
Figure 3.5 Main membrane forces – load case 1: uniform load q ¼ 20 kN/m on the upper edge
Fc = 31 kN
–8.17
z = 3.8 m
Membrane force in x-direction
in section x = 0.5lx (in kN/m)
Fs = 31 kN
36.64
Main membrane forces X Z
force is equal to z ¼ 3.80 m. The reinforcement requirements are obtained from dividing
the tensile force Fs by the permissible stresses of steel: As,req. ¼ Fs/sd. The stress in the
reinforcement should be chosen to fulfil the crack width requirements (sd 4 fyd).
We will now compare the results of the FE calculation with the widely used manual
design method given in CEB (Comité Européen du Béton, 1970) or ‘Heft 240’ of the
German Association of Reinforced Concrete (DAfStb; see Grasser et al., 1991). The
example considers four different load cases.
g Load case 1: uniform load q ¼ 20 kN/m on the upper edge of the deep beam.
g Load case 2: uniform load q ¼ 20 kN/m on the lower edge of the deep beam.
g Load case 3: concentrated load F ¼ 96 kN on the upper edge of the deep beam
(loaded width: 0.96 m).
g Load case 4: concentrated load F ¼ 96 kN on the lower edge of the deep beam
(loaded width: 0.96 m).
The results of the FE analysis are shown in Figure 3.5 and Figures 3.7 to 3.9 for the
various load cases. The resulting forces and the lever arms are obtained from numerical
integration of the horizontal membrane forces at midspan.
CEB (Comité Européen du Béton, 1970) and ‘Heft 240’ of the German Association of
Reinforced Concrete (Grasser et al., 1991) offer two different design methods (see
boxed text)
142
where:
MF is the midspan moment of a beam having the same span lengths as the deep
beam
MS is the moment over the supports of a beam having the same span lengths as
the deep beam
zF is the lever arm of the internal forces at midspan
zS is the lever arm of the internal forces over the supports.
Ftd,S FEd,S
l l l
The lever arm of the internal forces zF and zS can be estimated as follows:
1. Single-span member
0.5 < h/l < 1.0 zF ¼ 0.3h (3 h/l) (Grasser et al., 1991)
zF ¼ 0.2 (l þ 2h) (Comité Européen du Béton, 1970)
h/l 5 1.0 zF ¼ 0.6l (Comité Européen du Béton, 1970;
Grasser et al., 1991)
2. Two-span member and end span of a multispan member
0.4 < h/l < 1.0 zF ¼ zS ¼ 0.5h (1.9 h/l) (Grasser et al., 1991)
zF ¼ zS ¼ 0.2l (l þ 1.5h) (Comité Européen du Béton, 1970)
h/l 5 1.0 zF ¼ zS ¼ 0.45l resp. 0.5l (Comité Européen du Béton, 1970)
3. Intermediate spans of a multispan member
0.3 < h/l < 1.0 zF ¼ zS ¼ 0.5h (1.8 h/l) (Grasser et al., 1991)
h/l 5 1.0 zF ¼ zS ¼ 0.40l (Grasser et al., 1991)
143
4. Cantilever member
1.0 < h/lk < 2.0 zF ¼ zS ¼ 0.65lk þ 0.10h (Grasser et al., 1991)
h/l 5 2.0 zF ¼ zS ¼ 0.85lk (Grasser et al., 1991)
where:
h is the height of the deep beam
l is the span of the deep beam
lk is the span of a cantilever member.
Table 3.1 summarises the main results of the FE analysis and the simplified method of
‘Heft 240’ (Grasser et al., 1991).
A good agreement can be seen between the FE results and the manual calculation, except
for load case 3 (beam theory), which is mainly due to the simple single-span system.
However, greater differences may be estimated for other structures (for example,
multi-span deep beams), as shown in the following example.
The internal forces for a two-span deep beam are calculated, carrying the loads from the
facade columns of an 11-storey office building, above the entrance to the underground
car park (Figure 3.10). The concentrated loads in the columns are simplified to an
equivalent uniformly distributed load (gEd/qEd ¼ 400/200 kN/m). The deep beam is
modelled by 52 15 plane shell elements (element size 0.25 0.25 m). In the first
analysis, an infinitely stiff vertical support is assumed. The stiffness of the supporting
columns is neglected.
Figure 3.7 Main membrane forces – load case 2: uniform load q ¼ 20 kN/m on the lower edge
–4.80
Membrane force in x-direction
–3.96
in section x = 0.5l x (in kN/m)
Load case 2: Fc = 31 kN
uniform load q = 20 kN/m
at the lower edge –8.17
z = 3.8 m
Fs = 31 kN
X Z 36.64
Y
144
Figure 3.8 Main membrane forces – load case 3: Force F ¼ 96 kN on the upper edge
–95.59
Fc = 25 kN
1.37
Load case 3:
concentrated force
q = 100 kN/m (F = 96 kN)
at the top
z=6.1 m
–1.80
Fs = 25 kN
27.5
Table 3.2 shows the resulting tensile forces in midspan and at the intermediate support
and the support reactions. The tensile forces obtained from the manual design method
(in accordance with ‘Heft 240’; Grasser et al., 1991) are greater than those from the
FE analysis. In particular, the beam model results in much higher forces over the
intermediate support (þ172%). This is due to the small lever arm zS with respect to
the numerical analysis.
Figure 3.9 Main membrane forces – load case 3: Force F ¼ 96 kN at the lower edge
–6
–5
Load case 4:
concentrated load
q = 100 kN/m (F = 96 kN)
at the bottom Fc = 39 kN
z = 3.9 m
–9
Fs = 39 kN
117
145
Table 3.1 Resulting forces and lever arms – single-span deep beam
DAfStb
beam 29 4.2 29 4.2 40 4.2 40 4.2
tables 29 – 29 – 26 – 38 –
to normal beam structures, the stiffness of the supports and the resulting deflections have
a considerable effect on the stresses and resulting internal forces in deep beams, as will be
demonstrated in the following section.
First, the influence of the deflection of the supports will be demonstrated for the
slender (h/l ¼ 0.6) double-span deep beam shown in Figure 3.10. The structure is
loaded by a uniform vertical load of q ¼ 600 kN/m at the top edge. Furthermore, it is
assumed that there is a settlement of the intermediate support by up to 7 mm.
Figures 3.11 and 3.12 show the results of the numerical analyses for different
settlements.
The reactions and the resultant horizontal forces are very considerably influenced by the
amount of settlement at the intermediate support. A deflection of only 2 mm is needed to
increase the vertical forces at the end supports by 37% and reduce the reaction at the
intermediate support by 38%. The resulting horizontal tensile force is increased by
200% in midspan and 33% at axis B. With increasing deflections, the stress distribution
becomes similar to that of a single-span beam with a span length of 2 6.25 ¼ 12.50 m
(Figure 3.12). A deflection of only 7 mm ( ¼ 0.001l ) is needed to reduce the reactions
at the intermediate support to zero.
146
Figure 3.10 Two-span deep beam: System, loading and membrane forces
Column
24/24 cm
3.75 m
2.80 m
50 1.0 m 50
6.50 m 6.50 m
x
y
190
252
are often supported by slender columns or walls. The simplest and often used approach
to consider the stiffness of the bearing structure is to model them by using individual
springs. For simplicity, the column itself is not modelled. This will be shown in the
following section.
147
Table 3.2 Tensile forces and lever arms – double-span deep beam
kN % m % kN % m % kN % kN %
FE analysis 650 100 2.6 100 700 100 3.6 100 1692 100 4417 100
‘Heft 240’ 675 104 2.4 94 1201 172 2.4 68 1609 95 4802 109
Grasser et al. (1991) 713 110 – – 825 118 – – 1609 95 4802 109
A column has a bending and a normal stiffness. Both should be modelled by individual
springs. It should be noted that bending springs cannot be used in conjunction with plane
membrane elements, as they have no degree of freedom for rotation (see the following
section); hence, a different method of modelling the bending stiffness of columns has
to be found.
In general, deep beams are loaded by large vertical forces that require large bearing areas.
This continuous support can be modelled by using several spring elements or special
boundary elements. The advantage of the latter model is that the elastic support stresses
can easily be estimated. These have to be checked in the design.
The axial stiffness of springs CN is obtained from the following expression (Table 3.3):
CN ¼ E . Ac/l
where:
E . Ac is the axial stiffness of the column
l is the height of the column.
Figure 3.11 Support forces and tensile forces depending on the settlement s of the intermediate
support
pp C
4000 or
t fo A=
axis
rce force
ax por t
is Sup
3000 B s
50 1.0 m 50
2000 an
idsp A 6.50 m B 6.50 m C
in m
ce F s
,F
or
1000 Tens ile f
148
Figure 3.12 Horizontal membrane forces at the midspan and at the intermediate support B
depending on the amount of vertical deflection s at axis B
–1056
–1416
–1777
–2137
–2497
–1728
–2464
–3201
–255
–992
3.75 m
50 1.0 m s 50
s = 1 mm
A 6.50 m B 6.50 m C
s = 1 mm
s = 5 mm
s = 5 mm
461
801
1245
1773
2497
s = 1 mm –1188
s = 2 mm –267
s = 5 mm 2790
s = 4 mm 2426
s = 3 mm 2063
s = 2 mm 1700
s = 1 mm 1337
s = 5 mm 2790
s = 4 mm 1575
s = 3 mm 654
As bending springs cannot be used in conjunction with plane shell elements, the
rotational stiffness of the column is modelled by using multiple springs, having a given
distance from each other (Figure 3.13).
The principles are demonstrated on a single-span deep beam (Figure 3.14). The structure
is supported on two columns (b/h ¼ 0.24/0.48 m) that are clamped at their base. This
v
E c, I c E c, I c E c, I c
Ac Ac
l l Ac
l
149
v5 v4 v3 v2 v1
a1 vi = ai .
a2
F i = C . ai .
a3
Mi = Fi . ai = C . a 2i .
a4
M = Mi = C . a 2i . = C .
a5
C = C . a 2i
results in the following equivalent stiffness for the springs: (concrete grade C 25/30,
Ec ¼ 30 500 MPa).
Ec Ac 30:5 106 0:24 0:48
CN ¼ ¼ ¼ 1:464 MN=m
l 2:4
4 Ec Ic 4 30:5 106 0:24 0:483 =12
C’ ¼ ¼ ¼ 112:4 MNm=m
l 2:4
Two springs are used for each support. The supported nodes are coupled with each other,
and thus the supported area remains plane before and after loading in the numerical
model. The distance of the normal springs has to be determined in order to get the correct
h/b = 24/48 cm
t = 24 cm X
l = 2.4 m
Y
Rigid
coupling
lx = 7.20 m
150
where:
ai is the distance of the spring from the centre of the supported area.
For the given structure, it is necessary to use a distance that is smaller than the actual
supported width. As this is an exceptional case, the springs are located at the face of
the supported area. This results in a distance of 0.48 m between the springs. The analysis
gives a spring force of 31.6 kN and 40.4 kN at the outer spring and inner spring, respec-
tively. The total vertical force is equal to 72 kN. The resulting bending moment at the top
of the column is:
M ¼ ð40:44 31:6Þ=2 0:48 ¼ 2:12 kNm
The influence of the spring stiffness on the bending moment is shown in Figure 3.15. It
can be seen that, in the practical relevant region of CN 1000 MN/m, the spring stiffness
has a significant effect on the bending moment at the column head. An infinite stiff
support results in a bending moment of M ¼ 10.3 kNm.
Figure 3.15 Bending moment at the supports against normal spring stiffness CN
m
10.3 kN
8 max M =
Rigid
7 coupling
Bending moment M in kNm
6 CN
5
4 F1 F2
3
N M
2
1
0
0 2000 4000 6000 8000 10 000
Spring stiffness C N in MN/m
151
bending moments, shear and normal forces) that are required for the design of the
columns must be separately calculated by the integration of the stresses in various
sections. Both of these models will be further discussed.
The connection of the beam elements with the plane shell elements should be handled
with great care, as will be demonstrated in the following example, which uses a single-
span deep beam. The columns are modelled with beam elements located on the axis of
gravity of the column. The top beam node is rigid, connected to the node of the deep
beam element at the same location.
The results of the FE analysis are shown in Figure 3.16. No bending moments are
estimated at the column heads. This result is not expected, as a rotation of the supported
area of the deep beam is likely to happen (note the deflected structure in Figure 3.17).
Furthermore, considerable distortions can be seen for the elements in the vicinity of
the column heads.
h/b = 24/48 cm
t = 24 cm
l = 2.4 m
Y
lx = 7.20 m
0.78
152
Deformed structure
Compression
tension
54.8
As zero bending moments are highly unexpected at the stiff joint between the column and
the deep beam, the same structure is recalculated, using plane shell elements for the
columns. Figure 3.17 shows the results of this analysis. A rotation of the upper end of
the column can be seen that must result in bending moments in the column.
The error of the beam model can be traced back to the fact that plane shell elements
have only two degrees of freedom for deflection, vx and vy, but no degree of freedom
for rotation (Figure 3.18). Thus, coupling of beam and plane shell elements results in
a so-called ‘incompatible’ element mesh. This means that there is a missing degree of
freedom (rotation) between the two types of elements. Thus, a plane shell element can
be used to only estimate membrane forces and not bending moments.
The rigid connection between the column and the deep beam must be modelled by stiff
coupling of the top node of the beam element to some nodes of the deep beam
153
Figure 3.18 Degrees of freedom of a plane shell and a beam element (2D)
v1 v4
u1 u4
v2
v1
u2
u1
y 2
1
x v3
v2 u3
u2
(Figure 3.19). The rotation of the supported nodes of the deep beam should be identical
to the rotation of the upper end of the beam element. Distortions of the shell elements
above the columns as shown in Figure 3.16 should be avoided.
A rigid connection can either be modelled with a special coupling of the nodes
(Figure 3.19 left) or by extension of the beam elements into the deep beam (Figure 3.19
right). Both models only provide a rough approximation of the real structural behaviour.
The number of coupled nodes can have a great effect on the bending moments in the
column. This is demonstrated by the results shown in Figure 3.20, where considerable
differences between both approaches in the calculated bending moments at the column
head can be seen. For the case with horizontal coupling, a bending moment at the
junction of M ¼ 3.7 kNm is estimated, whereas for the other model the bending
moment is only M ¼ 2.0 kNm. The ratio of the first moment with respect to the
second one is 3.7/2.0 ¼ 1.85!
Shell element
154
Figure 3.20 Membrane forces and the deformed structure for two different models – coupling
with supported nodes (left) and extension of the beam elements (right)
Deformed structure
Here, it should be noted that there are also plane shell elements, which have a rotational
degree of freedom (based on Cosserat continuum). However, these elements are rarely
implemented in structural software used for the practical design of concrete structures.
A highly refined FE mesh has to be used to obtain a more accurate value for the bending
moment at the column head. The results of this analysis are shown in Figure 3.21.
The numerical integration of the vertical membrane forces at the top of the column
(Section 2.2) results in a force of F ¼ 72 kN (F ¼ 7.20 m 20 kN/m 0.5 ¼ 72 kN) and
a bending moment of M ¼ 2.6 kNm. This value is between the two results from the
earlier-mentioned models.
An unbalanced distribution of the vertical membrane force ny in Section 4-4 at the inner
face of the corner can be seen in Figure 3.21. This is a result of the simplifications and
assumptions of the used numerical model and does not occur in real structures. The
corner causes a model problem. The unbalanced boundary condition results in high (infi-
nite) stresses resp. membrane forces. If the element size is further reduced, the stresses at
the inner face of the corner will become infinite. However, the resulting bending moment
M is only slightly changed.
155
Figure 3.21 Vertical membrane force ny in various sections near the corner of a deep beam
3 4 Section 3 –3 Section 4 – 4
–80 –60
1 1 –131 –99
–142 –213
2 2
3.60 m
Detail
–152
–150 –182
7.20 m
3 4
–103
–117
–133
–214
Detail
Finite Element model
X
Section 1–1
Section 2–2 Y
The stress distribution in any corner region of membranes can only be approximated
when using plane elements with a linear elastic material behaviour. High tensile stresses
would cause cracking of the concrete, thus resulting in a redistribution of the internal
forces. The same effects occur in the case of large pressures at the inner edge of the
corner. Therefore, a highly refined element mesh does not model reality any more
precisely than a coarse element mesh.
The deep beam shown in Figure 3.22 is a very slender structure, having a width-to-depth
ratio of 1 : 2. The member can, therefore, also be considered as a normal beam with
the assumption of a linear strain distribution over its depth at midspan. The structure
is modelled with 14 30 plane shell elements. It is loaded with a vertical uniformly
distributed loading of q ¼ 20 kN/m at its upper free edge.
156
ly = 3.60 m
asup = 24 asup = 24
t = 24 cm
lx = 7.20 m Supports
Figure 3.23 shows the distribution of the main membrane forces nI and nII and the
horizontal membrane force nx at midspan. Significant differences can be seen between
the two models.
If one uses the manual design models provided in ‘Heft 240’ (Grasser et al., 1991), the
calculated tensile force is F ¼ 53 kN. This is the same value as the resulting tensile
force of System 1. If the structure is fully restrained in the horizontal direction
(System 2), the resulting tensile force in midspan is only 1/5 of the value calculated of
System 1, even for this very slender structure.
System 1: nodes at the supports fixed System 2: nodes at the supports not restrained
in horizontal direction
Main membrane forces
59.6 23
157
g Concrete takes no tension (1c ¼ 0): This assumption is not true in the
serviceability limit state, where one has to consider a reasonable tensile strength of
concrete to get realistic results.
g Cracks are orientated orthogonal to the principal tensile stress: The orientation of
the cracks depends not only on orientation of the stresses, but further on the rebar
arrangement.
g Mohr’s failure criterion is applied: Research is still ongoing to develop consistent
models to describe the behaviour of reinforced concrete membranes. Mohr’s
failure criterion is an old model, but is still used quite often.
g Sufficient ductility: The capacity of concrete sections for redistribution of forces
after cracking is limited. This is also called as ductility demand. Therefore, the
assumed load path after cracking should be similar to the elastic flow of forces.
The maximum compressive strength of a concrete element having compressive and
tensile stresses is less than the uni-axial compressive strength. A reduction or
‘disturbance’ factor of at least ¼ 0.85 should be used to take this effect into
account ( fcd ¼ 0.85 . . fck/ c).
This gives the minimum amount of reinforcement for a shear wall or deep beam. As the
orientation of the principal axes changes in each point of a structure, and as various load
cases have to be considered, the trajectory arrangement of reinforcing bars is mostly not
possible.
The angle of the compression struts should be chosen with regard to the stress field
based on a linear elastic analysis to minimise the required redistribution of forces.
158
ρly · fyd $ σy
σy σy
σx σx
ρlx · fyd $ σx
ρlx · fyd $ σx
σy A
σx B σx τxy > 0
σy τxy
σx A σx
σx σx
σy σy
ρly · fyd $ σy
ρl
σ1 1 ·f ρly · fyd $ σy + |τxy| · cot θ
σ2
yd $
σ1
σy σy
σx σx
B
σ2
τxy Þ 0 B
σ2
τxy Þ 0
σ2
σ1
$ yd
·f
σx σx
2
ρl
σy σy
σ2
σ1
fcd $ ρlx · fyd + ρly · fyd – σx – σy
ρx · fyd,x
159
q = 20 kN/m
0.41
Horizontal reinforcement
3.6 m
Section in midspan
0.5 cm2/m
1.0 cm2/m
1.5 cm2/m
2.0 cm2/m
2.08
7.20 m
flow of forces in a cracked member does not differ much from that of an uncracked one.
Membranes having significant shear forces should not be designed with this approach.
Instead, strut-and-tie models can be used.
The same problem exists when the deep beam is loaded at the bottom edge. The amount
of vertical reinforcement is calculated correctly (Figure 3.27). However, it must be kept in
mind that the vertical reinforcement should not be staggered over the depth of the deep
beam in accordance with the vertical tensile force distribution. The external vertical force
has to be completely transferred to the top of the deep beam.
160
Figure 3.27 Horizontal and vertical reinforcement (loading with q ¼ 152.3 kN/m at the bottom
edge)
3.6 m
2.0
4.5 4.5
3.
0 3.0
4.0 4.0
3.5
7.20 m 3.49
1.0
1.0 2.0
5.0
stresses are much more efficient than rebars in regions with less tensile stresses.
Reinforcement in the upper part of the tensile zone will only yield under ultimate limit
loads if very wide vertical cracks are formed. Thus, the tensile reinforcement of a
single-span deep beam (as in an ordinary beam) has to be concentrated at the bottom
of the beam and should not be staggered over the depth of the tensile zone, in accordance
with the local tensile forces (Figures 3.26 and 3.27).
161
Figure 3.28 Arrangement of the horizontal reinforcement for the tensile force Ft over the supports
of multi-span deep beams
Adapted from Grasser et al. (1991) with permission
h/l = oo
h/l = 1.0
h/l = 0.75
h/l = 0.5
h
h
2/3Fs 0.84l
0.4l 0.6l 0.6l
h
Reinforcement
distribution
h 1/3Fs
0.4l
nx
0.2l 0.16l 0.1l 0.1l
Loads acting at the bottom of deep beams, that is, loads acting in the region of a half
circle with a radius of 0.5l (where l 4 h), including the self-weight, should be carried
by vertical reinforcement, which must have a length of l 4 h.
As can be seen from the preceding list, it is not sufficient to only determine the relevant
design forces.
The internal forces of ordinary beams can be estimated by using beam and plane shell
elements. This was already demonstrated through some examples in Chapter 2. FE
calculations based on linear elastic material behaviour may be useful to determine the
flow of forces in discontinuous regions, where the Bernoulli hypothesis is not valid.
Examples for this are regions with highly concentrated forces, which occur when
anchoring the tendons in prestressed structures, or where there are sudden changes in
cross-sectional depths, such as halving joints.
162
advantage of this simple structure is that the required reinforcement for the ultimate limit
state can be estimated manually by well-known formulae, whereas for the case of deep
beams only approximate methods are available.
The manually calculated values are in a very good agreement with the resulting forces of
the FE analysis (Figure 3.29).
Figure 3.29 Single-span beam: system, loading and main membrane force distribution
20 kN/m
h = 1.0 m
l/2 = 5.0 m
b = 0.3 m
163
A partial safety coefficient of ¼ 1.45 is used for estimating the required bending
reinforcement. This gives the required amount of horizontal reinforcement at the
midspan (concrete grade C25/30, fyk ¼ 500 MPa, b/d/h ¼ 0.30/0.95/1.0 m):
MEds 250 1:45 103
moment coefficient Eds ¼ ¼ ¼ 0:095
b d 2 fcd 0:3 0:952 ð0:85 25=1:5Þ
inner lever arm z ¼ ð1 0:6 Eds Þ d ¼ ð1 0:6 0:095Þ 0:95 ¼ 0:896 m
MEds 250 1:45
required reinforcement req As ¼ ¼ ¼ 9:3 cm2
z fyd 0:896 43:5
The computer program estimates the reinforcement in each element separately. For the
lowest element, this results in a required horizontal reinforcement of (Figure 3.29):
req As ¼ ðnx1 þ nx2 Þ=2 ðh=5Þ =fyd
¼ ð1500 þ 900Þ=2 ð1:0=5Þ 1:45=43:5
¼ 8:0 cm2 (over a depth of 0.2 m!)
Figure 3.30 Simply supported beam with uniform loading – distribution of the reinforcement
calculated by means of a membrane model
0
30
50
2.5
5
10
164
Reinforcement for the other elements is shown in Figure 3.30. If one extrapolates the
results in the middle of the bottom elements to the lower boundary, the estimated
maximum reinforcement As,required would be greater than 50 cm2/m depth.
The numerical integration of the tensile zone results in a required total reinforcement of:
req As ¼ ðnx1 þ 0Þ=2 ðh=2Þ =fyd
¼ ð1500 þ 0Þ=2 ð1:0=5Þ 1:45=43:5
¼ 12:9 cm2
Such a numerical calculation not only results in a wrong distribution of the reinforce-
ment, but it is also uneconomical: it overestimates the reinforcement requirements by
39%. This is due to the lever arm of the internal forces of a cracked member which is
greater than the one of an uncracked structural element (see Figure 3.31).
This is also true for deep beams. The distribution of the vertical reinforcement is
shown in Figure 3.30. The estimated area of stirrups in the critical section at a distance
of 1.0d from the face of the supports is as,w ¼ 5 cm2/m. This is much greater than the
required value calculated by the manual design. According to EC2, stirrups of area
as,w ¼ 0.9 cm2/m are sufficient to carry the loads in the ultimate limit state.
Figure 3.31 Strain and stress distribution and internal forces of cracked (elastic) and uncracked
members
Actions Resistance
Elastic Non-linear
c2 c2 –
fcd
a = kjx
x = 0.5h
x = d
–Fcd
MEd –Fcd
z = 2/3h
z = kzd
s Fsd s
Fsd
c1 c1
Strain Stresses and Strain Stresses and
stress resultants stress resultants
165
for this error in the shell model is the assumption of linear elastic uncracked material
behaviour. The design for shear of a reinforced or prestressed beam is usually based
on a strut-and-tie model, with a fully cracked member.
This very simple example of a simply supported beam should demonstrate that the
nonlinear material behaviour of concrete, as well as the cracking of the composite
material ‘reinforced concrete’, has to be considered in the design. Estimation of the
reinforcement requirements by integrating the tensile forces of a linear elastic model
may lead to incorrect and even sometimes unsafe values.
–4.80
Compression strut
Compression strut
–3.96
Compression strut Fc = 31 kN
ly
– –8.17
z = 3.8 m
z
z1
tie Fs = 31 kN
z2 +
0.55 m
X Z 36.6
Y
lx Horizontal membrane forces
at midspan x = 0.5lx (in kN/m)
asup asup
166
Strut-and-tie models consist of straight compression and tension struts, which are
connected by hinges. The truss model should be statically determined and should
represent the main force distribution in the member. Schlaich and Schäfer provide
further information regarding this design method (Schlaich et al., 1987; Schlaich and
Schäfer, 1998). The main principles are explained in the following for a simple single-
span deep beam.
At first, one has to determine the force distribution in the member. This can be done by
means of the force flow method or linear elastic FE analysis. The strut locations are
similar to the main compression stresses. The only unknown in the manual analysis
for the given single-span deep beam is the location of the horizontal compression strut
and the distance z1 and z2, respectively (see Figure 3.32). The horizontal compression
7 18
40 –189
–7 –7
–53
–53
–53
167
and tension struts are located in the centre of the stress field at the midspan. Thus, the
increase of the lever arm of the cracked member under ultimate loads is neglected.
The strut-and-tie model is a statically determined truss structure. Thus, all forces can
be estimated based on the geometry of the truss system only. The resulting tensile
force for the given structure (lx ¼ ly ¼ 7.20 m; a ¼ 0.24 m; q ¼ 20 kN/m) is:
lx a q l=2 7:20 0:48 20 7:2=2
Ft ¼ ¼ ¼ 31:8 kN
4 z 4 3:8
The calculated force is nearly identical to the value given in Grasser et al. (1991). There is
a small difference of 3 kN only (see Table 3.1).
In the preceding example, the reinforcement requirements are estimated from the
resulting tensile forces. However, as will be shown later, this reinforcement can be
inadequate with respect to the serviceability of a structure. Figure 3.33 shows the
Figure 3.34 Two-span deep beam (lx /ly ¼ 2 7.5/3.65 m) with an opening
33 –203
Vertical membrane force ny
+2
–10 –8
–31
–51
–54
–71
168
distribution of the main membrane forces and the resulting strut-and-tie model for a two-
span deep beam (lx/ly ¼ 2 7.5/3.65 m). The model is nearly identical to that of a single-
span deep beam. The tensile force at the intermediate support is the only addition to be
considered. The tensile reinforcement at the midspan must not be concentrated at the
centre axis of the tensile zone, but at the bottom of the deep beam.
Figure 3.34 shows the same system with a small opening near to the left support. Here,
the reinforcement calculated from the resulting tensile forces is insufficient. A linear
elastic calculation does not show any vertical tensile forces near the opening. This is in
contrast to engineering practice and contradicts with the strut-and-tie model. Vertical
reinforcement is required.
Therefore, the nonlinear behaviour of the material has to be considered when developing
a strut-and-tie model.
3.5. Singularities
In shell systems, infinite stresses and deformations (singularities) may be calculated,
which are caused by the assumptions of the numerical model. Examples of these singu-
larities are (see Figure 3.35, Table 3.4)
An FE analysis will always estimate finite results, whereby the maximum stress will
increase considerably with any decrease in the element size. The ‘exact’ calculation of
Concentrated force
Corners
Pin supports
169
Table 3.4 Singularity regions in plane shell structures (walls and deep beams)
No Yes
> 1808
Yes Yes
> 638
Yes Yes
No No
No Yes
No (x)
No No
No
the maximum value of the stresses is generally not required, as the problems are caused
by the simplifications of the numerical model, an inaccuracy of the boundary conditions.
A real structure does not show any singularities. In the case of high stresses, the material
will ‘yield’. In tensile regions, the concrete will crack. Furthermore, in discontinuity
regions, the stress distribution is a complex 3D one that will not be modelled with
plane shell elements. The main concern will be the good detailing of the reinforcement
in the corner regions and in the area of concentrated loading.
Numerical problems with concentrated loads can be avoided if the width of the loaded
area is considered. The same is true for pin supports. However, any refinement of the
numerical model is generally not required, as the stress distribution in these regions is
not needed for the design.
170
REFERENCES
CEN (European Committee for Standardization) (2004) Eurocode 2, Part 1: Design of
concrete structures – General rules and rules for buildings. December.
Comité Européen du Béton (1970) Recommendations for design and construction of
concrete structures, Appendix 3, London, June.
Fédération Internationale du Béton (fib) (2007) Practitioners’ guide to finite element
modelling of reinforced concrete structures. FIB-Bulletin No. 45, Lausanne,
November.
Grasser E et al. (1991) Hilfsmittel zur Berechnung der Schnittgrößen und Formänder-
ungen von Stahlbetonbauwerken; Deutscher Ausschusses für Stahlbeton Heft 240.
Berlin.
Schlaich J and Schäfer K (1998) Konstruieren im Stahlbetonbau. In Betonkalender (Eibl
J (ed.)), part II, Berlin.
Schlaich J et al. (1987) Towards a consistent design of structural concrete. PCI
Journal, May–June: 77–150.
171
Chapter 4
Slabs
4.1. General
Slabs are thin plane spatial surface structures that are only loaded perpendicular to their
middle plane and where the minimum panel dimension is greater than five times the
overall slab thickness (Eurocode 2, 2004). The internal forces in slabs are the bending
moments mx and my, twisting moment mxy and shear forces vx and vy per unit length
(see Figure 4.1). Normal forces on the middle plane are not considered in this approach;
however, they can be taken into account by an additional shell analysis (diaphragm). It
should be noted that the linear combination of various load cases is only permissible if
the structural behaviour is fully linear elastic (Stage I, tensile stresses in the concrete
4 fctk;0.05/ c). Owing to the mentioned basic model assumptions, normal forces are
not even estimated if a slab is fully restrained: the middle plane is always stress-free.
In addition, the following assumptions are generally used
x
lx
Midplane mx
ly
my
h
w
g thin slab: h l; vertical stresses z ¼ xz ¼ yz ¼ 0; h ¼ depth of slab ¼ constant
g small vertical displacements (w h; first-order theory)
g linear strain distribution over the section depth (Navier)
g no strains at the middle plane (i.e. no normal or membrane forces)
g stresses in normal direction can be neglected
g plane sections remain plane before and after loading (Bernoulli-Euler).
The basics of a plate analysis are not mentioned any further in this book as extensive
information on this topic has already been published (e.g. see Girkmann 1978;
Timoshenko and Woinowsky-Krieger, 1959).
173
dx mxy
Shear forces v Moments m
vy x mx
x
my my = (my /x ) . dx
qz mxy
dy
vx
v x = v x + (vx / x) . dx mxy = (mxy /x ) . dx
mx = (mx /y ) . dy
y v y = v y + (vy /y ) . dy y
mxy = (mxy /y ) . dy
y
Stresses
Definition:
my m y is the moment,
which rotates around
x the y-axis
y
z
Slabs can be analysed manually by means of the strip method. In this approach, the 2D
load transfer is simplified to two ordinary beams or a grillage system. This simple method
is still applied to this day, as has been shown in Section 2.8. Nevertheless, the 2D
behaviour of slabs (dispersion of loads) is represented much more accurately by FE
models using plate and shell elements.
In recent years, FEM has become a standard tool for the analysis of spatial structures.
The practical applications have dramatically increased with the significant improvements
of the software. The graphical pre- and post-processing makes it fast and easy to handle
whole buildings, including all slabs, walls, columns, and foundations, within one big
numerical model. Also, the hardware has become much faster. However, complex FE
models include a big danger, as the collapse of the Sleipner platform has impressively
174
Finite
Element
Nodes
demonstrated (see Section 5.1.1). Therefore, only simple flat slabs will be mentioned in
the following.
In FEM, the slab is divided into small FEs (discretisation of a continuum), which are
connected to each other by their nodes (Figures 4.2 and 4.3). The deflections, strains
and internal forces within an element are interpolated from the node displacements by
means of so-called form functions, which are mostly polynomials. The form function
may but must not fulfil the compatibility conditions at the boundaries between the
elements. Therefore, the deformations, strains and stresses, as well as the internal
forces, may be discontinuous between the elements. The difference of the stresses at
the boundaries of the elements may be used to verify the quality of the FE model.
y,2 y,1
y
x,2 x,1
x
w2 w1
z
y,3 Midplane
y,4
x,3
x,4
w3 w4
175
The required effort to generate an adequate FE model has decreased dramatically due to
the availability of user-friendly graphical pre-processors. In the near future, it may even
be possible to perform the analysis and design of the slab for a ‘simple’ one-storey resi-
dential building (such as the one shown in Figure 4.4) by means of the FE techniques
instead of manual analysis. Therefore, one may worry that FE software will be
more and more used by designers without a proper understanding of the method and
experience in design of concrete structures. This problem becomes even worse as a
main advantage of FEM is that the user does not have to simplify the load transfer of
a structure as by manual analysis. Therefore, young engineers without structural experi-
ence tend to use the numerical method. This may lead to mistakes in the results that may
even cause the collapse of a structure (see Section 5.1.1).
The real system is simplified for an FE analysis. A simple line support is assumed under-
neath all supporting walls (breadth of support neglected), whereby small openings like
176
that of doors or windows are neglected. The slab thickness is h ¼ 0.18 m. Concrete
grade C20/25 is assumed for the building material. The slab is discretised by 451 plane
plate elements. Figures 4.4a–d show the internal forces and deflections for a uniformly
distributed load of gd ¼ 5.0 kN/m2. For simplicity, the most unfavourable arrange-
ments of live loads are not considered here. Very high bending moments and shear
forces can be observed in axis B-2 and B-3. The slab cannot be designed for these high
member forces. Even an inexperienced user can recognise this problem. The structural
engineer has to know that the design for such high internal forces is not required.
These peaks are caused by simplifications of the numerical model (singularities; see
Section 4.12). In the following sections, such problems in the model for reinforced
concrete slabs will be discussed with emphasis on the accurate representation of the
support conditions.
D C B A
1 1
2 2
Point support
Re-entrant corner
3 3
4 4
D C B A
177
0.2
0.8
1.2
1.4
1.2
1.0
0.8
0.2 0.6
0.4
178
2.0
4.0
2.0
0.0
6.0
2.0
6.0
2.0
0.0
0.0 -2.0
0. 0
0 0
-
6- 4.
0 0.000.0.021.4-16-1-1-
8-
0.
0
0.0 8.0
2.0
6.0
4.0
2.0
2.0
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Figure 4.4d Principal shear force v ¼ vx2 þ vy2 in kN/m
6.00
4.0
2.0
4.0
2.0
0
24.0
6 .0
2.0
8.0
6.0
33
6.0
4.0
2.0 2.0
2.0
4.0
4.0 6.0
179
Figure 4.5 Member forces of a one-way slab for different numbers of FEs
–31.25 (analyt.)
–30 q = 10 kN/m –27.0 (40 el.)
–25 –23.4 (20 el.)
5m 5m
–20
Bending moment in kNm/m
–5
0 1 2 3 4 5 6 7 8 9 10
0
10
15 15.7
(10 el.) 17.3 (40 el.)
17.6 (analy.)
20
10
0
0 1 2 3 4 5 6 7 8 9 10
–10
q = 10 kN/m 40 elements
–20 20 elements
10 elements
5m 5m
–30
180
The greatest field moment with five elements is 11% less than the ‘exact’ analytical
value only. Nevertheless, this small difference does not apply for support areas or near
single forces and area loads. Here, very high gradients in the member forces appear,
which cannot be modelled by a coarse mesh. Furthermore, it is to be noted that an
FE program based on four-noded elements calculates the member forces and the
reinforcement mostly in the centre of the elements only. Hence, with a coarse
discretisation, the moment peak in midspan is cut off (Figure 4.5). This can result in a
considerably unsafe design near the inner supports. Similar problems also appear with
flat slabs.
The Poisson’s ratio is defined for an elastic member as the ratio between the lateral
strain and the axial strain. According to EC2 (Eurocode 2, 2004: Section 3.1.3), the
Poisson’s ratio in an uncracked compression region of the cross-section may be taken
as ¼ 0.2, which is the mean value of a ‘homogeneous’ uncracked concrete under
compression. In the tension zone, may be assumed to be equal to zero. However,
this information is not very helpful as a slab under pure bending always has a
compression and a tension zone (Figure 4.6).
Figure 4.6 Strain in transverse direction of a linear elastic plate element under bending (one-way
slab)
Top face
Deformed
Plate element
Undeformed
Elevation
Bottom face
181
Bittner (1965) has conducted theoretical investigations to determine the correct value
for Poisson’s ratio. He proposed that a value of ¼ 0.0 should be used in the design
of reinforced concrete slabs. With this value, the compressive stresses are under-
estimated. This is not mostly critical, as the compressive stresses are usually not relevant
in the design.
So, the structural engineer can choose any value for between 0.0 and 0.2 for the design.
The influence of the Poisson’s ratio on the internal forces will be discussed in the
following box.
The Kirchhoff plate theory leads to the following expressions, which show the influence
of the Poisson’s ratio on the bending moments, the shear forces and the corner tie-down
force.
182
Figure 4.7 Bending moment my, twisting moment mxy and corner tie-down force Fe for different
Poisson’s ratios (lx /ly ¼ 5.0/5.0 m and 5.0/7.5 m, and h ¼ 0.20 m)
mym
Values relate to a Poisson’s ratio of 0.0
1.9 ly/ly = 7.5/5 m
Fe Fe
mx
1.7
my
1.5
Fe Fe mym
1.3 ly/ly = 5/5 m
1.1
0.9
0.7 Fe
mxy
0.5
0.0 0.1 0.2 0.3 0.4
Poisson’s ratio
In slabs, the secondary transverse reinforcement should not be less than 20% of the
principal reinforcement (Eurocode 2, 2004). Thus, the value of the Poisson’s ratio is
often not critical. However, this minimum transverse reinforcement has to be considered
in the FE design of slabs.
Normal and shear forces, as well as bending moments, can be transferred if the slab is
monolithically connected to the support. The stiffness of the support can be represented
either by individual springs or by special boundary elements. In both cases, the rotational
stiffness of the springs should be determined with regard to the nonlinear material
behaviour of the concrete and the reduction of stiffness in the case of crack formation.
Where the supports have different vertical stiffness values, any possible deflection of the
supports should be considered. This can be the case for flat slabs that are supported by
columns with different cross-sections, normal forces or building materials. Slabs usually
have only a small bending stiffness. Thus, the influence of the differential deformation on
the internal and support forces can be mostly neglected.
183
If the slabs are supported on different building materials, having different time-
dependent properties, such as aerated concrete brick and clay brick walls, then the
time-dependent deformations of the supports (e.g. due to creep or shrinkage) should
be considered in the analysis. A wall made of gypsum blocks – or sand-lime concrete –
may become shorter with time, whereas the height of a clay brick wall remains nearly
constant. Differential settlements of the foundations should be considered if relevant.
Various numerical models are used in practice to describe the support conditions of
slabs. These may result in rather different internal forces in the restrained regions.
However, one should not compare the maximum peak values. Only the bending
moments and shear forces in the sections, which are relevant for design of concrete
slabs, should be used.
The peak bending moments and shear forces over the centreline of an intermediate
support are generally not relevant in the design. In the case of a stiff monolithic connec-
tion between the slab and the supported columns or walls, the design in bending is carried
out for the forces at the face of the support. If the slab is restrained in vertical direction
only, for example, it is supported on block or brick walls, the maximum design bending
moment is calculated from the values at the face of the supports and smoothed para-
bolically (see Figure 4.8).
The relevant section for shear design is either the face of the support, in the case of
indirect supports, such as support on upstand beams, or at a distance of 0.5–1.0d
from the face for direct supports (d ¼ effective depth of the slab). The slab should be
designed for punching in regions of concentrated single loads or pin supports by columns
and not for the high resultant shear forces. To summarise, the maximum values and the
Pin support:
MEd smoothed bending moment
M*Ed M |M*Ed | = |MEd| – |FEd,sup| . bsup/8
Ed,left MEd,right
FEd,sup
Monolithical connection:
relevant moment at the face of the support
Shear force
VEd,right |MEd,left| = |MEd| – |VEd,left| . bsup/2 min|MEd|
|MEd,right| = |MEd| – |VEd,right| . bsup/2 min|MEd|
VEd,left
where:
FEd,sup = design support forces
bsup = breadth of the support
bsup
184
distribution of the bending moments and shear forces over the supports are generally not
required for the design of concrete slabs.
Therefore, in the following comparison calculations, only the values that are relevant for
the design are considered.
First, the various possibilities of modelling a ‘hinge’ line support are examined for a
simple one-way slab. This simple structure is used since the member forces from a
beam system can be used to check the numerical results. The outcome of this investiga-
tion will then be used to determine the ‘accurate’ model of flat slabs, where analytical
solutions are not available.
The different models for a support on walls that are used in practice are shown in
Figure 4.11.
(a) 3D model of the whole structure (column and walls) by volume elements: If the
whole structure behaves as linear elastic, then the load-bearing behaviour of the
system is modelled with great accuracy by means of volume elements. The
nonlinear strain distribution in the support region (discontinuity region) is
considered in this model. However, the assumption of an uncracked section over
the intermediate support is not valid, as it is very likely that the tensile stresses
h = 20 cm
25 cm
5m 5m
20
30
40
50
60
70
80
1
185
5m 5m
–31.3 kNm
–29.3 kNm (smoothed moment)
–27.3 kNm
(moment at the face of the support)
Bending moment
Support
25 cm
17.6 kNm
Bending moment
Support
24.0 kNm
exceed the tensile strength of the concrete. Moreover, great effort has to be taken
to handle a 3D volume or even a 2D shell model, which is often not required. It
should be noted that the problems arise when the reinforcement requirements are
being estimated. Therefore, in practice, a 3D shell or volume model is very rarely
used.
(b) Pin support of one node: The node in the centre of the support is restrained in the
vertical direction, as in a beam model.
(c) Pin support of all nodes at the support – restrained in vertical direction only: This
model is sometimes used to consider the breadth of a very rigid wall.
(d ) Fully restrained of all nodes at the supports: The vertical deformation as well as the
rotation of all nodes at the support are completely restrained. The elements over
the support are, therefore, free of stresses and show no internal forces.
186
d) Fully restrained of the nodes e) Pin support and f) Bedding of all nodes
at the face of support coupling
Diaphragm
(e) Coupling of nodes: The nodes at the support are coupled with the node in the
centre of the support to simulate an infinite stiff element that can rotate around
the centre node. This model has already been discussed in Section 2.6.3.
( f ) Bedding of the supported elements: The elements over the wall are elastic
supported. A flexible, plane support is simulated.
The different bending moment distributions, resulting from the models described
earlier, are shown in Figures 4.12 and 4.13. The relevant results are further summarised
in Table 4.1.
Model B (pin support of one node only) gives a good correlation between the beam and
the FE (plate element) results for both load cases. Any further refinement of the element
mesh does not change the results. Knife-edge-supported one-way slabs do not show any
singularities.
If three nodes are fixed in a vertical direction at the support (Model C), the resulting
bending moments at the face and those at the centre of the support are underestimated
for uniform loading over both spans by 89% and 47%, respectively. On the other hand,
if only one span is loaded, the support bending moment is overestimated by 226%.
187
Figure 4.12 Bending moments – uniform load at both spans – plate elements
–31.2
–27.4
q = 10 kN/m2
Pin support of 1 × 2 nodes
5m 5m
Axis of support
Support
17.6 –28.0
–24.3
Axis of support
Support
17.4
–25.9
Axis of support
Support
16.7 –28.0
Hinge coupling
Axis of support
Support
17.3 –29.1
–27.4
Axis of support
Support
17.6
Therefore, this model should not be used. A further refinement of the element mesh
would result in a fully restrained intermediate support. Doubling the number of elements
increases the bending moments by 15% (Msup,face,max ¼ 28.1 kNm/m). The influence of
the element size decreases by 1%, when only one span is loaded.
188
Figure 4.13 Bending moments – uniform load at left span only – plate elements
Axis of support
Support
–27.8 q = 10 kN/m 2
23.9
5m 5m
Pin support of 3 × 2 nodes
Axis of support
Support
17.5
–25.9
Axis of support
Support
16.7
–15.5
Hinge coupling –12.5
Axis of support
Support
23.7
–16.9
Bedding C = 9000 MN/m 3 14.6
–10.5
Axis of support
Support
21.9
The results of Model D (all nodes at the supports are completely restrained) show little
difference from those of Model C. The internal forces of the elements over the
support become zero, as the nodes can neither move nor rotate. Furthermore, there
are no bending moments or shear forces in the unloaded part of the structure. Therefore,
189
qEd = 10 kN/m 2 Field moment Moment at the edge of Moment at the cl. of
max MF: kNm/m support MSA: kNm/m support MSM: kNm/m
M S,A M S,M
max M F max M F
5m 5m
(b) Pin support of one node 17.6 (100%) 27.4 (100%) 31.2 (100%)
(c) Pin support of three nodes above the wall 17.4 (99%) 24.3 (89%) 14.5 (47%)
(d) Fully restraint of the nodes at the face of the column 16.7 (95%) 25.9 (95%) 0
(e) Hinge coupling 17.3 (98%) 28.0 (102%) 14.5 (47%)
(f ) Bedding (C ¼ 9 106 kN/m3) 17.6 (100%) 27.4 (100%) 29.1 (94%)
qEd = 10 kN/m 2 Field moment Moment at the edge of Moment at the cl. of
max MF: kNm/m support MSA: kNm/m support MSM: kNm/m
MSM
MSA
max MF
5m 5m
(b) Pin support of one node 23.9 (100%) 12.2/15.2 (99/99%) 15.6 (99%)
(c) Pin support of three nodes above the wall 17.5 (73%) 27.8/0 (226/%) 7.5 (48%)
(d) Fully restraint of the nodes at the face of the column 16.7 (70%) 25.9/0 (211/%) 0
(e) Hinge coupling 23.7 (99%) 12.5/15.5 (101/101%) 14.5 (92%)
(f ) Bedding (C ¼ 9 106 kN/m3) 21.9 (91%) 16.9/10.5 (137/68%) 14.6 (92%)
191
Finite-element Design of Concrete Structures
this model should not be used. Refining the element mesh results in a higher bending
moment at the support (see Section 4.13.2).
The ‘hinge’ coupling (Model E) shows a good correlation with the beam model. Only
the support moment Msup in load case 1 shows greater deviations from the beam
values. Further refinement of the element mesh only changes the moments at the
centre of the support (from Msup,max ¼ 17.9 to 7.4 kNm/m) that are not relevant
for design.
An elastic bedding of the supported elements at the wall (Model E) results in a sufficient
correlation with the beam moments and shear forces for both load cases. When only one
span is loaded, the bending moment at the face of the support is approximately 37%
greater than the values that are manually calculated. However, it should be noted that
the results are highly dependent on the bedding modulus C used in the analysis. High
values may result in fully restrained nodes such as those in Model C.
When soft supported elements are used, the bedding modulus C has to be calculated first.
In practice, C is often derived from the normal stiffness of the wall (C ¼ EA/l) or the
bearing (see Figure 4.14). It must be emphasised that this approach considers only the
global deformation of the supporting structure and the resulting force redistribution.
The local deformations, the stresses and strains in the slab in the vicinity of the wall
with respect to the column head, which are required to determine the ‘exact’ internal
forces, cannot be modelled with this simple approach either.
The main assumption of all plate elements, the linear strain distribution, is not valid in
the region of the support (see Figure 4.15). Therefore, the bedding of the elements is an
engineering tool, to get a smooth distribution of the bending moments over the support,
in a manner similar to those in manual analysis.
Assumptions
• linear elastic material
• uniform stress distribution
u • Poisson’s ratio ν = 0 (no transverse deformation)
σ = C · u → C = σ/u
∂u u u·E u·E E
ε= ø = σ/E → σ = →C= =
h ∂x h h h·u h
where:
E is the modulus of elasticity of the elastic bearing
h is the thickness of the elastic bearing
192
20 cm
0.4
-1.61
–4.3 –6.9 –6.9 –4.3
20 cm
–3.4 –3.4
Vertical stress x y
Slab supported on bricks 12.5 cm 12.5 cm
x
Table 4.2 shows the bending moments for different values of C from C ¼ 9 105 kN/m3
to C ¼ 9 107 kN/m3. The results are not sensitive to the bedding modulus C used for a
uniform loading over both spans, as the deflection curve over the support show a hori-
zontal tangent. On the other hand, when only one span is loaded, the bending moments
are highly dependent on the value of C (see Table 4.2). A refinement of the element mesh
in the support region does not change the internal forces significantly.
A slab has to be designed for both bending and shear. The section at a distance of 1.0d
from the face of the support is relevant for the shear design for the given direct support of
the structure. The shear forces are listed in Table 4.3 for different models. A constant
static depth of d ¼ 20 2.5 cm ¼ 17.5 cm is used for comparison.
193
qEd = 10 kN/m 2 Field moment Moment at the edge of Moment at the cl. of
max MF: support MSA: kNm/m support MSM: kNm/m
max M F max M F
5m 5m
max MF
5m 5m
195
196
Table 4.3 Shear forces at a distance of d ¼ 17.5 cm from the face of the intermediate support
5m 5m
(b) Pin support of one node 28.0 (99%) 24.9 (99%) 3.1 (97%)
(c) Pin support of three nodes above the wall 28.1 (100%) 28.4 (113%) 0.0
(d) Fully restraint of the nodes at the face of the column 28.5 (101%) 28.5 (114%) 0.0
(e) Hinge coupling 28.1 (100%) 24.9 (99%) 3.2 (100%)
(f ) Bedding (C ¼ 9 106 kN/m3) 28.0 (99%) 25.8 103%) 2.2 (69%)
The results of the various models are similar if both spans are loaded, whereas in load
case 2, significant differences can be seen. A restraint of all nodes at the support
(Models B and C) results in much higher shear forces in the supported elements
(maximum þ13%) of the loaded span. The shear forces are much smaller than the
beam values in the unloaded span, and the forces are even zero in some models. This
is also the case if the elements over the support are bedded by spring or contact elements.
In conclusion, in order to model a support on walls, only the centre node should be fixed
to avoid numerical restraints.
In this section, the internal forces and moments calculated manually by a beam structure
have been used to verify the numerical model. It should be noted that the major
assumption of a beam model, a linear strain distribution over the supports, is not
valid. This is illustrated in Figure 4.15 where the stresses of a plane shell model
(diaphragm) of a one-way slab are shown near the intermediate support for a monolithic
connection and a free, non-restrained support (l ¼ 2 5.0 m).
A nonlinear strain distribution can be seen at the inner face of the support. High
compressive stresses are estimated for the fixed connection (singularity). In both cases,
the tensile stresses do not increase over the supports. Therefore, it is generally sufficient
to use the bending moments at the face of the support for obtaining the smoothened
values for design.
Even complex shell models can only approximate the real load-bearing and deformation
behaviour of a slab at the supports, as the basic assumption of a linear elastic material
behaviour is not valid in this region. The concrete slab will show some cracks in reality.
This support condition can easily be modelled by individual vertical springs or special
boundary elements that have no tensile stiffness. An iterative solution is required due
to this nonlinearity. Load superimposition is generally not valid in this analysis.
Figure 4.16 shows the bending moments, shear forces and deflections of a rectangular
slab (lx/ly ¼ 5.0/7.5 m) that is free to lift from the supports. It can be seen that approxi-
mately 20% of the supported length on both edges has been uplifted under uniform load.
The internal forces and midspan deflections of a rectangular, simply supported slab that
is fixed to the supports and one that can uplift, are given in Table 4.4. If the deformation
of the slab is not restrained in the vertical direction, the bending moments at midspan
increase by approximately 13%. The support forces per unit length increase (due to
197
Figure 4.16 Internal forces and deflections of a rectangular slab free to uplift from the supports
under uniform load
Part Part
Uplifting
Uplifting
2.7
Uplifting
Uplifting
0.6
10
9 7
21 7,5 0
9
7
10
24 24
54 52
24 –23
198
Table 4.4 Bending moments and deflections at midspan and shear forces at the supports of a
rectangular slab with and without restraint in vertical direction at the supports
my
lx = 5.0 m
where:
the reduced supported length) by approximately 38% (shorter edge) and 6% (longer
edge).
199
Interrupted
Top view
line support
Unrestraint node
Fixed node
application. Hence, three different models may be used to overcome the numerical prob-
lem (Figure 4.17). First, one may neglect the discontinuity of the line support in the
numerical model. The design of the opening region has to be done separately on a
fully or partially restrained equivalent beam system (‘flush beam strip’, see
Figure 4.18). This additional effort can be avoided when the opening is considered in
the numerical model. The nodes at the end of the line support may be fully or partially
restrained in the vertical direction. Another variation is to use a 3D volume model of the
slab and the wall underneath. However, this model again requires a big effort and is not
for practical use.
In the FE model, the missing line supports can be neglected if their length is smaller than
15 times the depth of the slab (leff/h < 15). The dimensioning of this region can be done
manually using an equivalent beam system (‘flush beam strip’) in accordance to ‘Heft
240’ (Grasser et al., 1991) (see Figure 4.18). More detailed investigations based on
FEM are only necessary if the opening length exceeds 15 times the section depth h of
the slab.
‘Theoretical’ singularities at the unsupported edge can be found if the wall opening is
considered by the FE model. These load peaks can be reduced significantly by a soft,
elastic support, by elastic bedding of the nodes or the elements, as the results of a
comparison calculation demonstrates. Figure 4.19 shows the internal forces for a soft
and a stiff support. A bedding modulus of C ¼ 2327 MN/m2 is used, which corresponds
to a concrete wall (grade C25/30) with a thickness of h ¼ 20 cm and a height of l ¼ 2.75 m
(C ¼ E . h/l ¼ 32.000 . 0.2/2.75 ¼ 2.327 MN/m2) (modulus of elasticity, see Table 4.5).
A large reduction in the moment peaks at the unsupported edges can be seen
(Figure 4.19), if the stiffness of the support is considered.
Bending moments for fixed and elastic supports and the values of Stiglat (Stiglat and
Wippel, 1983) are given in Table 4.6. Stiglat used the finite difference method and a
200
Figure 4.18 Beam system (‘flush beam strip’) and load influence areas for a discontinuous line
support, according to Grasser et al. (1991)
Simplified system:
leff fully restrained beam
Top view
Load influence
0.866 l
area
60°
Wall Wall
60°
Poisson’s ratio of ¼ 0.0 to estimate the internal forces in the region of the missing
support. With this assumption, the bending moment becomes nearly twice as high as
that of the fully restrained supported slab. This large difference may be caused by the
influence of the element size. The bending moments at midspan of the FE analysis
correspond well to Stiglat’s values.
Figure 4.21 shows the bending moment distribution of a slab with a partition wall. For
this structure, a singularity occurs at the face of the opening. To demonstrate the large
influence of the vertical stiffness of the support on the internal forces, a very small
bedding modulus was used (C ¼ 50 MN/m2), which correspond to a wall made of
hollow or aerated concrete bricks.
As shown in Figure 4.21, the peak of the bending moment and the large concentrated
shear force at the unsupported edge disappear. This load redistribution causes a
considerable increase of the bending moments and the deflections of the slab.
201
Figure 4.19 Bending moments and shear forces for a slab with a discontinuous line support
24.65
9.65
9.52
–21.88
10.52
25.21
(12.02)
–21.88
Elastic support
(–13.78)
–21.88
9.65
19
99
System
29
10 m
2m
43
80
108
108
5m
Elastic support
Slab thickness h = 20 cm
Concrete grade C30/37
Poisson’s ratio = 0.2
Loading: q = 10 kN/m2
202
Figure 4.20 Support forces of the outer wall near the face of the opening (nodes fully restrained in
the vertical direction)
System
Interrupted
support
2m
10 m
where:
fck is the characteristic concrete compressive strength in N/mm2;
0 is the permissible compressive strength of a brick wall.
l'
0.15l '
mxer
y
mxer
x
0.30l '
xer
mxrm
xm
ly
ym
lx
203
Bending Bending
moment moment
(12.05)
(9.85)
(–18.35) (–2.65)
10.02
8.35
–47.77 –36.43
(9.56)
-36.43
(6.83)
(–12.11)
7.00
8.55 (12.35)
Rigid 8.07
support (12.36)
–47.77
Deflections
System
Elastic Shown section
support
(1.33)
2.50 m
(1.00)
0.88
Line
5.0 m
support
(1.09)
(0.99)
2.50 m
0.65
0.28
5.0 m 5.0 m
Slab thickness h = 20 cm
Support forces at the inner wall – node fixed Concrete grade C30/37
Poisson’s ratio = 0.2
Loading: q = 10 kN/m2
The dimensioning of the slab can be done either with a stiff or elastic support. In the case
of a stiff wall, it should be noted that a redistribution of the bending moment at the
unsupported edge is only possible if the concrete slab cracks. Therefore, reinforcement
to limit the crack width should be inserted in the top face of the slab.
204
It must be emphasised that a flexible support of the slab can only represent the overall
deflection behaviour of the supporting wall. The complex 3D stress and strain distribu-
tion at the unsupported edges cannot be modelled with plain plate elements based on a
linear strain distribution.
A more refined model has to be used if this assumption does not apply. The joist has to be
discretised together with the slab in the same system. In this case, the real T-beam system is
idealised as a slab of constant depth and with an additional separate beam or a plate element
having a greater thickness, located at the midplane of the slab. In actual dimensioning, the
depth of the equivalent beam element has to be greater as the eccentricity of the joist is
neglected. The moment of inertia of the real T-beam should be identical to that of the
idealised system. The effective width of the flanges of the joist beff has to be considered in
this calculation. It may be estimated from EC2, Part 1 (Eurocode 2, 2004) as follows:
X
beff ¼ beff;i þ bw
0:2 l0
beff;i ¼ 0:2 bi þ 0:1 l0 4
bi
where:
The flexural stiffness of the FE model of the beam-plate system has to be identical to that
of the real structure. Thus, the depth of the equivalent beam hequiv
w has to be greater than
the sum of the depth of the web hw and the flange hf (see Figure 4.22). The width beff,i,
hw
bw
bw
205
T-beam
Moment of inertia of equivalent beam
1.2
Depth of the equivalent beam b1 b1
Moment of inertia of T-beam
which is part of the effective width beff, depends on the structural system and the loading.
However, it is not necessary to consider the change of beff in the longitudinal direction of
the beam in detail. It can be seen from Figure 4.23 that the depth of the equivalent beam
hequiv
w is nearly independent to that of the effective width beff.
The dimensioning of the joist is made by using the resulting forces of the equivalent beam
multiplied by the relation of the moment of inertia. This factor has to be considered,
especially if the depth of the web hw is small (see Figure 4.23).
Another approach for joist floors increases the height of the plate element instead of
an extra beam element. The relevant member forces of the T-beam have to be
calculated from the internal forces of the slab, as explained earlier, for the dimensioning
of the joist.
Furthermore, one can model the real load-bearing behaviour (folded plates) by means of
a 3D FE model (plane shell elements þ eccentric beam element). This alternative, much
better method will be explained in Section 5.3.
The numerical analysis has to model the actual behaviour of the pin support at the
columns. Different approaches are available, which will be discussed in the following
text.
206
Element
mesh
Top view
Columns
A A
Slab Nodes
Section A–A
Column Column
It must be emphasised here that the behaviour of the structure in the region of the columns
cannot be calculated exactly due to the underlying assumptions of the FE slab model. In
the supported area, the strain distribution is very complex. This behaviour cannot be
modelled by plate elements that are generally based on a linear strain distribution over
their cross-section depth. In Figure 4.25, the stresses of a circular slab are shown near
the supported area. The slab is simply supported at the outer edge and has a circular
column at its centre. The slab is subjected to a uniformly distributed load of q ¼ 10 kN/
m2. The axisymmetric model represents very well the load-bearing behaviour of a flat
slab for the region of the column. Even in the case of a rectangular arrangement of the
supports, the force distribution around the columns is nearly axisymmetric.
207
Figure 4.25 Normal stresses in a flat slab near the column (FE analysis)
Top view
-16.22
Pin support
Horizontal stress
m in the slab near
= 10 the column
dius
Circular Ra
column 0.20 m 0.20 m 0.20 m 0.20 m
10 cm
10 cm
Vertical stress
in the column
10 cm
h = 0.40 m
Radius = 10 m Radius = 10 m
Circular column 20 cm
= 0.40 m
An ‘exact’ determination of the internal forces and the stresses of a pin-supported slab is
only possible by means of a 3D (volume) model. The nonlinear behaviour of the concrete
due to cracking has to be considered. Such a complex model of a flat slab is too extensive
for practical use. Furthermore, the dimensioning cannot be done automatically by the
software.
Therefore, in practice, a 2D plane plate model is used, and the columns are modelled
by special support conditions. Various models, as shown in Figure 4.26, can be
used. These have already been explained in detail in Section 4.5 for a simply supported
one-way slab.
208
Figure 4.26 Flat slab – various models for the column support
Column
Top view
Rigid region
Top view
The main structural differences between the various models for the line support discussed
in Section 4.5 and that of a pin-supported flat slab is that the latter is a 2D spatial
structure, carrying the loads in two dimensions.
The ‘correct’ bending moments, shear forces and stresses of a flat slab in the region of the
column supports are unknown. Therefore, the results of the various models are examined
in a simple structure, using an interior panel of a flat slab with regular panel dimensions
of lx ¼ ly ¼ 5.0 m. For comparison purposes, the bending moment at the face of the
column is used. This value is relevant for dimensioning the slab. A uniformly distributed
load of g ¼ 10 kN/m2 is applied to the slab.
209
x Modelled
region
aaaaaa
lx
y
aaaaaa
aaaaaa
ly
Owing to the symmetric structural system and the load arrangement, it is sufficient to
only analyse one quarter of the slab and to consider the boundary conditions at the
symmetry lines (Figure 4.27). Therefore, in the following figures, only a quarter of the
whole slab is shown.
The results of the FE analysis are compared with the values estimated by the widely used
equivalent frame (Eurocode 2, 2004, Tables 4.7 and 4.8) and beam method (Grasser et al.,
1991). Here, the bending moments of a flat slab are estimated by means of an equivalent
frame or beam system and distributed in a transverse direction with factors given in the
tables for a column width of b/h ¼ 0.25/0.25 m and 0.5/0.5 m (see Figures 4.28 to 4.30).
The method given in Grasser et al. (1991) results in the following bending moments for
the inner slab:
Input variables:
" ¼ lx/ly ¼ 5.0 m/5.0 m ¼ 1.0
bsup/l ¼ 25/500 ¼ 0.05 resp. 50/500 ¼ 0.1
uniform load g ¼ 10 kN/m2
where:
bsup is the width of column
l is the span length.
210
Height of column
c2 c1
Column
above
l1
Torsional moment
MT
e
m
fra
Strip
le nt
l2
h u iva
eq
of
h
Column idt
Torsional W
beam below
Strip
0:224 1:0 52 ¼ 56 kNm=m ðcolumn 25=25 cmÞ
¼
0:160 1:0 52 ¼ 40 kNm=m ðcolumn 50=50 cmÞ
in the other strip beside the column:
mSG ¼ 0:7mSS ¼ 39:2 kNm=m resp: 28 kNm=m
in the field strip:
mSF ¼ kgSF g l12 þ kqSF q l12 ¼ 0:03 10 52 ¼ 7:5 kNm=m
Span moments
In the inner strip:
mFG ¼ kgFG g l12 þ kqFG q l12 ¼ 0:052 10 52 ¼ 13:0 kNm=m
in the field strip:
mFF ¼ kgFF g l12 þ kqFF q l12 ¼ 0:041 10 52 ¼ 10:3 kNm=m
211
Figure 4.29 Moment distribution of a flat slab according to Grasser et al. (1991)
l1
mSS mFG
(
1.25
MF
l2 ( (
1.25
MF
l2 (
0.1 * l 2 each
( M
1.4 SG (
( (
l2
Field strip MF
0.84
l2
considered direction of effective span
0.6 * l 2
l2
mSF
( (
MS mFF
0.5
M l2
0.1 * l 2 each
mFG
mss
½ inner strip
Moment distribution
beam or frame model
MS MS
MF
Region of the support moments
where:
k represents the moment factors (see Grasser et al. (1991); Tables 3.1–3.5)
The equivalent frame analysis (Eurocode 2, 2004) gives the following bending moments:
Table 4.7 Simplified appointment of bending moment for a flat slab#Data taken from Eurocode
2 (2004)
Note: Total negative and positive moments to be resisted by the column and the middle strips together should
always add up to 100%
212
Figure 4.30 Equivalent frame model of a flat slab according to Eurocode 2 (2004)
l x1 l x2 l x3 l x4
lyi/4
l y1
Column
strip
Middle
strip l y2
lyi/4
l y3
l y4
lx > ly
Equivalent frame
A B C D E
213
The calculated negative bending moment in the column strip msup,max ¼ 33.3 kNm/m
is significantly less than the value from the equivalent beam method msup,max ¼
56.0 kNm/m.
The distribution of the bending moments in the direction of the main load bearing is
treated first (see Figure 4.29 bottom). Next, we consider the distribution of the bending
moments in transverse direction (see Figure 4.29 right).
The moment distribution of the models with 8 8 and 16 16 elements are well matched
for the span and up to x/l 0.06 from the centreline of the pin support. Consequently, a
further refinement of the element mesh would only slightly affect the moments at
x > 0.06l. Both curves meet in the region of x/l 0.02.
The bending moment at the face of the column is needed in the design. Therefore, the
considerable differences in the peak moment at the pin support (mS ¼ 241/
471 kNm/m) are of no practical importance. As illustrated in Figure 4.31, the
results are highly dependent on the element size in the region of x/l < 0.02 and bc /
l < 0.04. Therefore, a pin-supported node should not be used for a column width
of bc < 0.04l. Such slender columns (bc < 20 cm for lslab ¼ 5 m) are rarely used in
practice.
Furthermore, one can see that the greatest bending moments at the supports for
all meshes correlate well with values calculated by the analytical method given in Grasser
et al. (1991). This is not surprising, since the moment factor k had been estimated for a
pin-supported plate.
The distribution of the bending moments in the symmetry lines for two different
cross-sections of the column (bc ¼ 25 cm and bc ¼ 50 cm) is shown in Figures 4.32 and
214
Figure 4.31 Moment distribution myy for y ¼ 0 in the region of an interior column for different
element meshes – pin-supported node
1000*m
q*l 2
51.44 (50 × 50 el.)
100
51.64 (16 × 16 el.)
50 52.44 (8 × 8 el.)
0.025 55.36 (4 × 4 el.)
0
0.1 0.2 0.3 0.4 (x /l ) 0.5
–50
–100 4 × 4 elements
–150 y
–200 8 × 8 elements
–241
–250
l/2
myy
–300 –315
x
16 × 16 elements
–350 l/2
–383 50 × 50 elements
–400
–471
1000*m
q*l 2
100
50
0.01 0.05 0.10
0
(x /l )
–50
4 × 4 elements
–100
50 × 50 elements
–150
8 × 8 elements
–200 Heft 240 for bc/l = 0.1
–241 Heft 240 for bc/l = 0.05
–250
16 × 16 elements
–300 –315
–350
–383
–400
–471
4.34. It can be seen that the bending moments in the span are only slightly dependent on
the width of the column. In contrast, the internal forces at the face of the column increase
considerably when the width of the column becomes smaller. This is caused by the
concentration of the support reactions.
215
Figure 4.32 Element mesh, bending moments and deflections of a flat slab – all nodes above the
column fixed in vertical direction
2.5 m
2.5 m
Deflections
Support
(20 nodes
fixed)
myy m xx
–53 –40
–40
–53
As illustrated in Figure 4.37, the size of the elements has a great influence on the
maximum support bending moment. The bending moments near the face of the
column (maximum value) are compared for various mesh refinements with the values
calculated according to Grasser et al. (1991).
where:
mSG is the support bending moment in the inner strip
mSF is the support bending moment in the field strip
mFF is the span moment in the field strip
mFE are the moments from FE analysis
m240 are the moments according to Grasser et al. (1991).
216
Figure 4.33 Numerical restraint caused by the pin support of all nodes above the column
A Section A–A
Top view
Column
Slab
A
Figure 4.34 Bending moment myy and mxx in the section y ¼ 0 – pin support of all nodes above the
column
1000*m
q*l 2
100
0
0.1 0.2 0.3 0.4 0.5
(x /l )
–18
–50
–64 myy-column 50 × 50 cm
–90
–100 mxx-column 50 × 50 cm
y
–142
–150 myy-column 25 × 25 cm
–151
–164
l/2
myy
mxx-column 25 × 25 cm
–200 x
–210
DAfStb Heft 240 l/2
–250
217
Figure 4.35 Bending moment myy and mxx in the section y ¼ 0 – all nodes above the column are
fully restrained
1000*m
q*l 2
–50
myy-column 50 × 50 cm
–100
y
myy-column 25 × 25 cm
–150 –164 mxx-column 50 × 50 cm l/2
myy
–200 x
mxx-column 25 × 25 cm l/2
–250
Figure 4.36 Bending moment in the symmetry lines – all nodes above the column are fully restrained
myy mxx
– 62
– 34
– 34
– 62
218
Figure 4.37 Bending moment near the face of the column for different refined element meshes;
column size 25 25 cm and 50 50 cm – deflection of all supported nodes fixed
Figure 4.38 shows the moment distribution in the column axis for a bedding modulus of
C ¼ 9000 MN/m3. 40 40 elements are used for a quarter of the slab. The column has a
cross-section of 25 25 cm. Hence, 2 2 elements and 4 4 elements for a column of
50 50 cm are bedded. The results of the numerical analysis do not change much if
the element mesh is further refined.
The desired smoothed distribution of the bending moment over the column can be seen
in Figure 4.38. The maximum bending moment decreases with the size of the column. A
good correlation of the results with the values provided by ‘Heft 240’ Grasser et al. (1991)
is obtained.
As shown by the parametric study with different bedding modulus C, only the greatest
bending moment at the support mS,max is effected by the stiffness of the support (see
Figure 4.39). As C changes, the bending moment at the face of the support keeps
nearly constant, whereas it decreases in the centre of the support with an increase in
C. The distribution of the internal forces becomes similar to that for all nodes being fixed.
219
Coupling Coupling
Column Column
Z Z
Rigid Restraint
diaphragm
Deflections Deflections
in Section 2.6.3. A stiff and hinge coupling can be used for flat slabs. In the case of fixed
coupling, all nodes of the supported area are restrained in the vertical direction
(vz ¼ vz0 ¼ 0). Thus, the rotation of the elements is prevented. The slab is fully restrained
for uniform loading.
Figure 4.38 Bending moment in section y ¼ 0 – elastic bedded elements
1000*m
q*l 2
100
Face of column 50 × 50 cm
50
Face of column 25 × 25 cm
0.1 0.2
0
(x /l )
0.3 0.4 0.5
–50
–100 y
–150 –149
bc/l = 0.10
–212 column 50 × 50 cm
l/2
–200 myy
x
–250
bc/l = 0.05 – column 25 × 25 cm l/2
–300
220
Figure 4.39 Bending moment in section y ¼ 0 – elastic bedded elements with different bedding
modulus C
Column 25 × 25 cm
1000*m
q*l 2 0.04 0.08 0.12 0.16 (x /l ) 0.2
0
Face of column 25 × 25 cm
–50
–100
y
–150 C = 9 × 108 kN/m3
1000*m
Column 50 × 50 cm
q*l 2
0.04 0.08 0.12 0.16 (x /l ) 0.2
0
–20
Face of column 50 × 50 cm
–40
C = 9 × 108 kN/m3
–60
–80
y
–100
–120
C = 9 × 107 kN/m3
–140 myy
l/2
–160
x
–180 C = 9 × 106 kN/m3
l/2
–200
C = 9 × 105 kN/m3
221
Therefore, a stiff coupling should not be used for asymmetric load arrangements, as has
already been demonstrated in Section 4.5.
In contrast to the previously mentioned model, the rotation of the nodes (’x, ’y, ’z ) are
not combined for a hinged coupling. Only the deflections are coupled with the master
node, according to the expressions listed earlier. This model is similar to a stiff coupling
for a symmetric system and loading, since the rotation of the slab in the centre of the
supported area with respect to the master node (’x0 ¼ ’y0 ¼ ’z0 ¼ 0) is equal to zero.
This statement is confirmed by the results shown in Figure 4.40. There are no internal
forces and no bending moments for the elements in the supported area. Hence, this
model does not give a smooth bending moment distribution at the support. The results
shown in Figure 4.40 demonstrate that the bending moment at the face of the column is
nearly independent of the element size.
As expected, the model of the column support has only a minor influence on the field
moments, as can be seen in Figure 4.41. The smallest bending moment in midspan calcu-
lated with FEM is 40% less than that of the simplified analytical model. It is only the
values in the support strip that are different for the various models (see support strip,
Figure 4.41 top). The support strip moments for the model with 3 3 fixed nodes and
the stiff coupling are some 25% lower than the values provided by Grasser et al. (1991).
Overall, a very good correlation is found between the different models. This is because of
the use of the simple system and uniform loading.
222
1000*m
q*l 2 Column 25 × 25 cm
100
50
–50
y
–100
–150
Maximum bending moments:
l/2
myy
–216 (20 × 20 elements)
–200 –207 (40 × 40 elements) x
–209 (80 × 80 elements)
l/2
–250
1000*m
q*l 2 Column 50 × 50 cm
60
40
20 0.1
0
0.25 (x /l ) 0.5
Face of column 50 × 50 cm
–20
–40
–60 y
–80
–100
l/2
223
Table 4.9 Span and support bending moments for a flat slab according to different models
(bc ¼ 25 25 cm, lx ¼ ly ¼ 5 m, 40 40 elements)
Inner strip 1 (face) mSS 56.0 (44.8)* 63.3 41.2 59.1 41.2
(100%) (113%) (74%) (106%) (74%)
The node in the centre of the supported area is fixed in the vertical direction. Thus, a
small cantilever slab is created that shows bending moments perpendicular to the free
edge.
The following calculations are carried out on a simple system, a rectangular flat slab with
equal span length lx ¼ ly ¼ 5.0 m supported by four edge columns (b/h ¼ 25/25 cm). Due
to this being a symmetric system, only a quarter of the whole slab has to be modelled. A
uniform load of q ¼ 10 kN/m2 is applied.
Bedding moduli between C ¼ 9 106 kN/m3 and 27 106 kN/m3 are used. The first value
corresponds to the normal stiffness CN ¼ E/l of a 3.55-m-high column with a Young’s
modulus of Ec ¼ 32 000 MN/m2 (concrete grade C30/37). If the bending stiffness is
considered, a value of CN ¼ 3E/l ¼ 27 106 kN/m3 has to be used.
Figure 4.43 shows the bending moment distribution myy at the edge of the slab ( y ¼ 0) for
different models. The moment at the face of the column and at midspan is very sensitive
to the numerical model of the support by the edge columns.
224
Modelled zone
l = 5.0 m
of the slab
l = 5.0 m
Figure 4.41 Transverse distribution of the bending moment perpendicular to the main load
transfer direction according to Grasser et al. (1991) (shaded area) and FEM
m in kNm/m
–70 Pin support –118 –70
Pin support –118
Elastic support –59
–60 Elastic support –59 –60
–56
–56
–50 Rigid body –50
–41 3 × 3 nodes –41
–40 –39 –39 –40
–10 –10
–7.5 –7.5
0 0
0.10 0.20 0.30 0.50 0.60 0.90 x/l 1.00
0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 x/l 1.00
2 2
8.3
5 5
10
10
10.3
13.0 Finite Elements
m in kNm/m DAfStb
225
M M
B B
l
h
A A
B = M l B = M l
B = B =
B = 3EI B = 4EI
M = CN · I · M = C ·
where: CN = bedding modulus of the slab C = 3E · I / R C = 4E · I / R
I = second moment of area of the column
M = CN · I · = C · CN = C/ R CN = 3 · E / R CN = 4 · E / R
Figure 4.43 Bending moment myy in section y ¼ 0 of an edge supported slab for different bedding
moduli C
1000*m
q*l 2
200
Elastic support C = 9 × 106 kN/m3
150 134
119
100 104
Pin support 69
50
All nodes fixed
0
0.1 0.2 0.3 0.4 (x / l ) 0.5
–46
–50
–92 y
–100
Elastic support C = 27 × 106 kN/m3
–153
–150
l/2
–200 myy
x
–250
l/2
–300
226
The span moment at the outer edge of the slab is 30% greater, if bedded elements are
used instead of a pin support. Significant differences can be observed at the relevant
sections for design at the face of the column. The extreme values are m ¼ 0 kNm/m to
m ¼ 38 kNm/m.
These differences become clear when the 3D distribution of the bending moments
above the column is checked (Figure 4.44). The peak values are not located on the
symmetry axis. The largest bending moment is between m ¼ 14 kNm/m for bedded
elements with C ¼ 9000 MN/m3 and m ¼ 73 kNm/m for fixed nodes.
Figure 4.44 Bending moments at an edge column for different bedding moduli C (thin shell
elements)
20 80
70
15 60
50
10 40
5 30 5
5 4 20 4
0 3 10 3
2 0
–5 –10 2
1 2 1 1
3 4 2 3 1
5 4 5
25 14
12
20
10
15 8
10 5 6 5
4 4 4
5
3 2 3
0 0
–5 2 –2 2
1 2 1 1 2 1
3 4 3 4
5 5
System
Column
Bending moment distribution 25/25 cm
over the column
m m
5.0
5.0 m
227
(a) footings: used for high concentrated forces (e.g. those resulting from column
loadings)
(b) strip foundations: used underneath walls or series of columns
(c) foundation slabs: used under buildings as load-bearing elements and for
watertight structures.
The design of a strip foundation in the direction of the longitudinal axis can be done by
means of a flexible supported beam model. The member forces in the transverse direction
can be determined either analytically, if the deformation of the structure can be
neglected, or by means of a 2D plate model (see Section 4.10.2).
The internal forces of a foundation slab are mostly estimated by means of the
stiffness modulus method. The inaccuracies of this approach have already been discussed
in Section 2.4.1. The main advantage of this model lies in the fact that the bedding
behaviour can easily be implemented within the element functions. Thus, no contact
or interface elements are required, which simplifies the mesh generation and the numer-
ical analysis. Furthermore, a 2D plane FE model is sufficient, whereas the constraint
modulus method requires a 3D model of the structure and the ground or an iterative
solution. The following calculations are based on the stiffness modulus approach.
4.10.1 Footings
Footings are used under high concentrated loads, for example, those caused by columns.
The panel dimensions necessary for these foundations are mostly determined by the
permissible soil pressures. The slab depth is fixed by the design for bending or by
shear (punching). For small loads, massive block foundations can be used that may be
not reinforced. Such structures can be designed by means of strut-and-tie models.
Numerical analysis is not required.
The following example focuses on the design of the foundation slabs. The internal forces
can be calculated by means of an FE model or simplified analytical approaches, such as
the equivalent strip method given in Grasser et al., 1991. Hence, different models are
available to verify the results of the numerical analysis.
The columns, which transfer the load to the foundation, can be modelled with different
approaches such as those used for flat slabs. The variants are (Figure 4.45)
228
Figure 4.45 Moment distribution of a single slab foundation in the central axis
1890
Concentrated
load
Loaded area
50 × 50 cm
Loaded area 1010 950
1.5 × 1.5 m
DAfStb 869
Nodes of the
elements under the Heft 240 700
column coupled 630
500
282
350
279
Bending moments
262
Face of column
190 MPa
Soil pressure
System
F = 5000 kN
Column Concrete grade C 25/30
5.0 m
50/50 cm
Bedding modulus C = 75 MN/m3
1m 20 × 20 plane elements
g hinge coupling of the elements in the column area with the node at the centre of
the column (coupled area is equal to column dimensions, 50/50 cm).
The bending moment distribution in the symmetry axis calculated for the different
models are presented in Figure 4.45. The rectangular slab with lx ¼ ly ¼ 5.0 m is loaded
by a concentrated force of F ¼ 5000 kN. The stepped curve is calculated with the
simple analytical approach given in Grasser et al., 1991. With the given FE model, a
single concentrated load results in a very high moment peak (singularity) of
mmax ¼ 1890 kNm/m. If the load is distributed over the area of the column
(mmax ¼ 1010 kNm/m ¼ 100%), it results in only half of this value. A coupling of the
elements under the column results in a further decrease in the maximum moment by
approximately 14%. A bending moment of mmax ¼ 630 kNm/m ¼ 62% is obtained
when the column load is dispersed up to the midplane of the foundation slab at an
angle of 458, resulting in a loaded area of 1.50 1.50 m2. It can be observed that the
size of the loaded area has a significant effect on the bending moment distribution. On
the other hand, the soil pressure is not sensitive to the loaded area in this example.
229
2.0
2.0
4.5
2.5
3.0
3.5
3.0
2.5
3.5
4.0
2.0
1.5 3.5
1.5
3.0
2.5
1.0
1.0 1.5
2.0 1.5
The correct model respectively the correct loaded area can only be found by verification
of the used method with experimental results or by means of a 3D FE analysis. The main
bending moments show an axisymmetric pattern in the column region, as can be seen in
Figure 4.46. Thus, an axisymmetric model (diaphragm), and not a 3D model, is sufficient
to determine the correct linear elastic values. Figure 4.47 shows the main membrane
forces and Figure 4.48 the horizontal stresses in three sections of a circular foundation
slab with a diameter of d ¼ 5 m.
The numerical integration of the horizontal stresses at the face of the column results in a
bending moment of mmax ¼ 637 kNm/m. This value is similar to that of the slab
model with a loaded area of 1.5 1.5 m2. However, such agreement seems to be only a
coincidence, since the essential assumption of a plate model, a linear strain distribution
over the width of the slab, does not apply to most parts of the structure (Figure 4.48). The
external load is carried by bending and by inclined compression struts. Strut-and-tie
models should be used for the design of such a structure. The compression struts form
a truncated cone for the given axisymmetric foundation slab (Figure 4.49).
230
Test results are used in the following to verify the earlier-mentioned results. Figure 4.50
shows the measured and calculated steel tensions in the reinforcement and the soil
pressure on the support and the field strip of a footing. The footing has a height of
0.32 m and a square base of 1.50 m 1.50 m. In the test, the footing was loaded by a
uniform load on the slab bottom instead of a single force in the column. The measured
–4.15
0.06 –0.71 –8.29
Centreline
231
3926 kN
Compression
cone
d
90 cm
steel stresses are smaller in the supporting strip than at the face of the column (column
strip). Therefore, the face of the column is relevant for the design of the footing and not
the middle of the column.
In the right figures, the results of the FE analysis are given for different loaded areas. The
footing was modelled with 20 20 plane elements. For the calculation of the steel
tensions from the bending moments, a steady inner lever arm from z ¼ 0.9d ¼ 0.27 cm
was used. As one sees from the curves on the right in Figure 4.50, the relevant
bending moment in the face of the column can be determined with sufficient accuracy
within the scope of the measuring tolerance with all models. The point-load should
not be used, however, because the maximum moment increases with a refinement of
the element mesh and, at the same time, the relevant moment at the face of the
column decreases. In the field region, slightly conservative values are determined with
almost all FE models.
The bending moment distribution for the various loaded areas is shown in Figure 4.52. It
must be noted that a uniform continuous load does not cause a moment peak. The
loaded area slightly changes the maximum bending moments. There is a difference of
approximately 15% in the mean value.
232
Figure 4.50 Stresses in the reinforcement and soil pressure of a footing – test results against FE
analysis
30 cm
600
F = 1350 kN
σB = 600 kN/m2
400
a/b = 150/150 cm
100 100
0 0
y = 0.03b (column region) y = 0.03b (column region)
200 200
Test results
100 100
50 50
0 0
y = 0.29b (field region) y = 0.29b (field region)
q = 1000 kN/m
Elevation 1.0 m
Top view
1 10 20 26
Shell
elements
5.0 m
233
Figure 4.52 Bending moment distribution of a strip foundation (plane shell model)
5.0 m
The same system has been analysed with a shell model. The results are shown in
Figure 4.54. It can be seen that the distribution of the horizontal stresses over the
cross-section depth is almost linear. The numerical integration of the horizontal stresses
results in a bending moment of mmax ¼ 449 kNm/m. This is similar to that of the shell
model with coupled nodes. Thus, extending the loaded area up to the midplane of the
slab at an angle of 458 results in unsafe bending moment distribution. This model
should not be used.
A tensile force of F ¼ 500 kN(125 12.5/90) ¼ 625 kN/m ( ¼ 398) can be calculated
from a simple strut-and-tie model (Figure 4.53). This force is greater than that of the
plate model, as a uniform soil pressure distribution has been assumed for simplicity.
0.5 m
1.0 m
234
Membrane forces
Membrane force in
horizontal direction in kN/m 2817
2900 3010
1302 2816
Settlement in mm
2.83 mm
2.45 mm
5.0 m
the support condition and the loaded area. In practice, more complicated slab systems
are built that have complex support conditions and irregular layout (Figure 4.55). It is
one of the most important advantages of concrete that structures can be built in arbitrary
shapes.
Skewed slabs, where two parallel edges are simply supported and the other edges are free,
are often used for single-span bridges. A line support results in infinite shear forces and
bending moments in the obtuse corner, as can be seen in Figure 4.56, where the distribu-
tion of the internal forces at a simply supported edge are presented. The slab has a width
of 13 m and an angle of 458 to the supported axis. The singularity of the support forces
can be seen clearly.
235
Figure 4.55 Triangular slab of an office building (details see Section 6.2)
Column
Column 24/24 cm
every 1.50 m
The peak moments can be reduced significantly if the stiffness of the individual bearings
is considered. In Figure 4.56, the internal forces are plotted for a slab with three
elastomeric bearings (50 25 cm2, height h ¼ 10 cm) closed to two parallel edges. The
main shear force in this system is one-third of that of the simple continuous supported
slab. The bending moment myy in transverse direction is increased from 234 kNm/m
to 409 kNm/m. The transverse bending moment can be reduced if the number or the
size of the bearings is increased.
The internal forces at the midspan are almost independent of the support conditions, as
shown by the figures given in Table 4.10.
236
Figure 4.56 Bending moment and main shear force at the supported edge of a skewed slab
Elastomer
(–35) bearing
Continuous
line support
y myy
12 m
(–409)
45°
2135
12 m
Elastomeric bearings
4.12. Singularities
Singularity problems may arise in linear elastic slab calculations, as has been previously
shown. Some problems can be avoided by a flexible support or an extension of the loaded
area. Singularities may occur in the regions of (Figure 4.57 and Table 4.11)
237
Re-entrant
corner
Re-entrant
corner Discontinuous
line support
Pin
support Discontinuous
line support
Opening
Knife-edge support
Two more singularity regions that are relevant in practice are treated next
g openings
g re-entrant corner ( 5 908)
g concentrated loads.
The singularity is caused by a sudden change of the boundary conditions at the corners of
the unsupported edges. There are no bending moments in the direction of the free
238
free edge
hinge support
fully restraint
F
No Yes (mx, my, vx, vy)
A
y M
dw
A Yes ’y ¼ Yes (mxy, vy)
x dx
y
No Yes (vx)
A A
x
239
Figure 4.58 Moment distribution near the corner of an opening (FE results)
Detail shown
left
mx = 0
mx
mx 0
mx = 0
boundary, whereas bending moments perpendicular to the boundary may occur. This
can be seen in Figure 4.58. The bending moment mxx is not equal to zero at the left
side of the corner, whereas it has to be equal to zero at the upper and lower side of
the opening. This inconsistency leads to infinite internal forces and moments. An FE
model always gives finite results. The maximum peak value depends on the size of the
relevant elements. Only a small region near the corner is influenced by the singularity
(Figure 4.58).
The distribution of the principal bending moments and shear forces in the region of an
opening is presented in Figure 4.59. The high increase of the shear force in the corner
indicates a singularity problem. The shear design of the slab is not possible (and not
required) for these peak values.
Figure 4.59 Main bending moments and shear forces in the region of an opening
mx,r = 0 Y 90
X 10 10
my,r = 0
10
1m
240
Singularity problems can be avoided by smoothing the edges, by rounding the corners in
the FE model. However, this effort is not required in practice, as shown by the previous
analyses.
Figure 4.60 Internal forces of a slab that is supported by two perpendicular walls – no supports at
the outer edges
207
Stiff support
50 25
1m
65
Elastic support
25
241
Figure 4.61 Moment distribution of a rectangular slab loaded by a point-load at the midspan
X Y
moments are reduced by a factor of 3 and 2, respectively, relative to those of a slab with
rigid supports.
Concentrated loads may occur, for example, on deck slabs of hollow box or T-beam
bridges. The internal forces due to truck loading are calculated in the following example
using a simplified system of a partially restrained equivalent slab (Figure 4.62), as in
practice a 3D FE model of the whole structure is too extensive. A detailed estimation
of the effective span length is not required with regard to the assumptions of the
model. The rotational stiffness of the webs may be estimated by separate manual
analysis, for example, using an equivalent beam model. It depends not only on the
structural system, but also on the type of the loading.
The point-load has to be dispersed up to the midplane to avoid singularities (Figure 4.63).
242
Figure 4.62 Equivalent slab model for the deck of a hollow box girder bridge
Cross-section
leff = 6 m leff
Detail ln Detail 0.5h1 ln
h1
1.2 m
Action
e.g. wheel load
Pavement
Deck slab
243
Figure 4.64 Moment and shear force distribution of a plate strip loaded by a force of F ¼ 100 kN at
midspan – different loaded areas
14.6
54.6 83.5
–16.8 28.0
–17.8
The significant influence of the loaded area on the bending moments can be seen in
Figure 4.64. Here, the bending moments for a total force of F ¼ 100 kN and a loaded
area of 0 0 cm2, 24 20 cm2 and 60 96 cm2 are plotted. The slab has an effective
span length of leff ¼ 6.0 m and a constant thickness of h ¼ 20 cm. The inclined haunches
near the webs are neglected.
An FE analysis of the deck slab is required if more accurate internal forces are needed for
design or in the case of special load arrangements. Tables published in literature and
influence lines provided by Homberg and Ropers (Homberg, 1973; Homberg and
Ropers, 1965) can be used for normal conditions. The determination of the internal
forces by means of tables may be time-consuming, as various values have to be
interpolated; thus, an FE analysis may be economical.
244
Figure 4.65 Principal shear forces at the centreline of the support and at a distance of 0.2 m from
the support for 5 8 and 60 90 elements
9.4 145.
3.7 22.
5 × 8 elements 60 × 90 elements
at the support at the support
the shear force distribution along the supports can be recognised. The high values in the
corners result from a model problem. A sharp corner of a simple supported rectangular
slab is a singularity region, as can been seen in Figure 4.65. Hence, one does not need to
design the slab for such high shear forces.
In Table 4.12, the principal shear forces, as well as the concentrated edge force Fe, are
listed for both element meshes. It can be seen that the shear forces cannot be calculated
with a broad mesh of 5 8 elements. Furthermore, it becomes clear that the maximum
values in the middle of the support at 1.0d as well as the edge force are not sensitive to this
discretisation.
Table 4.12 Principal shear forces and concentrated edge force for different element meshes in
kN/m or kN
* Short/long span
245
ϕt ϕn mn mt
A w vz
Section A–A
‘soft’ support: ϕn Þ 0 ‘stiff’ support: ϕn = 0
Up to now, the nodes at the supports were fixed only in the vertical direction. This model
is called ‘soft’ support (Figure 4.66). The rotation (torsion) at the support ’n is not equal
to zero as it should be with a stiff simple support. In contrast, ’n is fixed at a ‘stiff ’
support. The differences of both models are mostly negligible, but not in the area of
the corners, as will be illustrated in the following.
‘Stiff ’ support w ¼ 0 ’n ¼ 0 ’t 6¼ 0 vz 6¼ 0 mt 6¼ 0 mn ¼ 0
‘Soft’ support w ¼ 0 ’n 6¼ 0 ’t 6¼ 0 vz 6¼ 0 mt ¼ 0 mn ¼ 0
The shear forces at the support and at a distance of 0.2d for a ‘soft’ and a ‘stiff ’ support
are plotted in Figure 4.67. The big differences in the corners are easy to recognise. With
the ‘stiff ’ support, no shear peaks are estimated, and the edge force is close to zero.
The fixation of the rotation ’n at the edges leads to a fully clamped support in this
region, which will rarely be the case in reality. Also, with this support condition, the
real behaviour of a slab near the corners cannot be modelled. Better results can be
achieved with a 3D model with volume elements. Due to symmetry of the slab and the
loading, only a quarter of the slab must be discretised. Approximately 100 000 elements
were used. Figure 4.68 shows that the support forces of the volume FE model are
identical to the shear forces in the slab.
In middle of the span, a good agreement can be seen (Figure 4.67), whereas the concen-
trated edge force Fe ¼ 8.4 kN is considerably lower than for a simple supported slab.
An automatic mesh generation can only consider the main geometrical boundaries.
Refinements are required in regions of great stress gradients and high concentrated
246
60 × 90 elements
at the support
0.2 m from
the support
4.4
126.
16.5 16.3
Fe = –26.3 kN Fe = +0.1 kN
loads, and have to be done manually by the user, if no adaptive mesh-refinement tools
are available. Hence, the mesh generation requires sufficient experience and skills in
use of FEM and knowledge of the material properties too.
The problems of mesh generation are demonstrated by the following two real examples,
where the author was involved as a designer and a checking engineer.
–8.4 –8.4
23.
23. 23.
46.
23. –8.4
Fe = –8.4 kN
247
6.50
1.30
42
7.10
20
23.55
30
14.60
2.00
1.00
80
4.95
45
1.00 5.10
80
the slab even in the case of hinge support. This problem is demonstrated in a simple
example.
Circular slabs are often used in tanks or bins, like the 48-m-high silo structure shown in
Figure 4.69, which is used for storing fly ash. Large openings are necessary in the lower
circular slabs to install discharge devices. In addition, high dynamic point-loads may
occur due to the machines. The manual analysis of such a slab becomes very difficult.
FEM can help us design these slabs.
The problem of mesh generation is not demonstrated for a real complex structure, but
using a simplified system, for example, a circular slab (radius r ¼ 7.50 m) which is
simply supported at the outer edge. A uniform load of q ¼ 10 kN/m2 is applied on this
structure (Figure 4.70). The internal forces and deflections can be estimated analytically.
These values are used to verify the results of the FE analysis. The bending moment in the
centre of the slab is:
q r2 10 7:52
mI ¼ ð3 þ Þ ¼ ð3 þ 0:2Þ ¼ 112:5 kNm
16 16
248
Figure 4.70 Circular slab – boundary nodes not accurately placed on the outer circumference
m
50
7.
Line support
loading : q = 10 kN/m2
0 70
11
80
40
0
10
90
80
80
40
70
60
Figure 4.70 shows the FE mesh and the resulting main bending moments. Due to the
symmetry conditions, it is sufficient to only model a quarter of the whole slab.
The orientation of the main bending moments at the upper edge of the slab is surprising.
A support moment of mI ¼ 80 kNm/m has been calculated. This moment is 70% of the
maximum bending moment in the centre, even for a structure that is simply supported.
This error is caused by the faulty location of the nodes at the support. The fixed nodes
at the upper right of the slab are located in a sawtooth manner, 1 mm away from the
exact circle. Even this small value, equivalent to 1/1500 of the diameter, has a significant
influence on the bending moments. The same problem arises in rectangular slabs, for
cases where the fixed nodes are not located accurately on a straight line.
249
This example has impressively shown that the fixed nodes should be precisely located at
the curved boundary.
Very often, only the graphical output of an FE analysis is checked. Errors in node co-
ordinates, like those referred to earlier, can then hardly be noticed. Therefore, the checking
of an FE analysis should not be restricted only to the graphical output of the software.
Figure 4.71 shows the bending moment distribution and the deflection in the centre of the
slab for different numbers of elements. The correct values can be estimated analytically.
m
50
7.
r=
1 2 3
250
Figure 4.72 Deflections and bending moment in the centre of a circular simply supported slab for
different number of elements
wFEM MFEM
wanal Manal
Deflection
in midspan
1.0
0.9
Bending moment
in midspan
0.8
0.7
0.6
Number of
0.5
0 10 20 30 40 50 elements
The results summarised in Figure 4.72 demonstrate that only 4 6 ¼ 24 elements are
required for the whole slab in order to get satisfactory results at the centre.
Nowadays, due to the increasing capacities of computers, the size of the numerical model
or the number of elements does not usually play an important role. Therefore, modelling
is often done quickly, without sufficient attention being paid to the relevant details. What
may be considered as unimportant details, such as a one-way cantilever slab or a small
opening in a flat slab near a column support, are not noticed. However, these are the
regions where high gradients of the internal forces may occur and which are relevant
for design.
The problems that occur due to neglecting such ‘unimportant’ details are demonstrated
by a simple cantilever slab. This is not a theoretical example. Such a problem had
occurred in a real big slab, where, unfortunately, the errors in the FE analysis were
noticed only after the slab had been poured already with insufficient reinforcement.
Excessive retro fitting was required. The system and the element model are shown in
Figure 4.73 (full system shown in Figure 4.4). The balcony had been modelled with
only one element layer, as it has a simple one-way load-bearing behaviour. A uniform
load of q ¼ 5 kN/m2 is applied on the whole slab. This results in a support bending
moment of the equivalent cantilever beam of ms ¼ 5 2 1 ¼ 10 kNm/m. Values
between ms ¼ 3 and ms ¼ 6 kNm/m, which are much smaller than the analytical
ones, are calculated by FE analysis. The differences are caused by the linear form func-
tions of the four-noded elements used, which results in constant internal forces within an
element. Thus, only the values in the centre of the element are calculated, and not the
values required for design at the boundaries. The dimensioning of the slab, based on
251
The mesh problems may be noticed by the contour plots of the bending moment my
(Figure 4.73). A large discontinuity between the elements can be seen. However, these
obvious differences only occur in this example since the size of the elements in the
195 221 237 254 271 288 305 322 339 1425
193 219 236 253 270 287 304 321 338 1424
192 218 235 252 269 286 303 320 337 1423
191 217 234 251 268 285 302 319 336 1422
190 216 233 250 267 284 301 318 335 1421
189 215 232 249 266 283 300 317 334 1420
188 214 231 248 265 282 299 316 333 1419
187 213 230 247 264 281 298 315 332 1418
186 212 229 246 263 280 297 314 331 1417
2m
A
Bending moment my Bending moment my
in section A–A
–6.16 –6.0
7.0
6.0
–5.0
5.0
4.0
–4.0
3.0
2.0
.0
–1.0
.0
–3
–2
1.0
–3.02
252
inner slab is much smaller than in the cantilever slab. The discontinuity would not occur
if the size of the elements were similar.
The size of the elements is relevant in regions with big gradients in the inner forces or
deflections. These problems are demonstrated with an easy example, a one-way slab
with uniform loading (l ¼ 2 5 m). Figure 4.74 shows the bending moments and shear
forces for a different number of elements.
With ten elements per field, a maximum support moment of ms ¼ 16.8 kNm/m is calcu-
lated, which nearly corresponds to the analytical value in the distance from x ¼ 0.25 m of
Figure 4.74 Bending moments and shear forces for a one-way slab for different elements
–30
q = 10 kN/m –27.0 (40 el.)
–25 –23.4 (20 el.)
5m 5m
–20
–16.8 (10 el.)
Bending moment in kNm/m
–15 40 elements
20 elements
–10 10 elements
–5
0 1 2 3 4 5 6 7 8 9 10
0
10
15 15.7
(10 el.) 17.3 (40 el.)
20
10
0
0 1 2 3 4 5 6 7 8 9 10
–10
q = 10 kN/m 40 elements
–20
20 elements
10 elements
5m 5m
–30
253
the intermediate midsupport. The maximum moment increases by about 40% with 20
elements (ms ¼ 23.4 kNm/m) and by about 61% with 40 elements (ms ¼ 27.0 kNm/
m). Hence, the number of elements is important in the case of small support widths.
Please note that the face of the supports is relevant for the design and not the centreline.
The curves of the member forces are quite similar for all element meshes except at the
intermediate support.
The amount of elements is not so significant if the shear force is regarded, owing to
the lower gradient in the support regions. Furthermore, the relevant section for shear
design is in the case of a direct support in a distance of 0.5–1.0d from the supported
edge. It should be pointed out that some programs estimate the nodal values by an
interpolation of the inner forces in the adjacent elements. In this case, a shear of
v ¼ 0 kN/m would be estimated, whereas the beam theory will give a maximum value
of v ¼ 31.25 kN/m.
The dimensioning of spatial structures is based on the internal forces, which are
estimated using the linear elastic behaviour of the concrete, in a similar manner to
that in beams. In addition, the orientation and the area of reinforcing bars have to be
chosen with respect to practical constraints and forces. Hence, two main input
parameters – orientation of the reinforcing bars in the x- and y-directions – are mostly
fixed for dimensioning.
Two unknowns remain: the internal forces and the orientation of the cracks from
bending. The latter does not usually coincide with the orientation of the main com-
pression forces. Among other things, the crack pattern depends on the level of the
loading, the ratio of the main bending moments mI/mII and the ratio of the reinforcement
in both directions. The compatibility conditions and the complicated stresses can only be
estimated approximately if the reinforcement bars do not cross the crack in a right angle
(Figure 4.75).
Hence, the design of a slab must be based on various simplifications and assumptions,
such as, for example
254
l2
l1
Compatibility
Rebars Dowel effect
not given l1 <
= l2
where:
is the angle between mI and y-axis (orientation of rebar).
jm1 j > jm2 j; m1 > 0; m2 ¼ k m1
The resulting bending moments are distributed into the (orthogonal) directions of the
reinforcement according to the ratio asx/asy.
my asy dy
kt ¼ ¼
mx asx dx
where 0 4 kt 4 1; d is the lever arm.
The parameter kt is fixed by the second equation of equilibrium. The main bending
moment m2 (resistance) should be greater than the bending moment caused by the
actions. This condition gives the lower limit of kt:
k tan2
kt 5
1 k tan2
255
1 m1
2
m2 mx m1
m2 my 2
mx m xy
m xy
It should be noted that the codes specify a minimum transverse reinforcement (20% of
the maximum reinforcement in the main load-bearing direction). Thus, the value kt
should always be greater than 0.2. Furthermore, for 5 258, the reinforcement
should be distributed in the orthogonal directions, as illustrated in Figure 4.77. In this
figure, the bending moment m2 is always greater than that caused by the external actions.
The design bending moments in the direction of the reinforcement bars can be estimated
by using the following expressions.
1
my ¼ k m1 ¼ m1
cos þ kt sin2
2
mx ¼ kt my
Figure 4.77 Distribution of the main bending moment in the direction of the reinforcing bars
mx + my
or kt
m1
2.0
my
m1
1.0
mx
m1
mx
m1
my
mx
0.0
0 20 25 30 40 45°
256
The forces in the reinforcement results from the equilibrium conditions on a flat plate
element (see Figure 4.80). The angle of the cracks ’ is determined by the principle of
the minimum energy. A linear elastic material behaviour is assumed.
tan þ k cot cot þ k tan 1
cot4 ’1 þ cot3 ’1 cot ’1 ¼ 1 cot4 ’1
1k ð 1 kÞ
where:
asx a E
¼ ; ¼ sx s
asy h Ec
k ¼ m2/m1 ¼ n2/n1
m1 and m2 are the principal bending moments
is the angle between direction of rebar ( y) and main tensile force n1 ( 4 458)
Figure 4.78 Transformation of bending moments into normal forces in an equivalent disc
n1d = m2/zm
n2 = m2/zm
zm
A m2 = k . m1
B B
Compression zone
due to bending
due to bending
Tension zone
Section A–A
Plate
m1 m1
element
n1d = m2/zm
m2 = k . m1
n2 = m2/zm
Section B–B zm
257
ire c n
ra
t io n
i
ck
a rs
or
Reb
ie
nt
y -d
at
io
n
n2 = 2 . h
n1 = 1 . h
/4
n1 = 1 . h
Reb
ars
x-d
irec in
tion
n2 = 2 . h = k . n1
If the reinforcement in both directions reaches the yielding point, the angle of the crack
’2 is given by the following expression.
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
tan ’2 ¼ C þ C 2 þ
where:
1 þ tan2 ðk Þ k
C¼
2 tan ð1 kÞ
Further simplifications are justified under the assumption that both reinforcement
directions are stressed equally, resulting in the most economical solution. However,
this case will not always be possible. Based on this assumption, the angle ’ is constant:
’ ¼ ’1 ¼ /4.
258
Figure 4.80 Equilibrium between internal forces and actions for an orthogonal arrangement of
rebars
N2 . b 2
b2 = sin( – )
Equilibrium of forces
os
y N1 . b1
1
bx
as
b1 = cos( – )
x
N2 . b2
1 N1 . b 1
y
as
Fsy . by
x
cra
by Fsx . bx
=s
ck
in
Fsy . by
= angle between the rebar (y)
and the cracks
N2 . b2 = angle between rebar (x) and the
greatest principal tension force, N1
b2 = cos( – )
Fsx . bx
in
2
=s
y
bx
as
x
b1 = sin( – )
x
N1 . b 1
cra
Fsy . by N2 . b 2
ck
by
Fsy . by Fc
Fc Fsx . bx
c
1
259
The following expressions are used to determine the forces in the reinforcement Fsx and
Fsy and the compressive force Fc:
A more refined design model is given in appendix LL of EC2, Part 2 (Eurocode 2, 2005).
This approach is suitable for numerical analysis only.
yes if
no
mx –|mxy|
yes if
no
my |mxy|
2
= –mx + l mxy l
mudx
= –mx + mxy
mudx / l my l
= –my + l mxy l
mudy
=0
mudy
260
Figure 4.82 Reinforcement ratio asx/as1; asy/as1 and (asy þ asy)/as1 in relation to the angle
between direction of rebar and main tensile force
1.4
S – k = 0.0
B – k = 0.0
asx/as1
1.2
1.0
asy/as1
S – k = 0.6 B – k = 0.6
0.8 B – k = 0.6
B k = 0.6
0.6
S – k = 0.6
0.4 B – k = 0.0
0.2
S – k = 0.0
0
2.0
Total reinforcement
1.9
1.8
B – k = 0.6
1.7
(asx + asy)/as1
S – k = 0.6
1.6
1.5
1.4 B – k = 0.0
S – k = 0.0
1.3
1.2
k = m 2/m 1
1.1
1.0
0 5 10 15 20 25 30 35 40 45
(= angle between x-rebar and n1)
The EC2 values are not shown, since they match with the approach of Baumann. The
results of both models match for an angle between direction of rebar ( y) and main tensile
force nI of ¼ 08 and 458. Significant differences can be seen in the region in between
these boundary values (08 < < 458).
261
most tables have been calculated in accordance with the Kirchhoff theory, the FE soft-
ware is generally based on the more consistent model of Reissner/Midlin. The main
difference between the two approaches is in the shear deformation. The Kirchhoff
model neglects shear deformations, while these are taken into account in the Reissner/
Midlin model. This may result in significant differences with regard to shear forces.
This is illustrated in the following example, a simply supported rectangular slab with a
panel dimension of ly /lx ¼ 7.50/5.0 m.
Figure 4.83 and Table 4.13 show the result of this comparison (q ¼ 10 kN/m2). A very
good agreement in the bending moments can be seen. However, the twisting moment
mxy of the FE model is 50% less than that calculated by Czerny (1999).
Figure 4.83 Bending moments and shear forces of a simply supported rectangular slab under
uniform load calculated by FEM and with Czerny tables (Czerny, 1999)
+ x x
+
Fe = –30.7 Fe
Fe
Fe = –25
mxye = 10 75
18
v
mymax = 7.5
mxm = 18.7 21
Finite
Element vx
results
mym = 7.2
18
75
Fe = –25 Fe
262
Table 4.13 Comparison of the internal force, support forces and deflection
Czerny (1999) FE FE
¼ 0.2 ¼ 0.0
ly = 7.5 m
mx
my,max in kNm/m 7.2 10.9 (151%) 7.5 (104%)
my
vyrm in kN/m 18.2 17.7 (97%) 17.9 (98%)
v yrm in kN/m 26.5 25.0 (94%) 25.0 (94%) x
Large differences can be seen in the shear force distribution. According to Czerny, the
distribution of the shear force at the supported edges is almost parabolic. The shear
force at the corners of the slab is zero. On the other hand, the FE calculation gives
maximum shear forces in the corners of the slab. This peak value is more than three
times greater than the maximum shear force calculated with Czerny’s tables. This may
cause some design problems, because usually one wants to omit any shear reinforcement
in a slab.
The internal forces for the slab shown in Figure 4.84 are estimated by FEM and the
Pieper-Martens model. This structure has been used in the original publication (Pieper
and Martens, 1966), but the loads and the thickness of the slab have been increased to
represent up-to-date conditions.
The bending moment distributions calculated by FEM in both axes for various sections
are plotted in Figures 4.85 and 4.86. Figure 4.87 shows the comparison of the bending
moments in the relevant sections. A good agreement in the field bending moments can
be seen for both calculations using the simple manual method and FEM. The support
bending moments are generally greater due to the assumption of high flexural restraints
(between 50% and 100%).
It should be noted that the minimum bending moments given in EC2, Part 1, are relevant
in the small slabs – numbers 4 and 5 – and not the values from the linear elastic FE
model. The minimum internal forces always have to be considered in addition to an
FE analysis.
263
264
Reproduced with permission.
ly/ly ¼ 1.0 1.05 1.1 1.15 1.20 1.25 1.30 1.35 1.40 1.45 1.50
mxm ¼ q lx2 : 27.2 24.5 22.4 20.7 19.1 17.8 16.8 15.8 15.0 14.3 13.7
my,max ¼ q lx2 : 27.2 27.5 27.9 28.4 29.1 29.9 30.9 31.8 32.8 33.8 34.7
mxye ¼ q lx2 : 21.6 20.6 19.7 19.0 18.4 17.9 17.5 17.1 16.8 16.5 16.3
Fe ¼ q lx2 : 10.8 10.3 9.85 9.5 9.2 8.95 8.75 8.55 8.4 8.25 8.15
vxrm ¼ q lx : 2.96 2.87 2.78 2.71 2.64 2.58 2.52 2.47 2.43 2.39 2.36
q lx4
fm ¼ : 0.0487 0.0536 0.0584 0.0631 0.0678 0.0728 0.0767 0.0809 0.0850 0.0890 0.0927
E h3
ly =lx ¼ 1.50 1.55 1.60 1.65 1.70 1.75 1.80 1.85 1.90 1.95 2.00
mxm ¼ q lx2 : 13.7 13.2 12.7 12.3 11.9 11.5 11.3 11.0 10.8 10.6 10.4
my,max ¼ q lx2 : 34.7 35.4 36.1 36.7 37.3 37.9 38.5 38.9 39.4 39.8 40.3
mxye ¼ q lx2 : 16.3 16.1 15.9 15.7 15.6 15.5 15.4 15.3 15.3 15.2 15.1
Fe ¼ q lx2 : 8.15 8.05 7.95 7.85 7.80 7.75 7.70 7.65 7.65 7.6 7.55
vxrm ¼ q lx : 2.36 2.33 2.30 2.27 2.25 2.23 2.21 2.19 2.18 2.16 2.15
v xrm ¼ q lx : 1.95 1.94 1.93 1.92 1.92 1.92 1.92 1.92 1.92 1.92 1.92
vyrm ¼ q lx : 2.75 2.74 2.73 2.73 2.73 2.72 2.72 2.71 2.71 2.70 2.70
v yrm ¼ q lx : 1.89 1.88 1.87 1.86 1.85 1.84 1.83 1.82 1.82 1.82 1.82
q lx4
fm ¼ : 0.0927 0.0963 0.0997 0.1029 0.1060 0.1093 0.1118 0.1145 0.1169 0.1195 0.1215
E h3
Note: h is the slab thickness; v xrm, v yrm is the vertical support force in the middle of the supported edge; vxrm, vyrm is the shear force at the line support in the middle of the
supported edge.
q . lx q . lx
vyrm = – vyrm = –
1.89 2.75
y y q . lx2
q . lx2 q . lx2 Fe Fe = –
mxye = mxye = 8.15
+ 16.3 16.3
+ 0.2ly
q . lx2 0.2lx
q . lx2 mymax = q . lx
mxm = + 34.7
13.7 vxrm = –
1.95
+
265
Finite-element Design of Concrete Structures
Slab thickness: h = 18 cm
6.00 m
1 3
Loading:
g = 8.0 kN/m2
2 q = 4.2 kN/m2
Floor covering
Cement screed
40
h = 4 cm
40
Foil g = 0 kN/m2
4 5 6 7 4.80 m Impact sound
180 insulation h = 4 cm
Concrete slab,
h = 18 cm
15
Gypsum plaster, 15 mm
1.60 m 2.00 m 5.40 m 3.00 m
266
9.39
6.26
4.09
9.51
6.26
–2.04
Maximum
–7.38 bending
–4.44 –3.88 moments
–5.01
–6.83
–6.50
–7.32
2.76
0.84
9.61
–1.36
–1.58
9.90
2.83
0.65
4.03
Minimal
–7.36 –5.22 bending
–12.05
moments
–14.99
–14.09
–10.97
–5.85
267
13.40
1.41
11.72
–10.54 Maximum
bending
moments
5.80
3.12
3.12 9.05
5.31
–7.13
–8.38
7.20
8.21
–17.84
Minimal
–2.49 bending
moments
2.49
–12.08
–13.60
1.32
5.48
1.82
268
Figure 4.87 Bending moments in the relevant sections – comparison between FE analysis and
Pieper-Martens analytical method (values in brackets)
9.4 (7.7)
6.3 (4.4)
6.00 m
13.1 (12.8)
(–17.6)
–10.5
11.4 (12.1)
–14.1
–11.0(–18.20)
(–15.7)
–13.6
(–4.0)
–1.9
9.1 (8.2)
4.80 m
5.8 (7.0)
3.1 (4.0)
1.2 (4.1)
REFERENCES
Bathe K-J (1982) Finite Element Procedures in Engineering Analysis. Prentice Hall,
Englewood Cliffs.
Baumann Th (1972) Zur Frage der Netzbewehrung von Flächentragwerken. Der
Bauingenieur 47(10): 367–377.
Bittner E (1965) Platten und Behälter. Springer Verlag, Wien.
Burton WS and Sinclair GB (1956) On the Singularities of the Reissner’s Theory for
Bending of Elastic Plates. Journal of Applied Mechanics, 53(March): 220–222.
CEN (European Committee for Standardization) (2004) Eurocode 2, Part 1: Design of
concrete structures – General rules and rules for buildings. December.
CEN (2005) Eurocode 2, Part 2: Design of concrete structures – Concrete Bridges – Design
and detailing rules. December.
Czerny F (1999) Tafeln für Rechteckplatten. In Betonkalender 1999 (Eibl J (ed.)), Volume 1.
Ernst & Sohn, Berlin.
DIN 1045-1 (2008) Concrete, reinforced and prestressed concrete structures – Part 1: Design.
Deutsches Institut für Normung. Berlin, August.
269
270
Chapter 5
Shell structures
Shells are spatially curved surface structures that exhibit the behaviour of both plates
(Chapter 4) and shear walls (Chapter 3). They can be loaded perpendicular (as a
plate) and normally (as a shear wall) at their middle surfaces. Thus, they can develop
bending moments, membrane forces and shear forces. Each element node has six degrees
of freedom: three for deflection and three for rotation (see Figure 5.1).
FEM allows the calculation of the internal forces for shells of arbitrary shapes. This
chapter only deals with thin shells, where the shell thickness is much smaller than the
main radius of curvature. With this assumption, the FE model can be reduced to the
midplane of a shell. Brick or volume elements can be used for thick shells or massive
structures.
The accuracy of an FE shell analysis is dependent on the numerical model, the shape
functions of the elements used and the order of the numerical integration. The following
calculations use a four-node flat shell element (quadratic displacement approach), where
the bending and membrane-load-bearing behaviour is not coupled (Hughes, 1987).
271
u3 u4 jx,3 jx,4
w3 w4 jz,3 jz,4
Elevation
+97.5 m +110 m
+32 m
0m
Plan view
m
24
C3
Detail A T23
D2 D3
272
Sea. There was no significant difference between this platform and the earlier-
constructed structures of the Condeep type.
The structure collapsed during installation. The construction of this type of platform is
carried out in three phases. In the first phase, the lower part of the foundation structure
is built in a dry dock. Then, the dry dock is flooded and the structure is shipped to a
deep-water construction berth, where the rest of the cylindrical caisson cells and the
four shafts are erected. Finally, the whole platform is lowered to nearly the seawater
level by partially flooding the cylindrical caisson cells, and the steel deck is lifted on
the shafts and fixed in place.
Detailed experimental and numerical investigations were carried out after the accident. It
was found that the collapse was initiated by the failure of a wall in the so-called tricell
T23 (see Figures 5.2, 5.3 and 5.4). This failure allowed flooding of the buoyancy caissons,
and the whole structure sank to the seabed. The financial loss from this accident was
estimated at US$250 million.
The failure was caused by the large differential pressure in the tricells. As noted
earlier, the caisson cells are partly filled with seawater during the sinking operation.
The area between these cells is open to the sea, and the high differential water pressure
between the caissons and the tricells results in very high pressures on the walls
(Figure 5.3). Investigations that were carried out after the collapse showed that the
failure of the walls must have occurred under a water pressure of 670 kPa, equal to
67 m water depth.
5.80 m
0.55 m
Water pressure pi
0.80 m
273
Deflection of
wall
Kinked
reinforcing
bars
Crack
Spalled concrete
The subsequent, very extensive numerical and experimental investigations identified two
main causes of the collapse
g erroneous FE analysis
g insufficient reinforcement and bad detailing of the tricell walls.
The bad detailing of the intersection between the cell walls can be seen in Figure 5.5. The
T-headed bar of diameter ds ¼ 25 mm was not anchored in the compression region of the
walls. More information about this problem is given in the available literature (e.g.
Holand, 1997; Jakobsen and Rosendahl, 1994).
The errors in the FE analysis of the platform lay in insufficient discretisation, and in the
poor geometrical shaping of some elements in the tricells (Figure 5.6). Simplifications
had been made with regard to the number and shape of the elements, due to the size
and complexity of the structure, and the limited computing capacity. This resulted in
the incorrect modelling of the load-bearing behaviour of the tricells. The walls of the
cells were modelled using only two element layers.
Using a coarse FE mesh, the internal tensile forces (i.e. the shear forces at the wall
supports) are underestimated by nearly 50%. Due to the large amount of input and
output data of the 3D model used, this mistake was not recognised by the design
engineers.
The basic model problem is demonstrated by the following parametric study. For
simplicity, a membrane model is used.
274
Figure 5.5 Arrangement of the reinforcement in the intersection of the caisson cells (tricells)
T-headed bar
ds = 25 mm/17 cm
70/70/16
Shown detail
d = 25 mm
975
Figure 5.7 shows the distribution of the main membrane forces in the region of the wall
intersections. The inclination of the compression struts result in a high tensile force,
which should be carried by the T-headed bars. However, the size of this force is highly
dependent on the size of the elements, as can easily be demonstrated by calculations
with different FE meshes. Figure 5.8 shows the distribution of the horizontal tensile
force in the walls for different mesh refinements (see Figure 5.9).
With a very coarse mesh, there appears to be no tensile force in the intersection region
(mesh number 1). Refinements result in a large increase in the horizontal force resultants
275
1m
(see Table 5.1). It should be noted that the results are highly dependent on the element
type and size used.
The resultant forces in the walls of the tricells loaded with water pressure of
w ¼ 0.67 MN/m2 can be easily determined with the truss model shown in Figure 5.10.
0.80 m
detail N1
+
0.55 m
- –3.9 D1
nx –3.9
[MN/m] –4.0
N4 N8 N2 T1
+
N1 : 9.7
N4 : 28.1
N8 : 27.2
N2 : 24.5
276
N1 N2
N4 N8
The normal forces N ¼ 1.69 MN/m are in good agreement with the FE results of mesh
N4 to N8 (T1 D1 ¼ 1.66 MN/m, see Table 5.1). The resultant tensile force from the
bending moment is difficult to estimate, as the internal lever arm is not known. For
the following manual analysis, the inner lever arm is taken from the FE calculations
as z ¼ 0.50 m.
m q l 2 0:67 4:382
T1M ¼ ¼ ¼ 2:14 MN=m
z 12 z 12 0:5
The FE analysis gives a value of (see Table 5.1):
T1M ¼ 3:45 1:66 ¼ 1:79 MN=m
Mesh number N1 N2 N4 N8
277
l = 4.38 m l
q = 0.67 MN/m2 q = 0.67 MN/m2
q q
60° 60° ql 2
-–
ql/2 24
ql 2
Bending moment 12
8m
8m
4.3
4.3 N
60°
l=
30° 30°
l=
ql /2 ql/2
ql 1 0.67 · 4.38
N= · = = 1.69 MN/m
2 cos 30° 2 · cos 30°
q·l
T= · tan 30° = 1.40 · tan 30° = 0.85 MN/m
2
What are the conclusions we can draw from analysing this failure (Bergan, 1998)?
Complicated shell models with various arbitrary loadings can hardly be checked by
manual analysis of simplified equivalent structures, whereas this is possible for flat
plates and shear walls. Furthermore, the checking of the equilibrium condition of the
external loads and the support forces is not sufficient to guarantee the correctness of
the analysis. Other checks have to be performed for shells with complicated shapes.
The essential problem in the numerical analysis of the platform was that the element
mesh used was not able to model the real deformation characteristics and load-bearing
behaviour of the structure. The element mesh used was too coarse; furthermore, the
assumption of linear elastic behaviour did not hold for all sections of the structure.
Thus, each structure has to be checked for regions where the assumptions of the numer-
ical model do not apply. Detailed knowledge of both FEM and the material behaviour
are required for this task. Numerical algorithms such as automatic mesh refinement
cannot compensate for the knowledge of the user. When the element mesh used is too
coarse, they are often not able to detect critical sections.
Critical regions of the structure can be analysed separately from the whole structure by
means of, for example, strut-and-tie models or refined FE models (substructure method).
The resultant forces at the outer surface of the substructure can then be applied to a
global model, and vice versa.
278
25°
60
25 m
+3.00 m 60°
0.00 m
–1.50 m
The cylindrical bin, which is used for storage of fly ash, has a height of 59 m and an inner
diameter of 24.4 m (see Figure 5.11). The following analysis is carried out on a simplified
model. Only the cylindrical shaft is modelled. The inverted cone and the partial restraint
of the walls in the pile foundation are neglected.
For axisymmetric structures, 3D models are only required if unsymmetric loads have to
be applied. This is the case for most silo structures. The main loading of a silo results
from the pressure due to the bulk material inside the bin, which can be estimated from
various codes (e.g. EC1, Part 4; see Eurocode 1, 1997). A uniform pressure distribution
around the perimeter of the bin is generally assumed. In addition, a horizontal partial
overpressure, a so-called ‘patch load’, has to be applied to the structure (Figure 5.12).
The size of the square loaded area, s, is equal to s ¼ 0.2dc ¼ 0.4r (where r ¼ internal
diameter) for cylindrical concrete shells. For the structure shown in Figure 5.11, this
results in s ¼ 4.88 m. Generally, it is sufficient to apply this load at mid-height of the cell.
The patch load causes bending moments in the cylindrical walls. Charts to calculate the
resulting internal forces for thin shells have been published by Hennig (1971). These
tables can only be applied to infinite long cylindrical shells. The restraint on the walls
by the foundation or the roof is neglected. More accurate values of the internal forces
can be determined by a 3D FE analysis. Here, a sufficient fine element mesh in the
region of the patch load has to be used, as shown in the following calculations.
279
Patch load
pp pp
s
h s
2r
The FE analysis is carried out with three different meshes (Figure 5.13). For simplicity,
the sizes of all elements are kept constant within each model, whereas the element size is
only relevant in the region of the patch load. The element size is chosen so that the loaded
area covers one element in mesh number 1, 2 2 elements in mesh number 2 and 4 4
Loaded
area
X Y
280
C B A
C B A
nx ny
200 kN/m 100 kN/m
Y
Y
X Z
X Z
C B A
B A
mx
C my
100 kNm/m
20 kNm/m
Y
X Z
X Z
elements in mesh number 3. A uniform horizontal patch load of q ¼ 100 kN/m2 is applied
to the walls.
The internal forces under the patch load increase considerably with an increase in the
number of the elements, as can be seen in Figure 5.14. Table 5.2 gives the maximum
values of the internal forces. The maximum bending moment, calculated with the
coarse element mesh A, is only one-third of that of mesh C. Calculating the reinforcement
Table 5.2 Maximum internal forces (nodal value)
ny in kN/m Min 487 (85%) 556 (97%) 571 (100%) 540 (95%)
Max 273 (71%) 356 (92%) 387 (100%) – –
281
Figure 5.15 Contour plot of the bending moment mx in circumferential direction (mesh C)
35 m
100
m
25
28 m
200
pp
28 m
150 25 m
–50
100
75 nx
ny ny
mx
my
50
nx
00
25
–25 -
15 m
Z
Bending moment MXX from
–12.5 0.0 –63.6 to 209.3 kNm/m 12.5
X Y
Figure 5.16 Contour plot of the membrane force nx in circumferential direction (mesh C)
35 m
0.0
m
25
28 m
500
0.0
0.0 pp
400
28 m
200 25 m
0.0
nx
ny ny
mx
my
nx
15 m
Z
Membrane force nx from
–12.5 X Y 0.0 –78.7 to 526.9 kN/m 12.5
282
requirements using such considerably lower value for the bending moments would result
in a lack of safety. A contour plot of the bending moment and the normal forces in
circumferential direction calculated with mesh C are shown in Figures 5.15 and 5.16
respectively.
Due to the large uncertainty in the actual loading conditions, a high accuracy in the
analysis is not required. The patch load represents only a rough idealisation of the
non-uniform pressure distribution in a silo bin in the circumferential direction caused
by the granular bulk material. In this respect, an extensive 3D FE shell analysis is
only useful in special cases. Table 5.2 shows that a simple manual analysis of the
maximum bending moments, using the parameters published by Hennig (1971), gives
a good agreement with the FE analysis.
As previously mentioned, the patch load represents a rough simplification of the real
pressure distribution. Therefore, material nonlinear calculations are not permissible,
as the theoretical bending moments in the walls would be considerably reduced.
5.2. T-beams
T-beams are widely used in concrete structures, whether as a main longitudinal girder in
bridges or as the support ( joist) of slabs (see Sections 2.8.2 and 2.8.3). An analysis can
easily be carried out manually on equivalent girder systems. However, the use of 3D FE
calculations is reasonable and would be required if the load-bearing behaviour of the
structure has to be modelled to greater accuracy. In these calculations, one does not
depend on simplifications such as the effective width of the flange beff. In contrast to
grillages, a 2D spatial load distribution is modelled. However, the increase in accuracy
requires a much greater workload.
Deflected shape
This area is
included twice
283
In this model, the behaviour of the T-beam is given with a high degree of accuracy.
However, this is only the case when the assumption of a linear elastic material behaviour
holds true and the flange and the webs are not too thick. Massive structures must be
analysed with volume elements.
The elements of the web are mostly arranged up to the midplane of the plate. This
approach results in a greater cross-section area due to the intersection of the flange
(plate) and the web (beam) elements (see Figure 5.17). In general, this inaccuracy can
be neglected, due to the relatively small influence of the bending stiffness on the internal
forces. As an alternative, the nodes in the web and the flange elements may be coupled.
The main disadvantage of a shell model is the considerable effort needed in the genera-
tion of the 3D mesh. Furthermore, an automatic dimensioning of the girder by the
computer program is not possible (see Section 3.3).
284
Strain (elastic)
Slab
Web
A Rigid support
Support
B Shell model
Schalenelemente
Plate or beam
C
elements in midplane
Coupling of nodes
E Eccentric shell
elements
Coupling of nodes
285
does not need an additional coupling node at the midplane of the slab. Thus, the element
mesh generation for Model D1 is easier and faster than D2.
The local connection of the beam and plate elements results in a non-uniform distribu-
tion of the bending moments and shear forces in the beam (web), which may be smoothed
for design purposes.
Both models require additional work in the dimensioning of the T-beam. The internal
forces in the flange must be integrated for the calculation of the bending moment My
and shear force V in the beams (member forces). The T-beam must then be dimensioned
for both group forces and not for the internal forces of the beam element only.
Centre of
160
gravity 15 m
Bending moment
421.9 kNm
15
Section properties
A = 0.510 m2
Iy = 0.1144 m4 Shear force
zs = 0.429 m
(from top of flange) 112.5 kN
286
The results of the shell model are shown in Figures 5.20–5.22. The structure and the
loading are symmetric. Therefore, it is sufficient to model only half of the whole structure
150 cm Midspan
System and
20 cm FE-model
160 cm
15 cm
15 kN/m
Y X
Z
Support 15 m
Axis of symmetry
–42 –87 –117 –142 –237 (–1580)
287
Figure 5.21 Membrane forces nx, ny, nxy in the flange in different sections
Membrane force ny
–77.8
41.8
1.96 14.1 28.8
0.00
–14.1 –41.8
0.0
–28.8 – –
Distribution in
longitudinal
direction
and to consider the special boundary conditions at midspan. The web and the flange of
the T-beam are each modelled by 20 50 elements. A pin support is used, since the
internal forces near the supports are of no further interest. A uniform load of
q ¼ 10 kN/m2 is applied to the flange.
288
1.5 m
Flange
(top view)
Membrane force nx
–117 kN/m
1.5 m
Web
Support
The distribution of the horizontal membrane forces and the horizontal strains across the
section depth are linear except near the supports (Figure 5.20). The main assumption of
any beam analysis, the linear strain distribution over the depth of the cross-section, is
valid for almost the whole structure. Therefore, it is not surprising that the normal
stresses of the FE model correspond well with the results of the beam analysis (stresses
at the upper and lower edges of the section at midspan (top ¼ 1580 kN/mm2;
bottom ¼ þ 4333 kN/mm2; see Figure 5.20).
The compressive stresses x ¼ nx/hf in the flange increase parabolically from the support
axis to midspan in agreement to the bending moment distribution (Figure 5.21). The
support force causes high transverse compressive stresses y ¼ ny/hf in the flange, at
the end of the beam (Figure 5.21 middle). The membrane shear forces nxy in the flange
are greatest at the intersection with the web. They decrease in a longitudinal direction
to zero at midspan. Consequently, the shear force is not constant as assumed in most
models for shear design of a flange in the transverse direction (e.g. see Eurocode 2,
2004, Section 6.2.4).
A 3D folded slab model of a T-beam is, in general, too extensive for most practical cases,
especially as the dimensioning task is difficult (see, e.g. shear walls, Section 3.3).
However, shell models are very helpful in determining the flow of forces in a structure
and to evaluate an accurate strut-and-tie model (see Section 5.2.3).
289
Figure 5.23 Model C: Plate elements (flange) with central equivalent beam elements (web)
Axis of symmetry
Cross-section
h equiv = 2.09 m
Y X
Axis of gravity of
Support equivalent beam Y X
Z
Z
Cross-section of equivalent beam
Figure 5.23 shows a model with beam elements located in the midplane of the plate. An
equivalent depth of hequiv ¼ 2.09 m and a width of bw ¼ 0.15 m (same as a real web) can be
calculated for the given dimensions of the T-beam. The models with an eccentric beam
element are shown in Figures 5.24 and 5.25.
The total internal forces of the T-beam are calculated by summation of the beam forces
Nbeam, Mbeam and the membrane forces in the shell nx, mx. The latter values are gained by
numerical integration.
Edge stresses of a beam: c ¼ N=A M=W
nx bF mxx bF nx mxx 6
Edge stresses of a slab: c ¼ ¼
A W hf h2F
Total normal force of T-beam: Ntot ¼ Nbeam þ Nslab ¼ Nbeam þ nslab bflange
Total bending moment of T-beam:
Mtot ¼ Mbeam þ Nbeam zbeam þ mslab bflange þ nslab bflange zslab
where:
b, bF are the widths of web and flange
h, hF are the depths of web and flange
290
Figure 5.24 Model D1: Shell elements (flange) with eccentric beam elements (web)
Axis of symmetry
Y X Axis of gravity of
equivalent beam Rigid connection
Z slab – equivalent beam
Support
Y X
Cross-section of equivalent beam Z
z is the distance between the centre of gravity of the flange or the web to that of
the T-beam.
The results of the various approaches are summarised in Table 5.3. A very good agree-
ment can be observed between all three different numerical models and the values of the
manual analysis. The small differences between the results are caused by rounding errors.
The distributions of the edge stresses are shown in Figure 5.26.
The bending behaviour of the T-beam is well represented in all the models. No differ-
ences in the deflection of the girder can be observed.
As was expected, the stress distributions in the slab and in the equivalent girder of Model
C are completely different from the other cases (see Figure 5.26). These values cannot be
used in design (e.g. checking the edge tensile stresses in the serviceability limit state for
partial prestressed structures). In contrast, the stresses of Model D agree well with
that of the T-beam girder, both in the web and the flange.
In Model D, single forces and single bending moments are introduced in the nodes due to
the local coupling of the beam and shell elements. This results in an unsteady sawtooth
distribution of the internal forces (Figure 5.27). For the purpose of the dimensioning of
the T-beam, the distribution of the internal forces may be smoothed.
291
Figure 5.25 Model D2: Shell elements (flange) with eccentric beam elements (web)
Axis of symmetry
Additional node
Y X
Axis of gravity of Rigid conncection
Z equivalent beam slab – equivalent beam
Support Y X
Z
Cross-section of equivalent beam
In case of automatic computational design for shear, there are significant differences
between the various models. These are caused by the difference in the inner lever arm
z. Therefore, the dimensioning should be done for the resultant internal shear forces
acting on the T-beam.
Table 5.3 Member forces at midspan over the section depth of different models (dimensions kN
and m)
Model Nbeam Mbeam bbeam hbeam zbeam top: bottom: Nbeam Mbeam
nslab mslab mslab mslab mslab kN/m2 kN/m2 Nslab Mslab
D2 245.0 2.43 1.5 0.2 0.429 0.1 1587 859 366.0 3.3 þ 108.8
Total 0.8 423.0
292
Figure 5.26 Stress distribution in midspan over the section depth of different models [kN/m2]
Model C
–3832
Model D1
Beam model Model D1 Slab Slab
–1582 –1584 Slab –1578 –1587 –363
–858
–845 –849 –840–859 +363
+3832
Support
Shear force in kN Y X
111
293
Axis of symmetry
Cross-section
210
10 12
46
19 strands
each
42
Isolated
Y X unbonded 153
Bonded
Z
39 + 13 = 52
strands
4.2
7.0
84 42 84
stresses and the required reinforcing steel in the support region. Therefore, accurate
information on the flow of forces is useful. A 3D FE shell analysis can be helpful in
this task.
The strut-and-tie modelling based on an elastic FE analysis will be illustrated for a precast
T-beam of the single-span concrete bridge shown in Figure 5.28 (see also Figure 2.98). This
structural system has already been explained in Section 2.8.3. The beam has a span length
of l ¼ 38 m. The following calculation only considers the prestressing forces. The T-beam
girder was stressed by 52 straight tendons of grade ST 1570/1770.
The analysis can be restricted to half of the system, as the structure and the uniform
loading are symmetrical to midspan. 20 50 shell elements are used to model the web
and the flange.
The results of the shell analysis are shown in Figures 5.29 and 5.30. The strut-and-tie
model was adapted to the orientation of the main membrane forces. The large transverse
tensile stresses in the flange at the support region are clearly shown in Figure 5.30.
The preceding analysis was based on linear elastic material behaviour. This applies
mostly to the web, as almost no tensile stresses occur under prestressing; nevertheless,
force redistributions may happen in the flange due to cracking.
294
Figure 5.29 Main membrane forces at the end of the girder (l ¼ 3.0 m) and membrane forces in
longitudinal direction in various sections – strut-and-tie model
–10968
482
1860
446
–11046
484
2022
361
–12618
534
519
1758
142
75.3
77
F1/2
F1/2
3.0 m
F1
F1
Flange
Web
295
Figure 5.30 Distribution of the membrane forces in transverse direction in the flange near the end
of the beam (top view)
+642
ny
3.0 m
With a slab supported on beams, the available ‘flange’ width of the downstand beam is
often considerably bigger than the real one, beff. This raises the questions: which part of
the member forces in the slab should be used, and which beff should be applied for design
of the downstand beam? Complex, time-consuming analyses does not make much sense
for this kind of common structure. An automatic design by a computer program is
required.
In the following, the member forces and the required amount of reinforcement are
estimated for a regular slab supported by beams with different model variations. The
folded plate model is regarded as the most realistic solution, and is used to verify the
other models. The analysis is done with a web height of hw ¼ 3hF and hw ¼ hF
(where hF is the thickness of the slab), as the relation hw/hf has a large influence on the
member forces. The span of the slab in both directions is chosen so that a design of
both structures is possible with a reasonable amount of reinforcing steel.
The slab is clamped on one side to simulate a multi-span structure. Furthermore, a stati-
cally indeterminate system is chosen, because here the member forces depend on the
distribution of the stiffness in the structure (width of the compression flange).
296
First, the member forces and the required reinforcement in the slab and the T-beam are
estimated manually. These values are compared to the results of the numerical calculation.
The bending moments and shear forces for the rectangular slab are as follows (Czerny,
1999; see Figure 5.32):
mxermin,d ¼ 50.2 kNm/m (46.7) mxm,d ¼ 23.2 kNm/m (24.3)
myerm,d ¼ 35.2 kNm/m (32.4) mymax,d ¼ 7.8 kNm/m (11.4)
vxerm,d ¼ 51.1 kNm/m (50.2) vyerm,d ¼ 45.3 kNm/m (41.6)
vyrm,d ¼ 23.9 kNm/m (23.0) vyrm,d ¼ 35.7 kNm/m (31.2)
f ¼ 3.4 mm (3.4 mm) (concrete grade C 20/25, Ec ¼ 24 914 MPa)
Modelled region
A
10 m
Y
X
B
20
60
3.1 3.1 3.1
20
297
Figure 5.32 Member forces and required reinforcement for the slab (FE analysis)
41.6 kN/m
myerm = –32.4 kNm/m as,yerm = 4.5 cm2/m
The values in brackets are gained from an FE analysis of the slab with Poisson’s ratio of
¼ 0.2. The other figures are from Czerny (1999), where ¼ 0 is assumed.
The manual design is done with a constant internal lever arm of dm ¼ 0.17 m in both
directions. This analysis results in the following amount of reinforcing steel (Figure 5.32):
x direction: as,xermin ¼ 6.6 cm2/m (6.7) as,xm ¼ 2.8 cm2/m (2.8)
y direction: as,yerm ¼ 4.0 cm2/m (4.5) as,ymax ¼ 0.9 cm2/m (1.5)
This concludes the manual design for the bending of the slab.
The vertical loads on the downstand beam can be estimated very easily, for example, by
means of influence surfaces according to Grasser et al. (1991). The effective widths beff
for the design of the downstand beams are given in EC2, Part 1 x5.3.2.1 (Eurocode 2,
2004):
Distance between the point of zero moment of the clamped edge:
lo ¼ 0.25l ¼ 2.50 m
Prop: beff;axis B ¼ 0:2 bav: þ 0:1 l0 ¼ 0:2 3:1 þ 0:1 2:50 ¼ 0:87 m
0:2 l0 ¼ 0:50 m
4 ) beff;axis B ¼ 0:50 m
bav: ¼ 3:10 m
Field: beff;field ¼ 0:2 bav: þ 0:1 l0 ¼ 0:2 3:1 þ 0:1 7:50 ¼ 1:37 m
0:2 l0 ¼ 1:50 m
4 ) beff;field ¼ 1:37 m
bav: ¼ 3:10 m
Figure 5.33 shows the loads, member forces and required reinforcement for the T-beam
with and without consideration of the effective width beff,axis B in the region of the
298
3.1 m 1.8 m
10 m
50 20 50 1.37 m 20 1.37 m
Support Field
cross-section cross-section
–1140.
(–1056.) Bending moments in kNm
689.
Values in brackets = analysis (721.)
with support and field section
1.0d –302.
(–311.)
52.
(46.) Top reinforcement in cm2
18. 20.1
(17.) (21.1)
1.0d
12. Stirrups in cm2/m
(12.)
1.0d
clamped support. The value of beff has an insignificant effect on the member forces. The
bending moment in axis B is 8% larger if a constant effective width of beff,field is used.
The influence of the different stiffness at the support and in the field can be neglected,
in particular because of the inaccuracy of beff.
It should be noted that the whole analysis is based on linear elastic material behaviour.
An effective width in the support region, where the slab is under tension, does not make
sense. The preceding equations are only valid for flanges under compression.
The previously shown manual analysis of the structure is an easy and quick task. There-
fore, complex and time-consuming FE calculations do not seem to be necessary.
Nevertheless, a disadvantage of the manual design lies in the fact that the real behaviour
299
of the system is rather simplified. Usually, the downstand beam is not a rigid support for
the slab, as has been assumed. Its deformation causes force redistributions and an
increase of the member forces in the slab. The real behaviour of the system can be studied
by means of a folded plate model with a higher accuracy. In the following, a pure folded
slab system (FW) and a shell with a downstand beam (FB) is analysed. The first system
serves as a reference.
Figure 5.35 shows the member forces of both models (FW and FB). One recognises a
very good agreement of the results in the slab. The structural behaviour can be modelled,
therefore, also with the shell/beam system (FB) very well. Different shear forces are
obtained only in the region close to the clamped edge. However, these are not relevant
for the design of the slab, as the relevant section is located in the distance
1.0d ¼ 0.17 m or 0.77 m from the edge of the support. The computer program makes a
design for punching due to the high concentrated forces, which does not make much
sense.
If one compares the member forces of FW or FB models with the manual calculation,
large differences in the bending moments of the slab my in the direction of the downstand
beam can be observed. The deformation of the beam leads to doubling of the bending
moment at the clamped support from myerm ¼ 32.4 kNm/m (see Figure 5.32) to
mII,erm ¼ 68 kNm/m and to a three-times-greater bending moment in the field mymax
(my,max ¼ 11.4 and mI,field ¼ 29 kNm/m).
For both models, the resultant bending moments were calculated in different sections
by numerical integration of the member forces in the elements (Figure 5.36). It turns
out that the distribution of the bending moments corresponds to a single-span
beam, clamped on one side, with a uniform load of 6.2 m 16 kN/m2 ¼ 99.2 kN/m
300
–5
2 0 –7 –7 31 40 38 40 31
10
4
5
6
15 10
15 0
8
25
20
5 10
04
20
25
15
10
30 30 1
20
1
12
13 10 13 24 24 41
25
12
-37
25
20
15
20
15 0
10
10
20
5 10
8 8
20
–5
5
20
15 0
6 10 10
4 4
0
5 –5
–14 0
10
2 2 0
–10 –5 –15
–5
2 0 30 42 30
2 42
4
5
10
15
6
15
20
20
15 0
03
25 20
20
5 10
20
–35
–10 –5
25
15 0
28 28
40
1 1
20
11.9 11.9 24 24 1199 109
40
10
10 -37
15
20
10
15 0
10
8
10
20
5 10
8 15
6 15
–5
5
–35
6
15 0
15
4 10
–15
10
5
2 –5
2 –16 0
–10
–10 –5 11
40
40
301
Finite-element Design of Concrete Structures
Figure 5.36 Bending moment for the pure shell and shell/beam models
1000
Folded shell
model
Resultant bending moment: kNm
500
Truss model
0
0 2 4 6 8 10 m
–500
–1000
–1500
(MField ¼ 698 kNm, MSupp ¼ 1240 kNm). The same also applies for the shear forces. This
is not surprising, because, regardless of the structure, equilibrium condition must be
fulfilled. An integration of the shell forces over the whole width of 6.2 m results in the
same forces, as if the whole load is carried by the T-beam only. This corresponds to the
manual analysis where the downstand beam is regarded as a stiff support. A safe design
of the T-beam is achieved if the whole width of the ‘flange’ is applied. More realistic results
for the member forces of the T-beam can be obtained if the integration of the shell forces is
limited to a certain width. This will be explained in Section 5.3.1.3.
To illustrate the differences between a stiff and a soft support by means of a downstand
beam, the required reinforcement of the rectangular slab (see manual analysis) and the
folded shell system is plotted in the relevant sections (see Figure 5.37).
Only minor differences between both models are experienced with regard to the amount
of reinforcement in the x-direction at the top of the slab over the downstand beam
(Figure 5.37a). Here, the deflection curve shows a horizontal tangent due to the uniform
load. Big differences are obtained at the clamped support (Figure 5.37a,b). With a stiff
support, a reinforcement of approximately asy ¼ 5 cm2/m is required, whereas the shell
model gives a four-times-higher value of asy ¼ 19 cm2/m. The manual design deviates
in this section considerably from the ‘exact’ FE analysis.
The big reinforcement at the clamped support asy with the pure folded plate model is caused
by the large support bending moments in the slab. In addition, big tensile membrane forces
of almost ny ¼ 980 kN/m are calculated in the shell in axis B. Due to the assumption of a
linearly elastic material behaviour, the computer program cannot realise that it should
design a T-beam. It calculates the member forces and the reinforcement for the flange
and the downstand beam separately from each other. This example shows again that an
automatic design by software is not possible in some areas of a structure.
302
Figure 5.37 Reinforcement in the slab: left: stiff support; right shell/beam system in cm2/m
6.7 1.35
10 m
a b
Top y Top
layer layer
6.8
x
19.0
B
1.0 5.0 2.1 4.5
10.5
19.0
3.1 m 3.1 m
1.5
3.1
3.3
0.0 2.5
c d
Bottom y Bottom
layer layer
x
1.9 5.3
0.0 1.2
Furthermore, the deformation of the T-beam results in load redistributions from the
transverse to the longitudinal direction. Therefore, the transverse reinforcement
decreases in the folded plate model with regard to the three-sided clamped slab by
approximately 25% (Figure 5.37c) and the longitudinal reinforcement in the y-direction
increases by about approximately 100% (Figure 5.37d).
303
The computer program estimates, at the clamped support on the lower side of the slab, a
reinforcement of at most asy ¼ 5.3 cm2/m (Figure 5.37d). This results from the big tensile
force in the flange. This reinforcement should be arranged on the top of the slab in
addition to the calculated value there.
Therefore, it can be concluded that the flexible support of the slab due to the downstand
beam results in a significant redistribution of the forces. A computer program is not able
to design such a structure.
The effective width beff is a simple approach to determining the resultant force Fcd in the
concrete compression zone, the location of the centre axis and the stiffness of the jointed
cross-section. The compressive stresses are not uniform in the flange of a T-beam. They
decrease with increasing distance from the web (see Figures 5.38 and 5.39).
Nevertheless, the distributions of the compressive stresses or the membrane forces in the
flange are not known in general. Therefore, for easy manual analysis of T-beams, the
effective width has been introduced. The distribution of the compressive membrane
force n in the flange was analysed by FE models for various structural systems, cross-
sections and loadings and in different sections.
Ð The effective width results from the
resultant compressive force in the flange Fc ¼ n ds divided by the maximum membrane
force max n.
ð
ð n ds
Fc ¼ n ds ¼ beff max n beff ¼
max n
The effective widths from the FE analysis are shown in Figure 5.39, together with the
values of EC2 (Eurocode 2, 2004). Good agreement is achieved in the field, whereas
ted
ppor
ly su
Simp n ped
Clam
10 m 6.2 m
ted
ppor
ly su
Simp
ped
Clam
100 kN/m
304
Figure 5.39 Effective width beff according to EC2 Part 1 (Eurocode 2, 2004) and FE analysis
y
3.2 m
7.50 m
beff acc. to
EC2 part 1
beff acc. to
FE-analysis
1.9 m
52 52 kN/m
3.1 3.1
big differences can be observed in the support region. In the region of zero moments,
negative and positive membrane forces occur in the flange. The estimation of the stresses
in this section based on beff is incorrect.
The regions in which the existing beff is smaller than the code value are relatively small
and have, hence, only a low influence on the member forces of the structure. Calculations
with a constant effective width over the whole span will generally give good results.
The effective width depends on the span length, the support conditions, the transverse
reinforcement and also particularly on the type of loading. The equations from the
codes are valid only for uniformly distributed loads on the flange. Concentrated forces
result in significantly smaller values of beff. Furthermore, an elastic material behaviour
is assumed. This only applies for uncracked concrete regions in the serviceability state.
In the ultimate limit state, the concrete in the compression zone will show a nonlinear
behaviour in the plastic regime by which the width beff will increase. In a tension zone,
the effective width is useful only as long as the concrete has not cracked yet.
If the compression stresses are integrated over the effective width only to estimate
the member forces M and V of the girder, a part of the actual load is neglected. This
load portion can be assigned to the slab in the case of a suitable support. The required
reinforcement necessary for this is determined by the FE program.
305
Figure 5.40 Modelling of a slab on girder structure by shell and truss elements
A
T-beam
B
Shell
The integration of the membrane forces and bending moments in the slab and in the
girder is very time consuming if it is not automatically done by the FE program. To
avoid this effort, the T-beam can be viewed as a separate truss in the numerical
model (Figure 5.40). The centre axis of the beam is located in the midplane of the
slab to avoid additional normal forces. This T-beam can be designed by the program
automatically. If necessary, one can delete the stiffness of the shell elements in the
longitudinal direction in the width beff. For the given system, the influence of this
measure is not significant.
In Figure 5.41, the calculated member forces and the deflections of the T-beam are
shown. The deformations are in agreement with the pure folded shell analysis. The
load-deformation behaviour is therefore properly modelled.
In Table 5.4, the results of the FE analysis and the manual analysis are listed (see
Figures 5.41 and 5.33). The differences between both models are relatively small. The
‘accurate’ FE model gives approximately 15% lower member forces. Hence, the greater
effort of a numerical calculation does not seem to be justified for this structure.
306
Shear force: kN
277
–404
–896
Bending moment: kNm
617
Deflections: mm
9.0
The system is analysed again with a pure folded plate (FW) and a shell/T-beam (FB)
model. The flange width of the T-beam is equal to the effective width, according to
Eurocode 2 (2004).
The bending moments of both models are shown in Figure 5.43. Both approaches are in
good agreement. This demonstrates that the FB model can be used for this system too.
The bending moments in the transverse direction (mx) correspond very well in the field
area. Significant differences can only be observed over the web of the girder. The bending
moment my in the direction of the T-beam differs by approximately 15%. Bigger values
are obtained with the shell/T-beam model.
Because the flange of the T-beam exists twice in the slab and in the beam, the stiffness of
the system is overestimated. This affects the deformation. The maximum deflection of the
Table 5.4 Member forces of the manual analysis against the FE model
307
Figure 5.42 Slab supported on a small downstand beam – system and FE model
Modelled region
A
7m
Y
X
B
20
20
2.6 2.6 2.6
20
folded slab model is 8.75 mm, whereas the FB model gives 7.81 mm only ( ¼ 12%). The
results can be improved if the normal stiffness of the shell elements in the effective width
of the flange and in the direction of the girder is set to zero. Nevertheless, the effect of this
measure is relatively low.
The computer program estimates longitudinal reinforcement on the upper side of the
slab in the region of the downstand beam due to positive bending moments. Due to
Figure 5.43 Bending moments in the plane shell – (FW) against FB model
Folded shell Shell with beam Folded shell Shell with beam
(FW) (FB) (FW) (FB)
mx my
7.8 7.4 22.0 25.4
12.9
2.6 m 2.6 m
308
Shear force in kN
–155.
Bending moment in kNm
132.
111.
Deflections in mm
6.8
the linear elastic analysis, the numerical approach cannot realise that the reinforcement is
more effective, if it is put in the upper layer of the downstand beam.
The manual analysis, which is based on a stiff support of the slab by the downstand
beams, should not be used for this flexible structure. The equivalent forces on the
2.6 m 1.5 m
7m
35 20 35 1.03 m 20 1.03 m
Support Field
cross-section cross-section
279.
Shear forces in kN
245. 1.0d
–163.
1.0d
309
T-beams are overestimated, as the results shown in Figure 5.44 illustrate. The member
forces of the T-beam in the manual analysis are 2.5-times bigger than the real value of
the FW analysis.
5.3.3 Summary
The member forces of a slab supported on beams can be estimated very well with a rather
easy shell/T-beam model. A difference of only 15% with regard to the folded shell model
is estimated in the aforementioned examples.
If a T-beam is arranged in the centreline of the shell (model FWPL), the automatic
design of the shell and the downstand T-beam is possible. Nevertheless, the stiffness
of the T-beam is overestimated, which may result in an unsafe design of the slab. In
both models, the amount and arrangement of the reinforcement has to be checked
carefully.
5.5. Singularities
Singularities of the internal forces can be observed in shell structures as well as in plate
systems. They are caused by the same numerical errors. Therefore, only one practical
problem – single forces on shell structures – is discussed in the following example.
310
Cross-section
(depth increased) Top view
7.8
6.5
3.9
Coupling
1.3
0.0
0.0
20.0
Centre of gravity Beam elements
Shell elements
of steel girder
As explained already for plate structures, single forces should only be applied to an
FE shell model if the internal forces close to the loaded area are not needed for
design. Otherwise, the estimation of high bending moments and membrane forces
close to the concentrated load may result in an uneconomical design. This will be
demonstrated on the silo structure (height h ¼ 56 m and diameter d ¼ 25 m), which was
already mentioned in Section 5.1.2 (Figure 5.48).
Figure 5.49 shows the distribution of the bending moments and normal forces in
circumferential direction for a single force of F ¼ 5 m 5 m 100 kN/m2 ¼ 2500 kN. A
comparison of this result with that of a distributed uniform load over an area of
5 5 m shows that the internal forces are higher but on the safe side (see Table 5.5).
The greatest bending moment in circumferential direction mx under a concentrated
load is three times higher than if a distributed load is applied.
28 m
ny ny
mx
pp my
F
nx
28 m
28 m
311
Figure 5.49 Bending moment mx and membrane force nx in the loaded area (point-load)
Bending moment mx
35 m
100
–50
–50
25 m
100 50
0.0
0.0
0.0
0.0
Z
Bending moment mXX from –73 to 667
–12.5 X Y 0.0 12.5
Membrane force nx
35 m
0.0
0.0
–100
200
100
25 m
0.0
0.0
0.0
0.0
–100
Z
Membrane force nX from
–12.5 0.0 –112 to 968 kN/m 12.5
X Y
It should be noted that the maximum values will greatly increase with the number of
elements in the loaded area.
312
Table 5.5 Internal forces (mesh: 64 44 elements) loaded area – single force
Loaded area 5 5 m: q ¼ 100 kN/m 113 627 571 387 64 209 24 116
Point-load: Q ¼ 2500 kN 117 968 590 576 73 667 31 541
The main purpose of such an analysis is to get results that are closer to the real behaviour
of a structure and to get a more economic design and construction.
In Section 2.11, it was demonstrated that significant problems already appear in the
nonlinear analysis of columns and beams. The materially nonlinear calculation of
shells and massive concrete sections is considerably more complicated.
In the following, the state of the art in the field of nonlinear analysis of concrete
structures and the software development in this field cannot be mentioned in detail.
This complicated field of research can only be understood by experts. Therefore,
only the fundamental problems of material nonlinear FE calculations are illustrated.
Solutions for the following problems are needed.
313
Figure 5.51 Stress–strain diagram for concrete under uni-axial compression (on the left) and uni-
axial tension (on the right)
σc σc
Stresses σ1 = σ2 = σ3
Concrete
fc dg = 16 mm ft
Mortar
σcu dg = 4 mm
εc (<0) εc (<0)
εc1 εcu Strain ε1 = ε2 = ε3 εct
314
σ3 = 0 =σ
2
–0.4 σ1
Deviator-
surface
fcm n
sio
–0.8 Ten
–σ2
–1.2
fcm
–σ1
Crack
Rebar
315
oo oo
Discrete Smeared
be analysed with the last-mentioned model. The same is valid for the numerical
modelling of cracks. The discrete consideration of the cracks generally requires a
modification of the element mesh during the calculation (Figure 5.54). In general
cases, this is connected to a big expenditure, which is why most FE programs use
a smeared crack model.
g Numerical procedures: Robust numerical integration algorithms that guarantee the
exactness of the solution are necessary for a nonlinear calculation.
The analysis of shear walls and beams with plane shell elements under high shear forces
presents big problems.
In the field of research, nonlinear FE analyses are used mainly to study the behaviour of a
test specimen after the test has been conducted. Within such calculations, the material
parameters, the structural system and the arrangement and amount of the reinforcement
as well as the actions are most exactly known. Furthermore, enough time exists to verify
the numerical model thoroughly. These conditions are generally not given in the
construction practice.
The reinforcement arrangement is outlined in Figure 5.56. It becomes clear that the
lapping of the horizontal cantilever reinforcing bars (ds ¼ 28 þ 16/10 cm) with the very
massive vertical reinforcement in the column (ds ¼ 2 28/10 cm) is absolutely insufficient
by the loops ds ¼ 16/10 cm. More than 50% of the required reinforcement is missing in
this region.
Unfortunately, this mistake was noticed after the construction of the column and after
the left span of the bridge superstructure was already erected. The load-bearing capacity
could not be proved by means of strut-and-tie models. Therefore, the engineer in charge
for the design of the structure conducted a complicated 3D material nonlinear FE
analysis (volume elements, discrete reinforcing bars). This analysis proved that the
load-bearing capacity of the construction is sufficient.
316
670 cm
Top view
550
150 180
25 120
ds = 28/10 + ds = 16/10
180
ds = 16/10
ds = 28/10 +
150
ds = 28/10
317
Nevertheless, the responsible checking engineer still had doubts about the results of the
FE analysis. Even by means of the complicated FE analysis, a simple strut-and-tie model
or a reasonable force transfer mechanism could not be formulated that could prove the
stability of the special column head. Furthermore, the bridge is a statically determined
single-span structure without any redistribution of loads in the ultimate limit state.
Thus, the stability of the special column is needed for the safety of the whole structure.
With the used commercial, well-known FE program, principal software mistakes are not
likely to occur. The graphics of deflections, stresses and crack patterns were reasonable,
but the exact values could not be verified. Nevertheless, the checking of the assumptions
used in the numerical analysis, in this case of the material parameters, uncovered a
serious mistake. In the FE model of the cantilever column, the tensile strength fctk;0.05
according to EC2, Part 1 (Eurocode 2, 2004), had been used. With this value, no
horizontal cantilever reinforcing bars are needed at all due to the massive cross-section
of the cantilever section, as can be easily demonstrated by manual checks. In the real
structure, the available tensile strength of concrete might be significantly lower due to
the hydration restraints of the massive sections and the dynamic loads on the bridge.
The design checker requested a second FE analysis with half the concrete tensile strength
(0.5fctk;0.05), which showed instabilities. The structural safety was insufficient and,
consequently, the cantilever column had to be retrofitted.
This example clearly demonstrates that with complex, nonlinear FE calculations, a large
amount of scepticism is warranted with regard to the results. In particular, detailed
engineering experience is necessary and not just theoretical knowledge of the FE
method. Hence, a complicated materially nonlinear FE analysis should only serve to
reduce the available safety margin of an easy and clear simplified model, and not as a
substitute for the design engineer lacking in knowledge about the force transfer and
behaviour of the system. The design engineer who is in charge must be convinced of
the plausibility of the inputs and results by detailed checking.
REFERENCES
Bergan PG (1998) Computational Challenges for the Finite Element Method in Offshore
Engineering. Conference Finite Elemente in der Baupraxis, Darmstadt, 1998.
CEN (European Committee for Standardization) (1997) Eurocode 1, Part 4: Basis of
Design and Actions on Structures – Actions on Silos and Tanks. December, 1996.
Brussel.
CEN (2004) Eurocode 2, Part 1: Design of concrete structures – General rules and rules for
buildings. December.
Czerny F (1999) Tafeln für Rechteckplatten. In Betonkalender (Eibl J (ed.)) 1999, Volume 1.
Ernst & Sohn, Berlin.
Grasser E et al. (1991) Hilfsmittel zur Berechnung der Schnittgrößen und Formänderungen
von Stahlbetonbauwerken; Deutscher Ausschusses für Stahlbeton Heft 240. Berlin.
Gudmestad OTG and Coker JWA (1988) The Sleipner Platform. SPE 18344: 111–126.
Hennig E (1971) Experimentelle und analytische Untersuchung der Spannungsverteilung in
örtlich belasteten dünnwandigen zylindrischen Schalen. Fortschrittsberichte der VDI-
Zeitschriften, Reihe 1, Nr.: 33, Düsseldorf.
318
Holand I (1997) SINTEF Report No. 17: Sleipner A GBS Loss, Main Report. Trondheim.
Hughes TJR (1987) The Finite Element Method. Prentice Hall, Englewood Cliffs.
Jakobsen B and Rosendahl F (1994) The Sleipner Accident. Structural Engineering
International 3: 190–193.
319
Chapter 6
Three-dimensional building models
The future points to more and more complex FE models, as one can notice from the
numerous contributions in technical periodicals and conferences of the past years.
Why should one disassemble a building into its different structural components, into
simple equivalent structures, when the analysis and the design of the whole building
by means of a 3D FE model is faster and also more accurate? Next, the numerous
load combinations according to the codes can hardly be handled manually.
Daily experience shows that results from computer analyses are often trusted with blind
faith. The required checking of the model as well as a clear documentation of the main
results are often lacking, because of time constraints. This development raises concerns
regarding the safety of structures in the future.
The collapse of the platform ‘Sleipner A’ (see Section 5.1.1) has demonstrated the
consequences of blind trust in FE analysis. The more complicated the numerical
model, the more difficult it is to maintain a global overview of the structure and to
check the accuracy of the results. The latter is the responsibility of the structural designer,
321
not the checking engineer, as is often assumed. In addition, a software company will
rarely be responsible for errors in their code. The structural engineer must be convinced
that his or her numerical analysis is correct. He or she should always be aware that a
faulty analysis may endanger human lives, for which the designer is responsible.
Not to be misunderstood; the author does not express himself against complicated FE
calculations. He would like to merely point out to the software user the dangers which
can result from insufficient checking of his FE analysis and a lack of experience in this
field. In special cases one will always try to model the structural behaviour of a building
as accurate as possible with nonlinear 3D models. Nevertheless, for normal buildings the
big expenditure is not justified in most cases in relation to the expected reduction in
required building materials.
Besides, it is often overlooked that the FE method is a numerical approach only, and is
based on numerous assumptions and simplifications (Bathe, 1982; Hughes, 1987;
Zienkiewicz and Taylor, 1989). Hence, the accuracy of FE analysis is limited. The results
could even be wrong and unsafe. Also, the nonlinear material behaviour of reinforced
concrete can have significant influence on the force distribution in a structure. Therefore,
the results of FE design are not automatically correct even if the input is error-free.
The essential problem of big FE models does not exist in the element mesh generation,
but in the control of the input and output of the analysis. Much effort and experience in
the field of concrete engineering and numerical modelling is needed to check a complex
FE analysis, because no easy, program-controlled procedures are available up to now.
The behaviour of a building must be modelled by simple equivalent members. This is
the only method that gives certainty to the correctness of results from FE calculations.
Experienced engineers are needed for this task (fortunately).
All principal problems of truss, beam, plate and shell elements have been discussed in
Chapters 1–5. It should be remembered that a slab element has no normal forces in its
midplane (pure bending). Discontinuity regions of beams and singularity regions of
spatial structures are common problems of FE models. In the following, the problems
and limits of complex FE building models are illustrated by two real structures, an
office building (Figure 6.1) and an arch bridge (Figure 6.13), in which the author was
engaged as a designer and a checking engineer, respectively. It will be demonstrated
that no software is able to design these structures, at least in some areas. A careful,
independent checking of the numerical results is always required.
Checking the safety and serviceability of FE building models basically does not differ
from simple structures. However, there are some basic differences between manual
and FE design.
1. Documentation of the analysis: With manual calculation, one limits the output to
the essentials because of time limitations. In contrast, the printing of all results of
a complicated FE analysis is quicker than summarising the relevant outcome to a
minimum, to the relevant results only (Section 6.5).
322
323
soil parameters show a big scatter. Therefore, upper and lower bound values should be
considered in the FE analysis. This limit analysis has to be conducted in general, whether
it is an easy arch bridge (Section 6.4) or a complex building (Section 6.2).
For an office building, the size of the elements must be mainly checked in the support
regions (columns, walls), where a big strain gradient appears. This will be illustrated by a
simple one-way slab. Figure 6.2 shows the distribution of the bending moment my parallel
to the boundaries and the shear forces vx in the region of the facade as a function of the
number of the elements in the outer span. The analysis is based on a simplified equivalent
system, a one-way slab that is clamped on one edge. The slab has a span length of l ¼ 7.8 m.
The distribution of the member forces close to the columns is not influenced by the number
of elements. Nevertheless, with a coarse discretisation, the member forces are not estimated
in the relevant section at the face of the columns, but in the centre of the nearest big element,
far away from the correct location. This may lead to an unsafe design. For the accurate esti-
mation of the transverse bending moment mx, the region between the façade columns has to
be discretised with a sufficient number of elements. With only four elements, half of the
correct field moment mx,f is estimated. The bending moments at the column heads are
nearly independent of the size and number of the elements.
The elements should be so small that the member forces are independent of any further
decrease of their size. Exceptions are singularity regions. Here, the member forces
increase with the number of elements to ‘infinity’. To check whether the high member
Figure 6.2 Member forces of a one-way slab (column strip) for different numbers of elements
100
Shear force
80
Face of column
mx
shear force: kNm/m, kN/m
Bending moment my and
60
1 my
40
1 9 elements
x
20 18 elements
11.2
6.6 36 elements
0
–10.7 0.5 1.0 1.5 2.0 m
–20 Bending moment 100 V M
50
–41.4 0
–40 1 2 3 4 5 6 7 8
–50
–50
324
forces of an FE analysis are real or only caused by a numerical problem, one can analyse
the structure with different element meshes. This is not done, since a big effort is needed
to create a different FE mesh. As shown before, local problems can also be examined by
equivalent substructures.
325
Figure 6.3 Principal moment mI (on the left) and main shear force vI (on the right) in the slab as well
as singularity regions
Pin support
Openings
Pin support
a beam element should not be connected to a plane shell element (see Section 3.2.3),
as the latter has no rotational degrees of freedom. Therefore, the rotation of a node is
not restrained, and the bending moment in the beam at the joint becomes zero. This
problem can be solved with special nodal coupling or by the extending of the truss
into the shell elements. The type of coupling should be checked carefully to avoid
unintentional constraints. The office building has, in this respect, a very complex
region. In the outside façade, a column is jointed with a spandrel beam, a slab and a
shear wall (see Figure 6.4).
326
shear wall-plate-
column-joint
beam-slab-
shear wall
joint
section
Wall t = 25 cm Columns 80/80 cm
shell elements beam elements
The minimum reinforcement As,min of a fully tensioned section can easily be estimated
from the equilibrium condition before and after the cracking. In the crack, the
reinforcing steel has to take the whole tensile force.
P
Fsd ¼ As,min . sd ¼ Fct ¼ Ac . fct,eff
where:
sd is the permissible tensile stress in the reinforcing steel, depending, for
example, on l, ds and wk
fct,eff is the actual tensile strength of concrete at time of cracking.
In Eurocode 2 (2004), some parameters (kc , k) are added to the preceding equation to
consider special effects, such as the thickness of a member (As,min . s ¼ kc . k . fct,eff . Act
EC2 eq. 7.1).
The design of the fully restrained slab between the elevator shafts (see Figure 6.5),
according to EC2, Part 1, x7.3.2, with fctm ¼ 3.2 MPa (C35/45), gives a reinforcement
requirement of as ¼ 9 cm2/m (top and bottom), which is much more than is needed to
carry the external loads. Inexperienced engineers who do not know the background of
the minimum reinforcement will just load the structure by the temperature change and
shrinkage effect, respectively. This procedure results in a very high, uneconomical
reinforcement rate of as ¼ 100 cm2/m, as linear elastic material behaviour is assumed.
Thus, it not recommended in an FE analysis to apply thermal or shrinkage effects to
the concrete structure if the big decrease in stiffness due to cracking of concrete is
not considered. On the other hand, if the nonlinear behaviour of concrete is applied in
the analysis, one must assume a high tension-stiffening effect (TS) to get realistic
results. The restraint diminishes if the stiffness of concrete between the cracks is
neglected (TS ¼ 0). A nonlinear FE analysis, which is based on this assumption, gives
as 0 cm2/m. The lack of minimum reinforcement may cause a sudden collapse of the
slab.
327
Figure 6.6 shows the deformation of the columns and the slabs in the most upper floors
of the building. Unexpected, big deflections of the slab close to the inner columns can be
328
seen. First, one may assume that these unexpected results are caused by errors in the FE
model, incorrect support of the system or an insufficient cross-section of both columns.
However, this is not the case. Neglecting the construction sequences result in such
unrealistic deformations of the slab. In a usual FE analysis, the final structure is
loaded by the dead load of the members in one step. This leads to big deflections of
the columns as compared to the stiff inner walls. In reality, the various floors are built
one after the other, and the previous deflections are compensated. The uppermost slab
shows no deformations due to the dead load of the floors underneath. The big deflections
point out the fact that the creep and shrinkage effects of the columns are to be followed in
the design of the building.
In reality, each floor is built one after another. This results in the deflections shown
in Figure 6.7. Total different deflections in the edge columns are obtained if the final
structure is loaded in one step only (Figure 6.8). Thus, different member forces are
obtained in the slab. As the slab has a small bending stiffness with regard to the span
length, the deflections result in small differences in the bending moments (see
Table 6.1). This is not the case if the span length is to be reduced, for example, to 1.95 m.
where:
329
Figure 6.7 Deflections of the structure during construction (after hardening of concrete)
l
1u 2
lc propping
1
end of phase 1
After hardening
of slab no. 3
1u 1u 3
F After concreting F
of slab no. 3
1u 1u 2u 2
F F 2F
1
end of phase 1 + load of slab no. 3 = end of phase 2
1u 1u 4
1F
F
1u 2u 3u 3
1F 2F
F
2u 1u 3u 2
F 3F
2F
1
end of phase 2 + load of slab no. 4 = end of phase 3
1u 1u 5
1F
F
1u 3u 4u 4
1F 2F
F
3u 2u 5u 3
2F F 3F
3u 1u 4u 2
3F F 4F
1
end of phase 3 + load of slab no. 5 = end of phase 4
1u 1u 6
1F
F
1u 4u 5u 5
1F 2F
F
4u 3u 7u 3F 4
2F F
5u 2u 7u 3
3F F 4F
4u 1u 5u 2
4F F 5F
1
end of phase 4 + load of slab no. 6 = end of phase 5
330
Figure 6.8 Deflections of the structure with and without construction phases
1u
15u
5u
14u
12u 7u
9u 7u
5u 5u
F · Ic
u1 =
E · Ac
models. This structural behaviour in the ultimate limit state cannot be modelled with an
FE analysis, based on a linear material behaviour.
331
Fz
T
Fx Cross-section
5.00 m
0.30 m
7.00 m
x
y
30 m
z
m
y 7
5m
This simple structure is used to verify the shell model as the deflections, member forces
(M, V, T ) and required reinforcement can be easily estimated by manual analysis.
First, only a horizontal force of FEd,x ¼ 1 MN is applied at the top of the cantilever beam.
This load requires a total vertical reinforcement of As ¼ 129 cm2 at the support. The FE
program estimates a slightly higher value of As ¼ 141 cm2 ( ¼ 10%) due to the smaller
internal lever arm of the linear elastic analysis (Figure 6.10 top).
The difference between manual and FE analysis increases if, in addition to the horizontal
force, a normal force is applied. With FEd,x ¼ 1 MN and FEd,z ¼ 6 MN (cd,N ¼
0.84 MPa), a total bending reinforcement As ¼ 62 cm2 is required due to the
manual analysis. The FE calculation gives a 64% bigger value (As ¼ 102 cm2). This
result makes it clear that a complex shell analysis must not always result in a more
economical design, as is believed quite often.
An FE program designs an element on the basis of elastic material behaviour and the
resulting member forces. Nevertheless, in the cracked state, external compressive
forces (Fz) are not equally dispersed over the whole cross-section, as is assumed in the
FE model. It directly reduces the required tensile reinforcement. This becomes clear
from the generally known design equation:
1 MEds
req As ¼ þ NEd
fyd z
332
Figure 6.10 Vertical reinforcement at the support for two load cases (shell analysis)
6.72
M = 30 MNm, Pz = –6.0 MN)
4.09
6.72
An FE analysis always gives member forces and reinforcement values on the safe side in
the case of bending. Nevertheless, this does not apply to the design for shear and torsion,
as will be shown in the following. Stirrups of as,w ¼ 6.1 cm2/m are needed to carry a
horizontal force of FEd,x ¼ 1 MN according to manual design for shear. The FE analysis
gives a maximum value of only 1.23 . 2 ¼ 2.5 cm2/m (Figure 6.11). Furthermore, the
arrangement of the estimated horizontal reinforcement has nothing in common with
stirrups. The FE analysis is unsafe. Furthermore, only a surface reinforcement for
every FE separately is given. Hence, it is not evident for the present hollow box cross-
section that stirrups are required, which should tie the compression zone and the tension
reinforcement together. The design problem of shell models can only be solved by
integrating the stresses in various sections numerically to get member forces (N, V,
M), which can than be used to design the beam manually.
For a torsional moment of TEd ¼ 10 MNm, the FE design as well as the manual design
gives a horizontal reinforcement, or stirrups, of asw ¼ 3.3 cm2/m, for each wall
(Figure 6.12).
Asw TEd 10:0
¼ ¼ 104 ¼ 3:3 cm2 =m
sw 2 Ak fyd cot 2 ð7:0 5:0Þ 435 cot 450
333
Figure 6.11 Horizontal reinforcement at the support of the beam, load FEd,x ¼ 1 MN
0.9
1.2 cm2/m
1.2 cm2/m
0.9
Nevertheless, this full correspondence between FE and manual analyses arises only if one
assumes an inclination of compression stresses to the vertical beam axis by an angle of
¼ 458, the same as in the linear elastic analysis. Nevertheless, smaller values for
(188 4 4 608), which result in considerably lower reinforcement, can be used. If the
section is loaded by a bending moment and normal forces, in addition to a torsional
moment, the main tensile stresses in the compression area are almost vertical. This is
illustrated in Figure 6.12, where the orientation of the main membrane forces as well
as the horizontal reinforcement in the support region due to a load of TEd ¼ 10 MNm
and FEd,x ¼ 1 MN is shown. The results of the FE design correspond neither to the
amount nor to the arrangement of the reinforcement for a cantilever beam with a
hollow box cross-section under bending and torsional loads.
Figure 6.12 Main membrane forces in the wall (left) and the horizontal reinforcement at the
support (right)
TEd = 10 MNm
FEd,x = 1 MN
30 m
z Shown
region
x 2.3 cm2/m
334
Due to the screwed superstructure and the soft foundation of the arch bridge, it seems
reasonable to use a 3D FE model of the whole structure to design the bridge, including
the foundations. Plane shell elements are used for the arch, the side and wing walls as
well as for the abutment beams. The piles are modelled by truss elements that were
horizontally supported by distinct elastic springs.
The structure is mainly loaded with the dead load, the ground surcharge of the
bridge, traffic loads and temperature effects. Furthermore, time-dependant concrete
deformations such as creep and shrinkage have to be considered in the design of the
‘integral’ bridge without bearings. The traffic loads are not significant due to the high
load by the ground on top of the arch. Hence, it does not seem necessary to consider
numerous load arrangements.
Abutment
beam
b/h = 2/2.2 m
Bored piles
d = 1.2 m
Plan view
335
Figure 6.14 Portal frame bridge: structure, loading and member forces (truss model) due to ground
surcharge
–188.
61.5 kN/m 61.5 kN/m
.
24
–3
–3
24
Abutment beam –387. –387.
.
24.005
–387. –387.
.
46
46
18
–6
63.
645.6 –646.
1.
–187. –187
.
81
–81. 81.
–1
These conditions rarely apply in reality. Thus, the experimental value has to be adjusted
to the real structure.
The reinforced concrete arch is supported by the abutment beams and the piles. The
latter are modelled by beam elements, whereas shell elements are used for the rest of
the structure. Because the plane shell (membrane) elements have no rotational degrees
of freedom, the joint of the piles with the abutment beam must be considered by coupling
the element nodes together or by extending the beam elements of the pile into the shell
elements of the abutment beam (see Chapter 3, Figure 6.16). No artificial restraints
should be generated in the case of nodal coupling. If the upper node of the pile is
fixed with two nodes of the shell elements, temperature loads may cause infinite stresses
in this region.
336
Figure 6.15 Member forces in the pile head in relation to the oedometric modulus of the soil (truss
model)
250
relation to Es = 50 MPa: %
Member forces of piles in
209
200
180
150 141
0
0 20 40 60 80 100
6.25
12.5
25.0
50.0
600
Member forces of piles
500
in kN bzw: kNm
400
300
0
0 20 40 60 80 100
6.25
12.5
25.0
50.0
Figure 6.16 Connection between the piles and the abutment beam
Coupling
Pile
337
FE analysis cannot give appropriate member forces. The abutment beam has a massive
cross-section of 2.0 2.21 m, and thus a nonlinear strain and stress distribution is likely
to occur. This will be examined further in the next section.
It should be noted that an FE analysis is based on nodal forces and displacements. There-
fore, arbitrarily distributed patch loads, such as, for example, wheel loads on bridge
decks, are automatically converted by the used program into equivalent nodal forces.
Hence, the size of the elements has a significant influence on the member forces in the
area of patch loads (see Section 5.1.2).
A temperature difference of TN ¼ 16 K was applied between the superstructure and the
substructure (abutment beam and the piles). This value is much smaller than the values
given in the codes as the whole substructure is under water and the arch is covered by the
ground. About 40% of the elastic stiffness is used in sections where the concrete is under
tension, in order to take into account the nonlinear material behaviour.
If the temperature of the screwed reinforced concrete arch increases with regard to the
substructure, big tensile forces are generated in the abutment beam in the horizontal
direction. The membrane force due to TN ¼ 16 K is approximately six times greater
than for the other loads (Figure 6.17). A design for this high tensile force would result
in a large amount of rebar, which in some sections cannot be arranged. A design
based on an elastic material behaviour seems not to be required anyhow. A minimum
reinforcement should be arranged in the beam rather than designing it for high-restraint
forces. The abutment beam should be designed for fully tensioned sections and a crack
width of wk ¼ 0.2 mm. This measure results, even for this massive cross-section, in a
significant reduction of the required reinforcement. It should be noted that linear elastic
analysis of a structure under restrain actions (temperature, shrinkage, etc) is, in most
cases, not useful.
Figure 6.18 shows the member forces of a rather simplified equivalent structure of the
arch, a single-span girder clamped on both ends into a massive beam. Analysis is
338
Figure 6.17 Horizontal membranes force nx of the abutment beam above the piles for various
loads
conducted with a span width of 20 m (system 1: fixed in the centre axis) and 18 m (system
2: fixed at the inner face of the support). The bending moment of model 1 at the face of
the support (x ¼ 1.0 m) is Ms ¼ 238 kNm, while model 2 gives a 13% higher value of
Ms ¼ 270 kNm. The differences between both models increase when the span length
decreases. A span length of ln ¼ 10 m, for example, results in a 28% difference between
both approaches. The fixed end moment of the 20-m-long beam is 40% bigger, with
Ms ¼ 333 kNm than the value at x ¼ 1.0 m. A design for the maximum value is
uneconomical.
For the given structure, the arrangement of the reinforcement in the parapet walls should
be examined. The wall shows very big tensile membrane forces (Figure 6.19). Hence, it
339
Figure 6.18 Member forces of a fully restrained beam with span lengths of 20 m and 18 m,
respectively
Real
structure 20.0 m
2
q = 10 kN/m
System 1
(l = 20 m) 20.0 m
System 2
(l = 18 m) 18.0 m
–400
Bending moment and shear forces
-333. M20m
–300 -270.
-238.
–200
M18m
–100 V
0
10 20 m
100 138.
167.
200
has to be clarified whether the tensile stresses are caused by tensile action effects (e.g.
temperature) or by bending of an upstand unsymmetrical T-beam (Figure 6.19). In the
first case, the reinforcement has to be arranged over the whole height of the beam,
whereas in the latter case it has been concentrated at the top edge (bending). From the
Figure 6.19 Horizontal membranes force and required reinforcement of the parapet wall
–793 kN
–1193 A
Shown region
A
340
Soil pressure
on shell
2.77
At abutment elements
3.36
3.95
beam
1.18
40
40
At midspan
distribution of the membrane forces (Figure 6.19), it becomes clear that the parapet
wall shows a bending behaviour and, therefore, the required reinforcement has to be
concentrated at the top edge. The parapet wall acts as a web of an unsymmetrical
upstand beam and has to be designed like a girder. Therefore, the local shear forces
have to be supported by stirrups that must tie the compression zone and the tensile
reinforcement together. The distribution of the shear forces should be examined
carefully.
The FE program cannot estimate the tensile forces that result from the dispersion of
the high pile forces into the abutment beam. For such massive structural elements as
abutment beams, strut-and-tie models are needed for design. Linear elastic FE analysis
may be used to develop a realistic stress-field model (Figure 6.22).
341
46
37 .
63
8.
1
6.
636. 12.6
Bending Reinforcement
moment
due to
surcharge
232. (plane truss) 18.5
wrong correct
(anchorage of rebars) 1 .3 (lapping of rebars)
lc =
1.11
1.10
lc = 0.90
2.00
342
M+N+V
Bending moment
2.21
1.20
2.00
based on the integration of the local vertical stresses in each element. Figure 6.23 shows
the result of such an analysis for the abutment beam. This approach is not correct here.
According to the numerical design, stirrups of approximately 12 cm2/m2 are needed in
the upper region of the beam. The arrangement of such stirrups would require a big
effort. By means of a strut-and-tie model (Figure 6.22), one can demonstrate that no
shear reinforcement is required.
11
72
11 11
11
12
44
11
20
11
11
11
6
6
6
11 4
11
11
11
25
18
13
12
14
14
Section
A–A
A
2.20 m
2.00
A
Bored piles d = 1.20 m
343
The high shear forces in the upper left part of the abutment beam are not relevant for
design because of the massive region.
The example makes it clear that the shear design of massive components must be
questioned.
The checks require an entire and clear documentation of the structural data, the actions
and the essential results of the FE analysis as well. Although this should be quite normal,
the quality and quantity of the documentation of an FE calculation is a permanent
matter of discussion between the design and the checking engineers. Hence, in addition
to the general remarks in Section 6.1, some further explanations are given in the
following for complex building models.
Numerous material parameters are needed for the design of a structure, like the modulus
of elasticity (see Section 2.11.1) and the Poisson’s ratio (Section 4.3). All parameters
show a significant scatter in practice, which may influence the results of the analysis.
Thus, all material parameters and the used material model, as well as the allocation to
the various structural elements, have to be checked.
Furthermore, the assumptions of the FE model have to be checked. Is the strain over the
height of the cross-section linear in all relevant regions of a beam and a plate, or do the
discontinuity or singularity regions need further attention? Does the reduction of stiff-
ness in a cracked concrete region due to tensile stresses or torsional moments have a
significant influence on the distribution of the member forces in the structure? The FE
mesh is to be checked for possible incompatibilities if different types of elements, for
example, beam and plane shell elements, are jointed together. Massive sections should
be designed with strut-and-tie models.
The control of the level and the arrangement of all relevant load cases (including
imperfections) are difficult and time consuming. The partial safety concept has led to
344
Dimension
X * 0.759
Y * 0.652
M 1: 200
30. 20. 10. 0. –10.
Scale
Contour
x
0.
Global co-ordinate
system
X System plot
Z
Name of data file
M 1 : 250
0.40
0.42
0.44
0.48
0.53
0.59
0.67
0.77
0.88
1.00
1.07
20.
Y Z thickness of shell elements in m (max = 1.07)
10.
System plot – plan view
(bogen03.dat)
0.
an enormous number of load combinations that can hardly be overseen. The arrange-
ment of the loads may be checked graphically. The total support force of a load case
given by the program can be used as a quick check, whether the actions are considered
in the correct integral size. Loads on fixed nodes are not considered and printed out by
some programs, because they give no contribution to the deflection energy of the system.
In such a case, one must add the missing nodal forces to get the correct load for the
supporting structures. Constraint effects are to be controlled with regard to the cracking
of concrete.
The distribution of the strains, deflections, and member forces must be plotted in the
relevant sections for the relevant load cases.
345
Tensor plots of the main moments or main membrane forces serve for the view of the
flow of forces in a structure.
The delicacy of the structure on slight changes of the input parameters, for example, the
soil parameters or the stiffness of supports and members, should be checked by means of
parametric studies with lower and upper bound values.
The effect of geometrical imperfections given in the codes (Eurocode 2, 2004) should be
checked.
The preceding (incomplete) listed items demonstrate that the checking of big FE
analysis requires a large effort, which must be considered in the financial budget of
the project.
Nevertheless, all these controls cannot give absolute certainty about the correctness of
numerical analysis. This is possible only by means of independent comparative
calculations with simple equivalent structural models (e.g. simple beams).
One problem of big FE models should still be pointed out here. During the planning and
construction phases of a building, numerous changes of the original draft are often neces-
sary. In the case of the arch bridge, one pile was drilled at a slightly wrong location. As
the structural behaviour of the bridge couldn’t be simplified by the design engineer as he
believed in his 3D FE model, the whole structural system had to be calculated again.
Moving a pile in the numerical model is a very easy task. Nevertheless, the clear docu-
mentation of the new results as well as the comparison with the original calculations
was connected with a large expenditure.
346
The formal demands for a numerical analysis are given in, for example, the regulation for
the conducting and checking of computer-supported structural analysis (RI-EDV-AP-
2001) (Verein Deutscher Prüfingenieure, 2001). This or similar regulations should
generally be fixed in the contract documents of any structural design. If this directive
is postponed, the designer must convince himself or herself of the plausibility of the
inputs and results by various checks. In the following, the essential items of this directive
are listed.
The numerical output should be reduced as much as possible in favour of clear, graphical
outputs.
Among other things, the documentation of a structural analysis must contain the
following items:
Formal demands
Input
347
Figure 6.25 Shading plot of principal bending moment due to uniform loading
g member forces and deflections for the relevant load cases and load combinations
g design member forces and estimated reinforcement with cross-section distortions
and cross-section dimensions (for relevant load cases and sections only)
g material grades
g construction sequences, if considered
g restrictions for the use of the structure
g necessary intermediate results (e.g. support forces including the used safety
coefficients)
Please note that coloured plots should only be made if all copies of the design document
are coloured. This may cause significant costs in bigger projects. Black-and-white copies
of coloured drawings are useless, as can be seen in Figure 6.25. The grey shadings cannot
be assigned to the level of the main bending moment mI in the superstructure.
As in every structural analysis, the documentation of the design should be clear and easy
to be verified. Sketches are very helpful.
348
the ‘Auditorio de Tenerife’ (Figure 6.27), are compared. The famous building in
Australia required a construction time of 14 years (1959–1973). The building cost that
was originally estimated at £3.5 million increased in this time period by a factor of
more than 14, to approximately £50 million. The dream of the architect could only be
349
realised when the complex shells of the roof were substituted by a frame structure made
of prestressed T- and V-beams. In contrast, the impressive ‘Auditorio de Tenerife’, which
opened in 2003, is a shell structure that could not have been built without FE analysis.
During the construction (1998–2003), the original estimated building cost of €30 million
increased to approximately €80 million.
Development in the field of numerical analysis continues. After the 3D building models,
where the engineer controls the meshing, even more complex ‘building information
models’ (BIM) will be used. These models contain all the data of a building, including
the structural elements and the required reinforcement for any member, the construction
sequences, quantity and qualities of building materials etc. The whole FE mesh is gener-
ated automatically with the available data from a general database. Any changes during
design and construction of the building are automatically considered in all elements of
the building. This raises the following questions: Who is controlling the flood of informa-
tion? Do we still need experienced structural engineers if the design of a complex building
can be done automatically by software in a very short time?
REFERENCES
Bathe K-J (1982) Finite Element Procedures in Engineering Analysis. Prentice Hall,
Englewood Cliffs.
CEN (European Committee for Standardization) (2004) Eurocode 2, Part 1 Design of
concrete structures – General rules and rules for buildings. December.
Hughes TJR (1987) The Finite Element Method. Prentice Hall, Englewood Cliffs.
VPI (Verein Deutscher Prüfingenieure) (2001) Richtlinie für das Aufstellen und Prüfen EDV-
unterstützter Standsicherheitsnachweise (RI-EDV-AP-2001). Der Prüfingenieur 18.
Zienkiewicz OC and Taylor RL (1989) The Finite Element Method. London.
350
Index
351
352
Modulus of Elasticity 118, 118 stress resultants 141–145, 142, 144, 145
nonlinear behaviour 43–45, 44, 45 strut-and-tie models 166–169, 166
steel, stress–strain relationship 111, 111, tensile forces 142, 146
116 singularities 169–170, 169
stress–strain relationship 112–113, 113, shear design 342–343, 343
114, 116–117, 116, 117, 314, 315 and slender beams 162–166, 164
yield surfaces 315 stress/strain distribution
concrete floor joists cracked members 165, 165
dimensioning 206 uncracked members 165, 165
models 205–206, 205 strut-and-tie models 166–169, 166, 167,
Condeep-type platform 3 168
see also Sleipner A platform support
constraint modulus method 39, 39 modelling of 145–157
cracked members horizontal restraint 156–157, 157
considerations 3 springs 146–151, 149, 150, 151
cracks beam elements 151–156, 152, 153
modelling 316, 316 incompatible 153, 154, 325–326, 327
force transfer 315, 315 two-span, strut-and-tie models 167–169,
torsional stiffness 63 167, 168
Czerny tables, slabs 262–263, 262, 263 deformation, beam
creep 119
deep beams cracked, with TS 123–128
column supports cracked, w.t. TS 120–123, 120
beam element modelling 151–156, 152, uncracked 117–120
153 dimensioning
spring equivalent modelling 146–151, beams 110–114, 114
149, 150, 151 concrete joist floors 206
corner region, stress distribution 155–156, shear walls 158–162, 159, 160, 161, 162
156 spatial structures
definition 139, 140 Baumann model 231–234, 231, 232,
degree of freedom 153, 154 233
dimensioning 158–166, 159, 160, 161, 162 cracking 254, 255
detailing 160–162, 162 internal forces 254
internal forces, linear-elastic calculations simplifications 254–255
139 Stiglat model 255–256, 256, 261
membrane force distribution 157, 157 discontinuity regions 13
membrane model 153 definition 13
multi-span truss systems 14, 22, 22, 29, 29, 30, 325
loading forces 147, 147 discretization
stress distribution 147, 149 definition 2, 5, 5
strut-and-tie model 167, 168 modelling errors in 8–12
support settlement 146, 148, 149 displacements, T-beam bridges 88, 89, 90
reinforcement documentation of FE-analysis 346–348
horizontal distribution 158–162, 160, double T-beam bridges 91
161, 162 beams
vertical distribution 160, 161, 165–166 locations 92–93, 93, 94
single span numbers, transverse beams 95
membrane forces 142, 143–145, 144, 145 stiffness 95–98, 97
reinforcement requirements 141, deformation 93, 95, 101
162–1662, 163, 164 design 91
353
354
355
modelling errors (continued ) portal frame bridges 15, 15, 16, 335
nonlinear materials 8 abutment beam 337
reinforcement determination 10 beam model 16
modulus of elasticity, building materials bending moments 17, 18
118, 118, 119, 119, 203 dimensioning 338–344
Mohr’s failure criteria 159, 159 frame corner 338–339
detailing 339
nodal displacements, and nodal reactions effective span length 338–339, 340
5–6, 5, 6 FE-modell
nonlinear behaviour, concrete support conditions 335–336, 337
foundation beams 43–46, 44, 45 singularities 336–338
beams 114–136 output data 345, 346
material parameters 133–136 flow of forces 341, 342
tension stiffening effect 123–126, 124, foundations 16–17, 16
125, 126 loadings 15–17, 338, 341
use 115 shear design 342, 343, 344
yield surfaces 315 shell model 325
numerical analysis strut-and-tie model 343
problems
modelling errors 8–12, 9 re-entrant corners, singularities 238–242,
program errors 8 238, 240, 241
real loading versus Finite Element nodal
openings loading 8–10, 9, 10
singularities, slab 238–242, 238, 240, rectangular slabs
241 bending moments 262–263, 262, 264, 265,
beam 267, 268
deformation 61, 61, 63 shear forces 262–263, 263
modelling 56–59, 58, 59 reinforced beams
see also deep beams
patch loadings, silos 250–252, 253 resistance 111, 111
Pieper-Martens model, rectangular slabs dimensioning 110–114, 114
263, 269 nonlinear analysis 114–135
pile foundations shear/torsion design 113–114, 114
bored strain distributions 112, 112
bedding modulus 46–47, 46 stress–strain relationship 111, 111
bridge columns 47, 48, 49 tension stiffening effect 112, 113
forces carried 46, 46 reinforcement
horizontal deformation 53–54, 53 deep beams
inclined, bending moments 54–56, 55 horizontal distribution 158–162, 160,
load-settlement curves 52 161, 162
manual analysis 47, 50 vertical distribution 160, 161, 165–166
numerical model 46, 46, 51 determination, by modelling 10
rigs 50 hollow box girder bridges 20
soil stiffness modulus 46–47 slabs 183, 254–261, 255, 256, 261
truss model 47–48, 51 Reissner/Midlin plate theory 262
vertical spring stiffness 48–53, 52 residential buildings
caps, horizontal restraints 54, 54 slabs 176–177, 177
uses 46 bending moments 177, 178, 179
plates, nodal forces 9, 10 deflections 178
Poisson’s ratio 181–183, 181 shear forces 179
356
357
358
359
Sydney (Australia), Opera House 349–350, tensile force, strip foundations 211, 212, 233
349 tension stiffening effect 123–126, 124, 125,
126
T-beam bridges 86, 91, 104 torsional moments
see also double T-beam bridges, T-beams beams 67, 68
bending moments 99, 101, 106, 107 bracing systems, high-rise buildings
deck slab 84, 85 73–74, 72, 74
design 84 double T-beam bridges 93, 95
eccentric loads, torsional moments 99, hollow box girder bridges 80–82
100, 102–103 T-beam bridges, eccentric loads, 99, 100,
longitudinal girders 102–103
eccentric line load forces 106–107 torsional stiffness, T-beam bridges 89–90,
placement 105, 105 90
torsional stiffness 89–90, 90 tricells, Sleipner A platform 273, 273
T-beams truss model verification, shell models 74–75,
additional beam elements 284–286, 285 75
bridges 84–105, 86, 104, 105 truss systems
eccentric beam elements 284–286, 285, bored pile foundations 47–48, 51
291, 292, 293 discontinuity regions 22, 22, 29, 29, 30
effective width 298, 304, 304, 305, 305 equivalent cross-sections 63–64
knife-edge supports 284, 285 hollow box girder bridge 17, 20
member forces 290–292 portal frame bridges 15
modelling 283, 285 bending moments 17, 18
shell elements 284, 285 foundations 16–17, 16
shell models 287–288, 287, 288 loadings 15–16
web/flange junctions 283 uses of 13
shell structures, uses 283, 294 shear walls with openings 56–61, 57, 58,
stress distribution 293 59
structure 294, 294
support region forces 289 wrapping torsion 67, 76–79
tensile stresses, anchorage region
293–294, 295, 296 Zuse, Konrad 2
360