Reportbakker PDF

A finite element model for the deck of plate-girder bridges in-
cluding compressive membrane action, which predicts

the ultimate collapse load
Gert Jan Bakker

Amsterdam, August 25, 2008
A finite element model for the deck of plate-girder bridges including
compressive membrane action, which predicts the ultimate collapse load
Gert Jan Bakker

Section Concrete structures – Structural engineering
Faculty of Civil Engineering and Geosciences
Delft University of Technology
Keywords:
compressive membrane action, concrete slabs, bridge decks,
lateral restrained, numerical analysis, test comparisons, ultimate limit state,
bearing capacity
August 25, 2008

Preface
This research was carried out as Master thesis project at the section Concrete Structures at Delft
University of Technology in co-operation with Witteveen + Bos. With this research I got a chance
to compare difficult and long theory’s on compressive membrane actions with a more workable
solution, using a finite element program. This is the first step towards using compressive membrane
action in design calculations and to use it for recalculating existing bridges. Including compressive
membrane action into design calculations might lead to thinner and thus more economical bridge
design and may prove that existing bridges do not need maintenance or replacement as yet.
I would like to thank all the members of my examination committee for sharing their knowledge
with me, while I was writing this paper. I would also like to thank the company Witteveen + Bos for
providing me a workspace and letting me make use of their facilities. I address special thank to Ar-
jan Luttikholt, who helped me with the finite element modelling in DIANA.
Gert Jan Bakker

Amsterdam August 8, 2008
Examination committee:
Prof.dr.ir. J.C. Walraven, Delft University of Technology – Section concrete structures
Dr.ir. C. van der Veen, Delft University of Technology – Section concrete structures
Dr.ir. P.C.J. Hoogenboom, Delft University of Technology – Section structural mechanics
Ir. F.J. Kaalberg, Witteveen + Bos
Ir. L.J.M. Houben, Delft University of Technology
Delft University of Technology

Faculty of Civil Engineering and Geosciences
Section concrete structures – structural mechanics
i
Summary
With traffic getting denser in a fast rate, and trucks carrying heavier loads using less axles, the re-
maining lifetime of concrete decks in viaducts and bridges becomes uncertain. The load of the traf-
fic gets closer and closer to the ultimate bearing capacity of these decks, calculated according to the
Dutch code. However, no collisions have occurred yet.
Tests on both full scale and small scale reinforced concrete slabs showed that, if the edges of the
slabs where laterally restrained, the bearing capacity was significantly higher than the slabs that did
not have laterally restrained edges. After these tests, performed in the 1960’s various people did
research on this phenomenon. They generally came to the same conclusion: after cracking of the
slabs a compressive force is introduced which enhances both the shear and bending capacity of the
slabs. This phenomenon is called compressive membrane action.
The theories however consisted of long and difficult derivations, ending up in big and hard to read
formulas, which are of no use in practice. Furthermore, different derivations where made for bend-
ing and punching failure, making things even more complex.
With the introduction of faster computers and especially better finite element programs, which can
include non-linear material behaviour, it can be checked if models can be made which takes into
account this compressive membrane action. These finite element programs are used more and more
in practice. Consequently, laterally restrained structures can be designed in a more economical way.
These models can also be used to demonstrate if certain repairs or replacement are really necessary,
or that the structure has enough extra bearing capacity to postpone the maintenance.
In this paper it is tried to include this compression membrane action in a finite element analysis and
the results are compared to a theory for both bending and punching shear that includes compressive
membrane action. The results are also compared to experimental data, which is presented in various
articles. The results of the finite element models look very promising, using a concrete strength of
fcu = 35 N/mm2. For lower values of the concrete strength, the finite element models seem to give
values that are to high, and for higher values, the finite element model seem to give values that are
to low.
For bridge decks, which commonly have a concrete strength that is about 35 N/mm2, the finite ele-
ment model gives results that lie in an acceptable range based on experimental found data. The fi-
nite element program is used to predict the ultimate load of bridge decks of a ZIP girder system,
which is commonly used in the Netherlands. It can be concluded that the enhancement factor for
this type of deck has a value that lies around 1,5. The enhancement factor is here defined as the col-
lapse load found by the finite element calculation divided by the lowest value of an analytical de-
termined flexural and punching shear capacity calculation according to the Dutch code.
This indicates that laterally restrained bridge decks have more bearing capacity than follows from
an analytical calculation.
iii
Contents
Preface........................................................................................................................................i
Summary ..................................................................................................................................iii
Contents ....................................................................................................................................v
List of symbols ........................................................................................................................vii
1. Introduction.......................................................................................................................1
1.1. The goal of this study ..................................................................................................3
1.2. The structure of this study ...........................................................................................4
Part I: The theory of compressive membrane action ..............................................................5
2. Theory of compressive membrane action.........................................................................6

2.1. Enhancement of the bending strength by compressive membrane forces......................7
2.2. Enhancement of the punching shear strength by compressive membrane action......... 13
2.3. Comparison between bending and punching failure ................................................... 19
3. Analytic solutions ............................................................................................................ 23
3.1. Standard calculations methods used........................................................................... 23
4. Comparisons between the analytical solutions and test results ..................................... 27
4.1. The bending model.................................................................................................... 27
4.2. The punching model .................................................................................................. 29
Part II: The finite element modelling..................................................................................... 33
5. Finite element models...................................................................................................... 35

5.1. The geometry of the model........................................................................................ 35
5.2. Material properties .................................................................................................... 45
6. Finite element model for bending action ........................................................................ 49
6.1. Total horizontal restrained clamped model ................................................................ 49
6.2. Partly horizontal restrained model ............................................................................. 54
6.3. Comparison with test results...................................................................................... 58
6.4. Brittle versus tension softening cracking model ......................................................... 60
7. Finite element model for punching failure ..................................................................... 63
7.1. 3D solids model......................................................................................................... 63
7.2. axi-symmetric model ................................................................................................. 68
7.3. Enhancement factors ................................................................................................. 72
7.4. Comparison with test results...................................................................................... 73
7.5. Partly horizontal restrained axi-symmetric model ...................................................... 74
Part III: A Practical example ................................................................................................. 79
8. The deck of a common girder bridge ............................................................................. 81

8.1. Analytical solutions ................................................................................................... 83
v
8.2. solutions including compressive membrane action..................................................... 85
8.3. Overview of results ................................................................................................... 87
8.4. Existing bridges......................................................................................................... 88
9. Conclusions and recommendations ................................................................................ 91
9.1. Conclusions............................................................................................................... 91
9.2. Recommendations ..................................................................................................... 92
10. References.................................................................................................................... 93
Appendix A: Excel sheet for bending capacity ........................................................................ I
Appendix B: Maple sheet for punch capacity......................................................................... V
Appendix C: Some DIANA input files ................................................................................XIII
Appendix D: Experiments based on the Ontario highway bridge design code .................XIX
Appendix E: New Zealand code ....................................................................................... LXXI
vi
List of symbols
a radius of a slab [mm]
c distance to the neutral axis in the sagging yield moment [mm]
c’ distance to the neutral axis in the hogging yield moment [mm]
ck parameter related to the ratio of compressive and tensile strength of concrete [-]
d effective depth of the cross-section [mm]
d’ difference between the height and the effective depth of the cross-section [mm]
d0 length over which the concentrated load is spread [mm]
d1 outer diameter of the punched cone [mm]
h height of the cross-section [mm]
fcc uni-axial concrete compression strength [N/mm2]
fct uni-axial concrete tensile strength [N/mm2]
fcu cube strength of concrete [N/mm2]
fs,fy yield strength of reinforcement steel [N/mm2]
fc cylindrical concrete tensile strength [N/mm2]
f’c cylindrical concrete compression strength [N/mm2]
kd factor related to the height of the slab [-]
mu resisting moment at the mid depth axis at the hogging moment per unit width [Nmm/mm]
m’u resisting moment at the mid depth axis at the sagging moment per unit width [Nmm/mm]
na dimensionless membrane force in the mid depth of a slab [-]
nr dimensionless radial membrane force working on the surface of the failure cone [-]
nu membrane force at the mid depth axis at the hogging moment per unit width [N/mm]
n’u membrane force at the mid depth axis at the sagging moment per unit width [N/mm]
p perimeter [mm]
p perimeter of cone that is punched out [mm]
q distributed load over a length [N/mm]
q reinforcement percentage in the code of New Zealand [-]
r radius [mm]
r function of the failure surface over the height [mm]
t outward lateral displacement at the restrained edge [mm]
w0 critical deflection, empirical determined as 0,5 h [mm]
wi deflection at which membrane action starts, empirical determined as 0,03 h [mm]
A cross-sectional area [mm2]

A,B constants [m], [1/m]
Ash cross-sectional area of hoop steel per unit width [mm2/mm]
C compression force at the sagging yield moment per unit width [N/mm]
C’ compression force at the hogging yield moment per unit width [N/mm]
DA the internal energy dissipation per unit area in the deforming zone [N/mm]
E modulus of elasticity [N/mm2]
F concentrated load [N]
I impact factor in the code of New Zealand [-]
L length of the span [mm]
M moment [Nmm]
N intern force [N]
Nrs sum of the radial membrane forces [N]
R radius of the edge beam [mm]
Ri unfactored ultimate resistance in the New Zealand code [N]
P ultimate load in punching shear failure [N]
Pa analytical ultimate load [N]
Pe ultimate load from tests [N]
Pp predicted ultimate load [N]
vii
Q load distributed over an area [N/mm2]
S stiffness parameter of a laterally restrained slab [N/mm]
T’ tensile force in the steel at the hogging yield line per unit width [N/mm]
T tensile force in the steel at the sagging yield line per unit width [N/mm]
W virtual work [Nmm]
 angle between yield surface and displacement rate vector [rad]

 factor (0 <  < 0,5) [-]
 angle between relative displacement and vertical axes [rad]
1 ratio of the depth of the equivalent rectangular stress-block to the neutral axis depth [-]
γ0 overload factor in the code of New Zealand [-]
γL live load factor in the code of New Zealand [-]
 deflection in the middle of the span [mm]
 strain [-]
 angular rotation [rad]
 shear stress [N/mm2]
1 shear stress at with transverse reinforcement is necessary [N/mm2]
2 ultimate shear stress capacity [N/mm2]
 flexibility factor of laterally restrained slabs [-]
 strength reduction factor in the code of New Zealand [-]
D strength reduction factor in the code of New Zealand [-]
u height of the compression zone of the concrete [mm]
0 reinforcement ratio [-]
L change in length [mm]

 virtual rotation [rad]
viii
Chapter 1: Introduction
1. INTRODUCTION
Dutch bridges will not collapse yet1.1
De Volkskrant, 4 October 2007
Twelve steel bridges need urgent maintenance.

1.180 concrete bridges need further investigation.
Details will be presented next year.
Fatigued brides
AMSTERDAM Although metal fatigue and stresses in the concrete assaulted the Dutch bridges at a
large scale, there is no reason to panic. `There is no immediate security threat`, so the ministry of
´verkeer en waterstaat´ ensures. Experts approve this statement. ‘I do not believe that bridges in the
Netherlands will collapse any time soon` claims Leo Wagemans, academic civil engineering at
Delft University.
According to a report that was presented yesterday by ´rijkswaterstaat´, 25 of the 274 steel bridges
suffer from metal fatigue. In the case of twelve of these bridges, including the ´Brienenoordbrug´
and the ´Moerdijkbrug´, the problems are so serious that short-term adjustments have to be made.
From the total of 2.020 concrete bridges, 1.180 have to be examined more closely.
It is clear that the lifetime of bridges in the Netherlands is less then is assumed in the design.
Cracks in the steel appear sooner and lumps of concrete fall out.
The increased traffic intensity, environmental load and the heavier trucks are the cause of the
shortening in lifetime. In the sixties and seventies a lifetime of at least sixty years was assumed.
Now it comes true that after just 30 years restoration is required.
‘Not only has truck traffic doubled, furthermore there is almost no empty truck left on the Dutch
roads’, thus Dick Schaafsma of ´Rijkswaterstaat´.
According to his colleague Frans Bijlaard from Delft- academic steel structures – it is not a matter
of carelessly or ignorance. The good reputation of hydraulics in the Netherlands is not a point of
discussion. ‘The Dutch ability to build bridges has not declined. Many of the problems involve the
greater rolling resistance of trucks. Due to technological developments the loads that needed to be
carried by two tires can now be carried by just one tire. This concentrated load results in savings of
fuel, but also in larger damage of the pavement.’
Collapsing of bridges happens more then men would suspect; last week in Vietnam, in august in
Minneapolis. In the United States five bridges have collapsed since 2000, due to heavy rainfall and
a collision.
1
Bijlaard: ‘Disasters due to a collision can also happen in the Netherlands. But the maintenance
mode in the United States is drastically worse than in the Netherlands.’ Schaafsma: `Problems at
the bridge in Minneapolis were already known. But nothing was done about it. This does not hap-
pen in the Netherlands.’
In May wear was detected at the ´Hollandse Brug´ at the A6 near Muiden and the bridge is now
closed for all truck traffic. Meanwhile ´rijkswaterstaat´ has inspected 2.020 concrete bridges and
viaducts and 274 steel bridges in the road infrastructure, which were build before 1975. Of the en-
gineering structures 1.180 need further examination. The remaining lifetime of those bridges might
be shorter then the lifetime of the original design.
At 25 of the steel bridges the problems are more urgent. Since necessary reparations have been
made, no traffic restrictions are needed according to the ministry. ‘On the mid-long period this is
not enough. To guarantee the traffic flow, the bridges need to be reinforced or replaced in the next
5 years’, according the ministry.
In the summer of 2008, a detailed report with all the needed adjustments will be presented. Then the
total costs of the renovation project will also be known. The renovation of the ´Moerdijkbrug´ itself
will cost 38 million Euro.”
The above article shows that bridges and viaducts build before 1975 are a point of discussion in the
Netherlands. The question is whether these structures really do need repairs or have to be replaced.
Another option is to check whether the structure is actually stronger then the calculations show in
the originally design. Varies studies have been performed to study the effect of compressive mem-
brane forces in laterally restrained concrete slabs. The conclusions of those studies are in general
the same: compressive membrane action enhances the ultimate load of laterally restrained slabs.
However, none of these studies present a calculation model that is usable in practice. Because of
this the effect of compressive membrane forces is neglected in current calculations. This report de-
scribes the development of a finite element model, which takes into account these compressive
membrane forces.
Figure 1.1: Typical Dutch bridge build-up
A large part of the structures discussed in the above article are build up of inverted T-beams with a
compression layer of concrete on top (see figure 1.1). The starting point for this study will be this
type of structure.
2
Chapter 1: Introduction
The traffic becomes denser in a fast rate. To get an indication how much the traffic load has in-
creased since 1975 a comparison is made between the bridge load model used before 1975 (accord-
ing to the VOSB 1963) and the bridge load model that is used nowadays (according to the NEN-EN
1991-2).
From figure 1.2 it can be concluded that the total load has not increased very much, but the loads
get more concentrated. This means that the structural integrity as the structure as a whole will
probably be not an issue, but locale failure (for example in the decks) might become governing.
This local failure due to higher axle forces will also be taken into account in the finite element
model. Another difference between the codes is the calculation of cyclic loading. In the VOSB 1963
no attention is paid to this type of loading, while in the NEN-EN 1991-2 this method is extensively
described.
DAF stands for

Dynamic Amplication Factor side view
side view square 400 x 400 mm
Lane 1 Lane 1
q = 4 x DAF kN/m2 q = 1 9 kN/m2
F = 4 x 150 kN
Lane 2 Lane 2
2m
q = 4 x DAF kN/m2 q = 2 2,5 kN/m2

F = 12 x 50 x DAF kN F = 4 x 100 kN
Lane 3 Lane 3
q = 2 2,5 kN/m2
q = 4 x DAF kN/m2
F = 4 x 50 kN
1,2 m
Left over area Left over area
top view configuration 1 q = 4 x DAF kN/m2 top view configuration 1 q = 2 2,5 kN/m2
1 = 1,0 for less then 3 lanes
rectangle 320 x 250 mm 1 = 1,15 for 3 or more lanes
2 = 1,0 for less then 3 lanes
Lane 1 2 = 1,4 for 3 or more lanes
2m
q = 0,8 x 4 x DAF kN/m2

F = 12 x 0,8 x 50 x DAF kN F = 2 x 200 kN
0,5 m 0,5 m
Lane 2
0,5 m
q = 0,8 x 4 x DAF kN/m2 rectangle 350 x 600 mm

F = 12 x 0,8 x 50 x DAF kN
1m 4m
Lane 3
q = 0,8 x 4 x DAF kN/m2
Left over area

q = 0,8 x 4 x DAF kN/m2 top view configuration 2
top view configuration 2
for local failure mode
LOAD MODEL ACCORDING TO VOSB 1963 LOAD MODEL ACCORDING TO NEN-EN 1991-2
Figure 1.2: Traffic load of VOSB 1963 compared to the NEN-EN 1991-2 traffic load
1.1. The goal of this study
The goal of this study is to develop a numerical finite element model that is able to calculate the
ultimate bearing capacity of one-way continuous concrete bridge decks, taking into account com-
pressive membrane action. This model takes into account two modes of failure. The first is due to
global failure in bending, the second due to local punching failure of the deck by concentrated axle
loads. This model can be used to define the enhancement due to compressive membrane action in
existing and new one-way concrete bridge deck structures.
3
1.2. The structure of this study
This study will be build up of three parts:
Part I: The theory

This part involves
 the theory on compressive membrane action in both bending and punching failure
 the comparison between analytical and the above mentioned theoretical solutions
 the comparison between test results presented in various articles and the above mentioned theo-
retical solutions
Part II: The numerical development of the finite element model

This part involves
 the creation of a numerical finite element model

 the comparison between the analytical and finite element results
 the comparison with some test results presented in various articles and the finite element results
 if necessary adjustments are made to the model
Part III: A practical example

This part will involve
 the application of the in part II found model on a practical bridge deck

 an estimation of the enhancement factor for these kind of structures
The study will finish with conclusions and recommendations.
4
Part I: The theory of compressive membrane action
5
Part I: The theory of compression membrane action
2. THEORY OF COMPRESSIVE MEMBRANE ACTION
Compressive membrane action forms when two conditions are met. First the horizontal translation
has to be (partly) restrained. The greater the restraint, the greater the compressive membrane force
will be. Secondly the net tensile strain along a longitudinal fibre must be non-zero when there is no
horizontal restraint 2.1. Figure 2.1 shows a concrete one-way slab, which is in the cracking state.
Due to the cracking, the slab wants to elongate, but the rigid lateral restrained supports prevent this
from happening, so a compressive membrane force is introduced.
lateral
movement Load
prevented
compressive
membrane
force
Figure 2.1: Compressive membrane action in a latterly restrained slab
The enhancement of the collapse load by compressive membrane action can be clearly seen in a
load-deflection diagram, as shown in Figure 2.2 2.5.
compressive membrane
tensile membrane
applied loading
action
action
bending action
central deflection
Figure 2.2: Load-deflection graph for a structure with compressive membrane action
6
Chapter 2: Theory of compressive membrane action
2.1. Enhancement of the bending strength by compressive membrane forces
The calculation method used here is derived by Park 2.2. The theory has its starting point as shown
in Figure 2.3. In this figure t is the outward lateral movement.
1 4
2 3
.L .L
t L t
Figure 2.3: Starting point for Park’s compressive membrane theory
Assumptions in his model are
 the tension steel in the plastic hinges is yielding

 the concrete in the plastic hinges has reached its compressive strength
 the tensile strength of concrete is neglected
 the rotations and strains are small
 the top reinforcement is the same on both sides
 the bottom reinforcement is constant over the length of the strip
 the top and bottom reinforcement may differ
 the 3 parts, 1-2, 2-3 and 3-4 of the beam remain straight
 the axial strain,  (sum of the elastic, creep and shrinkage axial strain) has a constant value over
the length
 shear forces are neglected, since their net contribution in virtual work equations is zero
The shortening of the middle part due to the strain is L23   (1  2  ) L . So points 2 and 3 will
move 0,5 (1  2  ) L to the middle of the system. The distance from point 2 to the boundary now
becomes L  0,5 (1  2  ) L  t . The parts 1-2 and 3-4 will shorten to the length (1   ) L . These
values are shown in figure 2.3.
7
L + 0,5 (1-2 )L + t
(1- L
A
T'
d'
B
top steel Cs
1 c Cc
d h
C'c d
c' 2
C's
bottom steel
T
d'
Figure 2.4: Section 1-2 in the deformed state
The distance between the points A and B can be calculated with the geometry of the deformations
in two different ways, and so the next equation is derived:
( L  0,5 (1  2  ) L  t )
dist A B   (h  c' ) tan   (1   ) L  c tan  (1)
cos 
( L  0,5 (1  2  ) L  t ) sin 
 (1   ) L  h  c  c' 
cos  cos 
L  L cos   L cos   0,5 (1  2  ) L  t

 h  c  c'
sin 
1  cos 2 x 
sin 2 x   L  L cos   2 L sin 2
2 2

2 L sin 2  L cos   0,5 (1  2  ) L  t
h  c  c'  2
sin 
8
Since  and  are small some simplifications can be used in this equation:
 
sin   2 sin 
2 L
cos   1
  ( L) 2 0,5L2  ( L) 2 tL
h  c  c'     
2    
2
 L  2t 
c  c'  h      (2)
2 2  L
The second equation is formed using horizontal force equilibrium:
C ' s C ' c T '  C s  C c  T (3)

'
C ' c  0,85 f  1c '
c
C c  0,85 f c'  1c
f c'  concrete cylinder strength
N
 1  0,85 if f c'  30
mm 2
N
 1 reduces linear wit h 0,05 for every increase of f c' by 7 until a minium of 0,65
mm 2
0,85
0,65
30 58 f 'c
Figure 2.5: β1 as a function of f’c
9
With this information equation (3) forms into:
T 'T  C s'  C s
c ' c  (4)
0,85 1 f c'
Solving equation (2) and (4) simultaneously give the following solutions:
h  L2  2t  T ´T  C s´  C s
c´       (5)
2 4 4  L 1,7 f c´  1
h  L2  2t  T ´T  C s´  C s
c       (6)
2 4 4  L 1,7 f c´  1
In figure 2.6 it can be seen how the forces Cs, Cc and T work on the slab, and which stresses they
generate for a field moment. A similar figure can be made for the support moment.
0,5a Cs
c a
0,5h
Cc
d n u n u neutral
h axis
a= c
m u m u
T
Figure 2.6: Forces and moments acting in the middle of the span of the slab
With this figure, the quantities ns and ms can be calculated.
nu  C c  C s  T  0,85 f c' 1c  C s  T (7)

mu  0,85 f c'  1c(0,5h  0,51c )  C s (0,5h  d ' )  T (d  0,5h ) (8)
nu'  nu (equilibri um) (9)
mu'  0,85 f c'  1c' (0,5h  0,5 1c' )  C s' (0,5h  d ' )  T ' (d  0,5h ) (10)
The sum of the moments of the stress resultants at the yield section about an axis at mid depth at
one end in the strip is given by the formula:
mu'  mu  nu (11)
The shear forces are neglected in this equation, since their net contribution to a virtual work analy-
sis will be zero.
10
Substituting equations (5) to (10) in equation (11) gives the next formula:
h  1   L2  2t  
 2 1  2   4  1  3  4  1  1   L  
   
mu'  mu  nu  0,85 f c'  1h  2

 2  1  L2   1  2t   1  2 L4  2t  
 2     1          
 8h  2  4h  2  L  16h 2  L 
1 h   h 
3,4 f c'
T 'T  C s'  C s   C s'  C s   d '   (T 'T ) d   

2
2 2  2 2
When a virtual rotation  is given to a plastic hinge in the system, the virtual work done in that
hinge is:
m '
u  mu  nu 
When the virtual work done by the load on a structure is known, a load-deflection relationship can
be derived.
A few notes when using this formula:
 the model assumes that plastic hinges form immediately, this is of course not the case
 the first part of the curve will thus be inaccurate
 for large deflections, not the compressive but the tensile membrane action (catenary action) will
be governing (see Figure 2.2)
 this formula can thus not be used for large deflections
 this formula gives good results for moderate deflections, when plastic hinges start to form
To get a feeling for the influence of the different input parameters, graphs are made. On the left
side, the enhancement of the ultimate load provided by Park’s theory, and on the right side the pre-
dicted load by Park is shown. For each set one of the input values is varied, while the other values
are kept constant. For the basic input values, see the excel sheet in appendix A. Mind that these
graphs hold for bending action only.
variation in the length of the slab

4000
4,5
predicted load by Park in kN/m
4 3500
enhancement factor
3,5 3000
3
2500
2,5
2 2000
1,5 1500
1
1000
0,5
0 500
0 0,02 0,04 0,06 0,08 0,1 0,12 0,14

0
L = 600, L/ h = 4
central deflection devided by the thickness 0 1000 2000 3000 4000 5000 6000
L = 1200, L/h = 8
length in m m
L = 2400, L/ h = 16
11
variation in the thickness of the slab
4,5 4000

enhancement factor
3500
3,5
3 3000
2,5 2500
2
1,5 2000
1 1500
0,5
0 1000
0 0,02 0,04 0,06 500
h = 75, L/ h = 16
0
central deflection devided by the thickness h = 150, L/ h = 8 0 50 100 150 200 250 300 350 400
h = 250, L/ h = 4,8
thickness in m m
h = 350, L/ h = 3,4
variation in the reinforcem ent percentage
4000
4,5
3500
4
enhancement factor
3,5 3000
3
2500
2,5
2 2000
1,5
1500
1
0,5 1000
0
500
0 0,02 0,04 0,06
w0 = 0,2% 0
0 0,5 1 1,5 2
central deflection devided by the thickness w0 = 0,4%
w0 = 0,8% reinforcem ent percentage
w0 = 1,6%
variation in the concrete strength
4,5
4000
enhancement factor
4
3,5 3500
3
3000
2,5
2 2500
1,5
2000
1
0,5 1500
0
0 0,02 0,04 0,06 0,08 1000
500
fcu = 25 N/ mm2
central deflection devided by the thickness 0
fcu = 35 N/ mm2
20 25 30 35 40 45 50 55 60
fcu = 45 N/ mm2
fcu = 55 N/ mm2 fcu in N/m m 2
12
variation in the steel strength

4000
4,5

enhancement factor
3500
4
3,5 3000
3
2,5 2500
2
2000
1,5
1 1500
0,5
0 1000
0 0,02 0,04 0,06 0,08 0,1 0,12 0,14 500
fs = 235 N/ mm2
0
fs = 335 N/ mm2 200 250 300 350 400 450 500 550 600
central deflection devided by the thickness fs = 435 N/ mm2
fs in N/m m 2
fs = 535 N/ mm2
Figure 2.7: Enhancement factors (left) and absolute values (right) for Park’s theory
From the graphs, the following conclusions can be made:
 the length of the slab is only of influence for short spans, the influence on the enhancement fac-
tor is rather small
 the thickness of the slabs has great influence on both the enhancement factor and the ultimate
load
 the higher the reinforcement ratio, the lower the enhancement factor
 the higher the yield strength of the steel, the lower the enhancement factor
 for higher concrete strengths, the enhancement factor and the ultimate load will increase both
2.2. Enhancement of the punching shear strength by compressive membrane action
The enhancement of punching shear failure is discussed in an article published by the American
Concrete Institute 2.3 and the international journal of mechanical sciences2.4. The theory used in this
study will follow this method. Assumptions made in this model are:
 the failure mechanism consists of a solid cone-like plug

 the compressive membrane force has a constant value
 the behaviour is rigid plastic
 the energy in hoop expansion outside the plug is neglected
F 0,5d0
r(x) nr nr
na na
d1
2a
Figure 2.8: Punching shear failure model
13
The authors did some tests, but instead of making horizontal restrained clamped edges, they used
hoop reinforcement to create a similar effect.
1200
support
hoop reinforcement to 1200

prevent horizontal movement
and rotations
loaded area
150
150
1200
150
150
top view cross-section
Figure 2.9: Schematic overview of the test set-up
An upper-bound solution is given by the virtual work theory:
Wextern  F cos 
h
Wintern   D A 2 1  r ' 2 dx  N rs sin 
0
h
N rs  f cc  n r 2rdx
0
w0  r
n r  na  1  
2h  a 
   w0  1  1  w 
n a   ke  h    n 0     0 
 2 2 2  4h 
 
 wi 
n 1  w   
k   0    i  e  h 
 2 4 4 4h 
N
n0  0
hf cc
Nrs is the sum of the radial compressive membrane forces working on the failure surface of the
cone. The value for na is derived by using the flow theory.
14
N 0  0,5hf c  Ash f s
af cc

2hS
1 aR a
 
S 0,8 Ec Ac  E s As 0,5E c ( x  h )
a  0,5L
R  0,5L
fs
x  1,76 w0
f cu
wi  0,03h (deflection at which membrane action starts, emperical determined )
w0  0,5h (critical deflection, empirical determined)
DA is the internal energy dissipation per unit area in the deforming zone and is given by the follow-
ing formula:
2
 ck 

D A  f ct  1  cot 2   sin 
 4 
f cc
ck  1  1
f ct
Substituting all the above equations in the virtual work equation gives the following solution:
h 2 h
 ck 1  r ' tan  2   w0  r 
P  2f ct  r r ' tan   dx  2f cc  na  1   rdx tan 
0 
4 r ' tan   0 
2h  a 
Minimising the first integral gives an equation that does not contain x and Euler’s equation has the
first integral. With two boundary conditions this integral can be solved.
15
2 2
 c 1  r ' tan   
F ( r, r ' )  r  r ' tan   k 
 4 r '  tan  
F d  F 
  0
r dx  r ' 
F
F  r' C
r '
d d
r ( 0)  0 ; r ( h )  1
2 2
tan 
r  Ae Bx 
B
d0 tan 
 A
2 B
d 1  d 0 tan   Bh tan 
  e 
2  2 B  B
When  is given, the constants A and B can be determined and the failure load P becomes:
 A2
P  2f ct  e 2 Bh  1  e  1 tan   h tan 2  
2 A Bh
 2 B B 
2
ck  h 2 A Bh A2 2 Bh 
 f ct  
2 B B
e  1 tan  
2
e  1 tan 2    N rs tan 

 w  A h 
N rs  2f cc  na  0   e 2 Bh  1  tan  
 2h   B B 
w0  A2 2 Bh 2 A Bh h 
 2 B e  1  B 2 e  1 tan   B 2 tan  
2
 f cc
ah  
P
For a minimum, the derivative must be zero

2
P c  2 A Bh   2 A Bh 2h 

 f ct k 
2 B
e  1  A2 e 2 Bh  1 tan    f ct  e  1  tan  
 B B 
 w h w 2A 2h 
 2f cc  na  0  tan   f cc 0  2 e Bh  1  2 tan   tan   N rs  0
 2h  B ah  B B 
To calculate the ultimate load for punching shear failure, the following method can be used:
16
d1
d 
 calculate d1 for a simply supported plate with  1   e ck h
 d0 
 choose d1 as a smaller value as just calculated
 assume a value for 
P
 calculate

P
 adjust  until 0

 calculate P
 reduce d1 and repeat the above steps till a minimum is found for P
f cc
According to the article by Salim and Sebastian, realistic values are gained for  400 . This
f ct
factor is used in the calculation model in the maple sheet, see appendix B. Mind that this factor
should not be used to determine ck, for which the standard value for fct should be used.
This analytical model is presented in a maple sheet, see appendix B.
To get a better understanding of the influence of the different input values, graphs are made. The
basic values are shown in the maple sheet in appendix B, and for each graph one of the input values
is varied. Mind that these graphs only hold for failure in pure punch.
5,00 800
4,50 700
predicted failure load in kN
4,00
600
enhancement factor
3,50
3,00 500
2,50 400
2,00 300
1,50
200
1,00
100
0,50
0,00 0
0 1000 2000 3000 4000 0 1000 2000 3000 4000
le ngth of the s lab in m m length of the slab in m m
5 800
4,5 700
4
600
enhancement factor
3,5
3 500
2,5 400
2 300
1,5
200
1
100
0,5
0 0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
height of the slab in m m he ight of the slab in m m
17
5 800
4,5 700

4
600
enhancement factor
3,5
500
3
2,5 400
2 300
1,5
200
1
100
0,5
0 0
0 0,2 0,4 0,6 0,8 1 1,2 1,4 0 0,5 1 1,5
reinforce m ent ratio in % re inforcem e ntratio in %
5 800
4,5 700
4
predicted failure load

600
enhancement factor
3,5
500
3
2,5 400
2 300
1,5
200
1
100
0,5
0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
fcu in N/m m 2 fcu in N/m m 2
5 800
4,5 700
4
600
enhancement factor
3,5
500
3
2,5 400
2 300
1,5
200
1
100
0,5
0 0
0 100 200 300 400 500 600 0 100 200 300 400 500 600
d0 in m m d0 in m m
5 800
4,5 700
4
predicted failure load
600
enhancement factor
3,5
500
3
2,5 400
2 300
1,5
200
1
100
0,5
0 0
0 100 200 300 400 500 600 700 200 300 400 500 600 700
fy in N/m m 2 fy in N/m m 2
Figure 2.10 Enhancement factors (left) and absolute values (right) for the Punching failure theory
18
Looking at he graphs, the following can be concluded:
 the length of the slab, the reinforcement ratio and the reinforcement yield strength do not have
much influence on both the enhancement factor and the total ultimate load
 the height has the largest influence on both the enhancement factor and the ultimate load
 the higher the slab, the smaller the enhancement factor becomes
 the concrete strength and the length of the loaded area have a positive influence on both the en-
hancement factor and the ultimate load
2.3. Comparison between bending and punching failure
To get an understanding if failure in bending or failure in punching occurs, the found results will be
compared to each other. The enhancement factors of the previous sections cannot be used. To com-
pare the results, two new enhancement factors will be used.
For bending:
Fpunch,code qPark
if qanalytic,bending  
L qanalytic,bending
Fpunch,code q park .L
if qanalytic,bending  
L F punch,code
For punch:
Fpunch,code Fpunch, predicted
if qanalytic,bending  
L qanalytic,bending .L
Fpunch,code Fpunch, predicted
if qanalytic,bending  
L F punch,code
In other words, the predicted values by the theory presented in chapter 2 are divided by the lowest
value of the analytic bending and punch failure load.
The starting point or calculations is shown below. For each graph, one of the values is varied.
L 1500 mm
b 1000 mm
h 150 mm
d0 300 mm
fck 35 N/mm2
fs 435 N/mm2
wo 0,37 %
19
2500 12
predicted ultimate load
2000 10
enhancement factor
8
1500
6
1000
4
500 2
0 0
0 1000 2000 3000 4000 0 1000 2000 3000 4000
le ngth of the slab in m m bending failure (Park) lenght of the slab in m m bending failure (Park)
punching failure punching failure
2500 12
2000 10
enhancement factor
8
1500
6
1000
4
500
2
0 0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
height of the slab in mm bending failure (Park) height of the s lab in m m bending failure (Park)
2500 12
10
2000
enhancement factor
8
1500
6
1000
4
500
2
0
0
0 0,2 0,4 0,6 0,8 1 1,2 1,4
0 0,2 0,4 0,6 0,8 1 1,2 1,4
reinforce m ent ratio in % bending failure (Park) bending failure (Park)
reinforcem e nt ratio in %
2500 12
10
2000
enhancement factor
8
1500
6
1000
4
500
2
0
0
0 10 20 30 40 50 60
0 10 20 30 40 50 60
fcu in N/m m 2 bending failure (Park) bending failure (Park)
fcu in N/m m 2
20
2500 12
2000 10
enhancement factor
8
1500
6
1000
4
500
2
0
0 100 200 300 400 500 600 0
0 100 200 300 400 500 600
d0 in m m
bending failure (Park) d0 in m m bending failure (Park)
2500 12
10
2000
enhancement factor
8
1500
6
1000
4
500 2
0 0
0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700
fy in N/m m 2 beding failure (Park) fs in N/m m 2 bending failure (Park)

Figure 2.11: Enhancement factors (left) and absolute values (right) for both Park’s and the punching failure theory
Conclusions that can be drawn from these graphs are:
 for a low slenderness, punch is governing

 if punch is governing, the flexural reinforcement is not a factor of influence
 the steel strength is not much of an influence factor
21
Chapter 3: Analytic solutions
3. ANALYTIC SOLUTIONS
A number of trivial structural cases are discussed and compared to the calculations methods used in
standard calculations nowadays. Before the comparison can be done these calculation methods will
be briefly discussed.
3.1. Standard calculations methods used
Both the ultimate load in bending and in punch will be described.
3.1.1. Ultimate bearing capacity in bending
The method to determine the ultimate load will be the virtual work theory 3.1. This theory makes use
of the fact that the work done internal and external needs to be in equilibrium. This means that the
work done by the displacements of the loads (external work), needs to be equal to the work done by
the rotation of the plastic hinges (internal work):
Wint ernal  Wexternal
How this works for beam-like elements will be illustrated by a simple example, see figure 3.1.
F Fu
Mu
Mu Mu
Figure 3.1Simple model with its failure mode
At least 2 plastic hinges need to be introduced to create a failure mechanism. It is assumed that the
deformations remain small, so the displacement  can be written as the angle times the length of the
rotated part. The virtual work equation now becomes:
Wexternal  Fu  0,5LFu

Wint ernal  M u  M u  M u  3M u
Wexternal  Wint ernal  0,5LFu  3M u
6M u
Fu 
L
If the ultimate moment of the cross-section is known, than the ultimate load Fu can be easily calcu-
lated. For reinforced concrete, the ultimate moment can be calculated with the help of figure 3.23.2.
23
N' b
Ns
cross section strain stress

Figure 3.2: Stress and strain distribution in a concrete cross-section, to determine the moment capacity
The ultimate moment for the cross section can now be calculated:
N s  N ' b (equlibrium)
N s  As f s
3
N 'b  u f 'c b
4
4A f
u  s s
3 f 'c b
z  d  0,39  u
 4A f 
M u  N s z  As f s  d  0,39 s s 
 3 f 'c b 
With Mu known, the ultimate load of system can be calculated.
3.1.2. Ultimate concentrated load
For the maximum concentrated load, the punching model described by Sagel and van Dongen 3.2
will be used.
Fu
a+d
d a
Figure 3.3: Punching shear failure for a concentrated load
24
Chapter 3: Analytic solutions
The height of the load that induces this failure mechanism can be calculated as follows:
Fa   1 pd
p   (a  d )
4a1
a  4a1  a 

 1  0,8 f c k d 3  0  0,8 f c
1   2
k d  1,5  0,6d  1
 0   0 x 0 y  2%
In the formulae for kd, d is in meters.
a1 loaded area
a1
Figure 3.4: Difference between a and a1
When there is a normal force acting on the structure, 1 may even be increased by n:
 n  0,15. bmd
 1,increased   1   n
25
Chapter 4: Comparisons between the analytical solutions and test results
4. COMPARISONS BETWEEN THE ANALYTICAL SOLUTIONS AND TEST RESULTS
To check whether the analytical models give an accurate solution, they will be compared to differ-
ent test results, which can be found in various articles.
4.1. The bending model
To calculate the solution by Park, a excel sheet is made (see Appendix A). In this sheet, the theory
is used as described in chapter 2.1. To calculate Cs and C’s the strain in the compression reinforce-
ment has to be known. The strain however depends on the concrete compression strain and the dis-
tance to the neutral axis. These values are variable as can be seen in the figure below.
tension zone
d
d
c = 0,25h ~ 0,5h
compression zone
s
c 0,175 ~ 0,35%
Figure 4.1: The strain distribution for different neutral axis depths
To get an estimation of the steel strain in the compression steel, the average of the above shown
values will be used. The value Cs thus becomes
c s

0,375h d  0,625h
 s  0,002625
d  0,625h 
0,375h
AE A E (d  0,625h)
C s  s s s  0,002625. s c
b 0,375bh
27
Park presents the following graph in his work 2.2, which shows the difference between an experi-
mental result and his analytical solution.
q over qanalytical
Park's theory
expirimental result
central deflection over the heigth
Figure 4.2: Difference between Park’s theory and a test specimen
As can be seen, Park’s theory gives a somewhat lower ultimate load then the actual specimen. This
graph is roughly the same for all test results. It is thus expected that the value of Park’s model will
lie in between the analytical virtual work solution and the ultimate strength of the tested specimens.
4.1.1. Test results from L. K. Guice and E. J. Rhomberg
The article 4.1 presents test results for a 1-way clamped slab. A sketch with the dimensions of the
experiment is showed below.
Figure 4.3: Dimensions of the test specimens
28
A table, which compares the virtual work results, the experimental results and results from the
bending theory presented in chapter 2.1, is presented in Table 4.1.
Slab fcu fy L b h d* w0 Pa Pe Pp Pe/Pa Pe/Pp

N/mm2 N/mm2 mm mm mm mm % kN/m kN/m kN/m - -
tests in bending failure with uniform loading
1 30,4 344,8 610 610 58,7 45 0,52 115,0 328,1 242,3 2,85 1,35
2 29,4 344,8 610 610 58,7 45 0,52 114,7 218,7 236,3 1,91 0,93
3 30,6 403,2 610 610 58,7 45 0,74 185,6 302,8 274,1 1,63 1,10
4 29,4 403,2 610 610 58,7 45 0,72 178,3 298,6 264,7 1,67 1,13
4A 28,7 403,2 610 610 58,7 45 0,72 177,8 290,2 261,2 1,63 1,11
4B 29,0 403,2 610 610 58,7 45 0,72 178,0 323,9 262,7 1,82 1,23
5 30,7 403,2 610 610 58,7 45 1,06 250,2 412,2 315,5 1,65 1,31
6 29,5 403,2 610 610 58,7 45 1,06 248,3 382,7 310,4 1,54 1,23
7 34,6 464,2 610 610 41,2 30 0,58 79,0 134,6 117,9 1,70 1,14
8 34,3 464,2 610 610 41,2 30 0,58 78,9 96,7 117,3 1,23 0,82
9 34,6 403,2 610 610 41,2 30 1,14 125,5 168,3 137,9 1,34 1,22
9A 34,5 403,2 610 610 41,2 30 1,14 125,4 172,4 137,8 1,38 1,25
10 34,3 403,2 610 610 41,2 30 1,14 125,3 - 137,5 - -
10A 34,2 403,2 610 610 41,2 30 1,14 125,2 - 137,3 - -
11 34,6 403,2 610 610 41,2 30 1,47 152,9 193,5 152,9 1,27 1,27
12 34,3 403,2 610 610 41,2 30 1,47 152,5 92,5 153,5 0,61 0,60
* this value is not mentioned in the article and is an assumption Mean value 1,59 1,12
Standard deviation 0,49 0,21
Table 4.1: Test values with the analytical (Pa), experimental (Pe) and predicted (Pp) solutions
As was to be expected, the value of the bending theory gives a better estimation than the analytical
result. Furthermore it can be seen that Figure 4.2 also holds for these experiments, the estimation of
Park is in most cases lower then the actual collapse load..
4.2. The punching model
4.2.1. Tests by J.S. Kuang and T. Morley
The article 4.2 presents the ultimate load for experiments, which failed in punch. The experiments
consisted of 2-way slabs with different edge beam widths. Dimensions as shown in Figure 4.4 and
Figure 4.5.
Figure 4.4: Overview of the experimental set-up
29
300 300
40 - 60
200
120
70 - 140 -280 1200
Figure 4.5: Dimensions of the test specimen
The results are presented Table 4.2.

Slab fcu fy L h d edgebeam w0 c Pa Pe Pp Pe/Pa Pe/Pp
N/mm2 N/mm2 mm mm mm mm % mm kN kN kN - -
S1-C03 48,7 400 1200 60 49 280 0,3 120 36 101 104 2,81 0,97
S1-C10 33,8 400 1200 60 49 280 1,0 120 29 118 64 4,07 1,85
S1-C16 41,2 400 1200 60 49 280 1,6 120 32 149 81 4,66 1,84
S2-C03 48,1 400 1200 40 31 280 0,3 120 20 49 63 2,45 0,78
S2-C10 45,8 400 1200 40 31 280 1,0 120 20 70 59 3,50 1,19
S2-C16 42,6 400 1200 40 31 280 1,6 120 19 68 53 3,58 1,29
S1-B10 45,9 400 1200 60 49 140 1,0 120 35 116 95 3,31 1,22
S1-B03 50,8 400 1200 40 31 140 0,3 120 21 42 67 2,00 0,63
S2-B10 59,5 400 1200 40 31 140 1,0 120 24 69 80 2,88 0,86
S1-A10 46,5 400 1200 60 49 70 1,0 120 36 99 96 2,75 1,03
S2-A03 47,8 400 1200 40 31 70 0,3 120 20 43 62 2,15 0,69
S2-A10 60,3 400 1200 40 31 70 1,0 120 24 63 81 2,63 0,78
Mean value 3,06 1,09
The factor Pe/Pp (Pp stands for predicted ultimate load and NOT Park’s ultimate load, this value is
calculated with the punch theory described in chapter 2.2) becomes smaller for smaller edge beam
widths, which is to be expected. The presented bending theory does not include a rate of inclination,
so for different edge beam widths, the predicted value will be the same. In reality the rate of inclina-
tion decreases for smaller edge beams, which will reduce the compressive membrane forces and
thus the ultimate collapse load.
4.2.2. Tests by W. Salim and W.M. Sebastian
The authors of the theoretical punch model 2.3 did some experiments. See Figure 2.9 for the test set-
up.
Slab fcu fy L h d w0 c Pa Pe Pp Pe/Pa Pe/Pp

N/mm2 N/mm2 mm mm mm % mm kN kN kN - -
S1 63 500 1200 150 113 1,06 150 156,9 369,4 466,7 2,35 0,79
S2 52 500 1200 150 113 1,06 150 136,3 290,6 367,6 2,13 0,79
S3 56 500 1200 150 113 1,06 150 143,8 402,2 403,4 2,80 1,00
S4 53 500 1200 150 113 1,06 150 138,2 394,1 376,6 2,85 1,05
30
As we put the data in a graph, we can see some scatter in the test results, since only the concrete
strength is varied.
520
470
failure load in kN
420
370
320
270
50 52 54 56 58 60 62 64
predicted load
fcu in N/m m 2
test result
Figure 4.6: Predicted result versus test results
4.2.3. Tests by Holowka, Dorton and Csagoly
In the article which presents the theoretical punch failure solution 2.2, some test results from other
parties where included. No information is given on how the test where carried out, but it still some
good comparison data. One of these test results is shown in Table 4.4.
Slab fcu fy h* d w0 c Pa Pe Pp Pe/Pa Pe/Pp

N/mm2 N/mm2 mm mm % mm kN kN kN - -
A1-A3 27,4 310 48,1 38,1 0,2 127 19,1 42,4 42,1 2,22 1,01
B1-B2 27,7 310 41,8 31,8 0,2 127 15,5 36,8 35,8 2,37 1,03
C1-C3 28,4 310 35,4 25,4 0,2 127 11,8 23,1 30,2 1,96 0,76
D1-D3 25,9 310 48,1 38,1 0,3 127 18,5 49,1 39,3 2,65 1,25
E1-E3 26,9 310 41,8 31,8 0,3 127 15,2 34,9 35,1 2,30 0,99
F1-F3 27,0 310 35,4 25,4 0,3 127 11,7 23,8 29,4 2,03 0,81
G1 28,1 310 48,1 38,1 0,8 127 19,4 44,6 43,1 2,30 1,03
H1 28,2 310 41,8 31,8 0,8 127 15,6 37,9 36,7 2,43 1,03
I1-I3 28,1 310 48,1 38,1 1,0 127 19,4 47,3 42,6 2,44 1,11
J1-J2 27,8 310 41,8 31,8 1,0 127 15,5 37,2 35,8 2,40 1,04
K1-K2 27,9 310 35,4 25,4 1,0 127 11,9 25,2 32,0 2,12 0,79
* this value is not mentioned in the article and is an assumption Mean value 2,29 0,99
The span of the specimen is not given, but it is earlier stated that the influence of the length is not of
much influence for a small d0/L ratio, as is already mentioned in chapter 2.2. It can be seen that the
predicted value gives a good indication. Only a few of the tested specimens have a somewhat larger
difference. It can be that the height of the slab is not assumed correctly, or the tests may not have
been carried out properly.
31
4.2.4. Full scale tests by S.E. Taylor, B. Rankin, D.J. Cleland and J. Kirkpatrick
Experiments done on a full-scale structure is described in the article 4.3. The tests where done on
several spans with loads up to three times the maximum wheel load of the British standards. The
crack widths and the deflection at the midsection of the span where measured.
160
300
1240 1740
1500 2000
Figure 4.7: Overview of the test setup
Since they did not loaded the bridges till they collapsed, for now it can be only check if the analyti-
cal ultimate load is higher than the load level during the testing. After the finite element calcula-
tions, it can be checked whether the deflection and the crack widths calculated at the loading levels
is in accordance with reality.
slab fcu fs L h d w0 anaytical collapse maximum test predicted ultimate

N/mm2 N/mm2 mm mm mm % load load load
A1 77,8 501 1740 160 82 0,5 + fibers 128,3 333 1240,1
A2 80,6 501 1240 160 74 0,5 + fibers 178,3 428 1279,6
B1 76,5 501 1740 160 75 0,25 + fibers 66,5 344 1280,5
B2 82,3 501 1240 160 75 0,25 + fibers 92,3 428 1314,9
C1 81,2 501 1740 160 75 0,25 66,6 333 1291,3
C2 78,2 501 1240 160 75 0,25 92,2 428 1239,3
D1 74,6 501 1740 160 75 0,5 127,9 368 1166,2
D2 74,6 501 1240 160 75 0,5 177,3 428 1169,1
E1 67,8 501 1740 160 105 0,6 202,1 392 1040,7
E2 67,8 501 1240 160 105 0,6 280,0 428 1043,1
F1 60,0 501 1740 160 103 0,6 199,5 371 899,8
F2 61,0 501 1240 160 103 0,6 275,2 428 919,8
Table 4.5: Test values and the predicted ultimate load
32
Part II: The finite element modelling
33
Chapter 5: Finite element models
5. FINITE ELEMENT MODELS
The theory describes a one-way laterally restrained slab. Cracking of the concrete is the main rea-
son compressive membrane forces are generated. The finite element package DIANA is capable to
calculate concrete structures, including the cracking of the concrete. For bending a 2D model will
be sufficient, but if punch is to be included, a 3D model must be used.
The loading steps on the structure can be applied in two different ways. There can be a given dis-
placement and displacement steps, at which DIANA calculates the load, or there can be applied a
load with a loading step and then DIANA calculates the displacement. Since Park’s theory de-
scribes a plate with a distributed load, the input will be a load with a loading step.
Then there is the issue of which elements to use for the calculations. This will be discussed first.
5.1. The geometry of the model
Since the model must describe a reinforced concrete one-way slab with a unit width, the following
models can be used:
 2D beam model bending

 2D plane stress model bending
 3D curved shell model bending
 axi-symmetric model bending and punch
 3D solids model bending and punch
To make a choice of the above models, they will be used to calculate the following simply sup-
ported beam as shown below.
1200
f cu = 35 N/mm2
f s = 435 N/mm2
115
150
11  8
1000
Figure 5.1: Simply supported beam used to determine a suitable finite element model
The analytical solution is

8M
qu  2 u
L
 4.552,9 
M u  552,9.435115  0,39   26,23kNm
 3.21.1000 
kN
qu  8.26,23.1,2 2  145,7
m
35
To make a choice which model to use the following factors are considered
 the deviation of the analytical solution presented in chapter 3

 the complexity of the input of the model
 the calculation time
 the output
For now the total strain rotating crack model, brittle in tension and ideal in compression will be
used. The analytical solution is based on brittle cracking without tension softening, so this material
model should give results close this solution and will be used for these calculations.
5.1.1. 2D beam model
The input model is presented in Figure 5.2.

shear reinforcement (if needed)
distributed load
flexural reinforcement
L L7BEN beam elements
0,5h-d'
Figure 5.2: Schematic overview and iDIANA input for the finite element model
The manual of DIANA gives the following overview for beam elements. Only the class-II and
class-III elements can be reinforced with embedded reinforcement bars.
Class Class-I Class-II Class-III

Theory Bernoulli Bernoulli Mindlin-Reissner
Type L6BEN L12BE L7BEN L13BE CL9BE CL12B CL15B CL18B CL24B CL30B
Dimension 2D 3D 2D 3D 2D 2D 2D 3D 3D 3D
Figure 5.3: Overview of the type of beam elements that can be used by DIANA
The L7BEN elements will be used for this calculation. The number of integration points over the
height will be increased, so that the cracking will be taken into account more accurately. The differ-
ence is shown in the load-displacement graph below. The calculation stops before reaching the
horizontal part when using only 2 integration points over the height of the beam.
36
160
140
120
100
load in kN/m
80
60
40
20
0
0 1 2 3 4 5
central deflection in mm
2 integration points (def ault)
21 integration points
Figure 5.4: Difference between2 and 21 integration points over the height of a beam element
Figure 5.2 load-displacement graph for the middle node calculated by DIANA
Advantages
 easy and fast to build model

 little calculation time
 analytical and numerical solution are almost equal
 moment and shear forces as output
Disadvantages
 graphical output is minimal, since it is only a line model

 loads act over the whole width of the beam
37
5.1.2. 2D plane stress model
The input is presented below. It is common to make use of symmetry whenever possible, to reduce
the calculation time.
symmetry axes
0,5L
distributed load
shear reinforcment (if needed)
CQ16M plane stress

elements
flexural reinforment
Plane stress elements are characterised by the fact that the stress components perpendicular to the
face are zero, σzz = 0. DIANA can use the following regular plane stress elements. For this calcula-
tion, the CQ16M element will be used.
T6MEM CQ8MEM CT12MEM CQ16M CQ18M
Figure 5.6: Overview of the type of 2D plane stress elements that can be used by DIANA
38
Figure 5.7: Output of iDIANA, stresses in the concrete and the reinforcement, the crack pattern and the load-displacement graph
Advantages
 easy and fast to build model

 little calculation time
 analytical and numerical solution are almost equal
 al the wanted output can be graphically presented
Disadvantages
 rotation is not a degree of freedom

 no moment and shear forces as output
 loads act over the whole width of the model
39
5.1.3. 3D curved shell model
For this model two symmetry axes will be used.
sy
m
m
et
ry
ax
e s
Z
0,
5L
X

distibuted load
CQ40S curved shell elements
x es 0,5b
ry a
m et
0,5h-d'
sym
Figure 5.8: Schematic overview of the 3d curved shell model
Figure 5.9: iDIANA input of the finite element model
The element CQ40S will be used for this calculation. The number of integration points over the
height will be increased, so that the cracking will be taken into account more accurately, in the same
way as for the 2D beam model, see chapter 5.1.1.
T15SH Q20SH CT30S CQ40S CT45S CQ60S
Figure 5.10: Overview of the type of 3D curved shell elements that can be used by DIANA
40
Figure 5.11: Some results, the load-displacement graph and cracking in the top, middle and bottom surface
Advantages
 analytical and numerical ultimate load are almost equal

 al the output can be graphically presented
 load can be distributed over an given area
Disadvantages
 rather long calculation time
41
5.1.4. 3D solid model
For the 3D solid model two symmetry axes will be used.
sy
m
m
et
ry
ax
is
0,
5L
Z

CHX60 solid elements
distibuted load
xis
etry a
symm
d
h
0,5b
42
TE12L TP18L HX24L CTE30 CTP45
CHX60 CTE48 CTP72 CHX96
Figure 5.13: Overview of the type of 3D solid elements that can be used by DIANA
The element CHX60 is used for this calculation.
Figure 5.14: Output of iDIANA , the load-displacement graph and σxx in a x and y cross-section
Advantages
 the largest collection of output data

 load can be distributed over an given area
Disadvantages
 long calculation time

 most complex to build of all the models
 output can be difficult to understand
43
5.1.5. Axi-symmetric model
The axi-symmetric model will only be used for concentrated loads and will be further discussed in
chapter 7.2.
5.1.6. Conclusion
The load-displacement graphs of all the models are compared below. All the models give a very
accurate outcome compared to the analytically found load.
160
140
120
100
load in kN/m
80
60
40
20
2D beam elements
0 2D plane stress elements

0 1 2 3 4 5 3D curved shell elements
central deflection in mm 3D soldid elements

Analytical collapsde load
Figure 5.15: Comparison of the load-displacement graphs of the four models
It can be seen that all the models give a good approximation of the ultimate analytical load. The 2D
beam model and the 3D shell model give almost the some solution. The same holds for the 2D
plane stress and 3D solid model. The latter react somewhat less stiff at large loads.
An overview of the different models is given in Table 5.1.
input graphical output realistic model punch behaviour calculation time

2D beam model ++ -- + irrelevant ++
2D plane stress model + + + irrelevant +
3D curved shell model o - + -- o
3D solids model - ++ ++ o --
axi-symmetric model + + -* ++ +
* : rectangular slabs cannot be modelled with this model
Table 5.1: overview of the different finite element models
44
5.2. Material properties
The material properties input in DIANA is numerous. Different models can be used for cracking,
the tensile and compressive behaviour of concrete and the behaviour of the reinforcement steel. All
of these will be discussed shortly in the next paragraphs. The most realistic material models will be
used, since the model will be compared to experimental results.
5.2.1. Cracking
There are two cracking models that can be used, smeared cracking or total strain cracking. The
smeared cracking model is depended on the principal stresses. This can be taken in to account for a
constant or linear function.
Figure 5.16: Two ways of smeared cracking
The total strain crack model describes the tensile and compressive behaviour of a material with one
stress-strain relationship. This makes the model very well suited for Serviceability Limit State
(SLS) and Ultimate Limit State (ULS) analyses which are predominantly governed by cracking or
crushing of the material. Within this model, there can be chosen for rotating or fixed cracking. The
difference between the two is that for fixed cracking the crack lies in the same direction for all the
load steps, while by rotating cracking, the direction of the crack is calculated separately for each
load step.
The total strain rotating crack model will be used for the calculations.
45
5.2.2. Concrete in tension
An overview of the different models DIANA can use for tension in the concrete is presented below.
Figure 5.17: Different models for concrete behaviour after cracking
The options (d) till (g) show the possibilities to take into account tension softening.
The Dutch code is based on models with a brittle cracking model, but including tension softening
may give results that lie closer to the real collapse load. So both options (c), brittle and (f), the
Hordijk model will be used for the calculations.
5.2.3. Concrete in compression
An overview of the different models DIANA can use for compression in the concrete is presented
below.
Figure 5.18: Different models for concrete behaviour after the yield strength is reached
Option (b), the ideal model will be used for the calculations.
46
5.2.4. Reinforcement steel
For the reinforcement there are three options, ideal plastic, a work-hardening diagram or a strain-
hardening diagram. The ideal plastic model is used.
5.2.5. Conclusion
The material model that is used looks like this:
 total strain rotating crack model

 both brittle and Hordijk tension softening in tension
 ideal plastic model in compression
 ideal plastic model for the reinforcement
47
Chapter 6: Finite element model for bending action
6. FINITE ELEMENT MODEL FOR BENDING ACTION
The 2D plane stress model will be used since this is an easy to build model with fast calculation
time.
6.1. Total horizontal restrained clamped model
The same model as described in chapter 5.1 will be used, but now the edges will be totally clamped
and horizontally restrained. The analytical and Park’s solution are given below.
q
L/2 L/2
Figure 6.1: Schematic overview of the structure
kN
qanalytical  291,4
m
kN
qPark  706,7
m
The schematic input in DIANA looks as follows

symmetry axis
0,5L
distributed load
shear reinforcment (if needed)
CQ16M plane stress

elements
flexural reinforment
Figure 6.2: Schematic input of the finite element model
The difference between the earlier shown input is that the edge is now horizontally restrained and
extra top reinforcement is added. Rotation is not a degree of freedom for these elements, but since
the whole edge is restrained horizontally and vertically, it can not rotate.
The ultimate load calculated by DIANA is even higher than Park’s prediction. This seems to be a
good result since Park’s values are in most cases somewhat lower than the experimental data (see
chapter 4.1).
49
Since the code does not take into account tension softening, the solution with a brittle cracking
model is compared to the tensile softening solution in the graph below.
1200
1000
800
load in kN/m
600
400
200
0
0 1 2 3 4 5 6 7 8
Brittle
Hordyk
Park's ultimate load
Analytical ultimate
load
Figure 6.3: Difference between a calculation with brittle cracking and tension softening
Figure 6.4: σxx over load steps 23 to 28, from 924 to 997 kN/m
In Figure 6.4, the shifting of the neutral axis can be seen.
Figure 6.5: Stresses in the steel, the steel is loaded in compression and tension
50
6.1.1. Model without shear reinforcement
The maximum analytical solution for the shear failure depends on 1 when no shear reinforcement is
used.
N
f cu  35
mm 2
N
f ' c  1,4
mm 2
k   1,0
k h  1,6  h  1,35
 0  0,37%
N
 1  0,4. f c' .k  .k h .3  0  0,4. f 'c  0,56
mm 2
Vd   1.b.d  64,4kN
2.Vd kN
qmax   107,3
L m
In the bending theory presented in chapter 2.1, a compressive normal force is introduced, and the
value 1 may be increased by n according to the Dutch code5.1. This value holds only for linear
elastic calculations, which means it may not be used in this case. Yet it is interesting to see what the
influence might be on the shear strength. The value of Nu follows for the calculation of the bending
theory and is calculated with the excel sheet from Appendix A.
N u  1293kN
Nu N
 bmd   8,62
h.b mm 2
 n  0,15. bmd  1,29
 1,increased   1   n  1,85
kN
qmax  354,6
m
51
1200
1000
800
load in kN/m
600
400
200
0
0 1 2 3 4 5 6 7 8
DIA NA tension softening model

central deflection in m m DIA NA brittle model
A nalytical (t1)
P ark (t1+tn)
Figure 6.6: Load-displacement graph for a model without shear reinforcement
The ultimate load calculated by DIANA is the same as for the model with shear reinforcement in
the case of a model which includes tension softening, which means shear failure is not governing in
this model. For the brittle material behaviour shear failure is governing. The ultimate load for the
brittle model is thus less than the model without shear reinforcement.
6.1.2. Variation in reinforcement layout
As can be seen in Figure 6.5 the flexural reinforcement is loaded in both tension and compression.
The following graph shows the ultimate load for a model with only reinforcement in the tensile
zones, and for a model with no reinforcements at all. The models are all without any shear rein-
forcement.
1200
1000
800
load in kN/m
600
400 no reinforcement ts
brittle
200 reinforcement in tension zones
only ts
brittle
0 reinforcement in tension and
0 1 2 3 4 5 6 compression zones ts
brittle
Figure 6.7: Load-displacement graphs for different reinforcement placements (ts stands for tension softening)
52
First of all, taking into account tension softening enhances the shear capacity considerably. All of
the tension softening models failed in bending. All the brittle models fail in shear. Noticeable is that
the model without any reinforcement has the highest shear capacity of the brittle models. This is
probably due to the fact that large crack widths occur in an un-reinforced structure, which enlarges
the compressive membrane force that is introduced by the lateral restraint edges. This compressive
force has a positive effect on the shear capacity of the one-way slab.
Testing is required to prove that un-reinforced lateral restrained slabs have a greater shear capacity
then reinforced lateral restrained slabs.
6.1.3. Enhancement factors
q DIANA
The enhancement factor is calculated for variation in the slenderness, the concrete and steel
q analytic
strength and the reinforcement percentage. The reinforcement is taken in both the upper and lower
layer, according to the first used model, which can be seen in figure 5.12.
6 6
5 5
enhancement factor
enhancement factor
4 4
3 3
2 2
1 1
0
0
0 0,2 0,4 0,6 0,8 1 1,2 1,4
0 5 10 15 20 25 30
s lenderness re inforcem ent ratio in % tension softening model
tension softening model
brittle model
brittle model
6 6
5 5
enhancement factor
enhancement factor
4
4
3
3
2
2
1
1
0
10 20 30 40 50 60 0
200 300 400 500 600 700
tension softening model tension softening model
fcu in N/m m 2 fs in N/m m 2
brittle model brittle model
Figure 6.8: Enhancement factors according to the finite element model
In all cases the brittle material behaviour gives slightly less enhancement factor then the model with
tension softening, as was to be expected. For low slenderness, the load capacity gets so high that the
value of τ2 gets exceeded. This means that the slab is failing in the line of pressure thrust, which
means it fails at a lower load than it would in pure bending. Therefore the enhancement factor pre-
dicted by the theory of chapter 2.1 is not reached for small values of the slenderness. All other val-
ues are in good comparison with the presented theory.
53
6.1.4. Conclusions
Thus far the following conclusions can be drawn:
 with a simple non-linear finite element calculation it can be shown that the ultimate load of a
horizontal restrained one-way slab is much higher than the analytical virtual work solution
 the theory presented by Park gives a better estimation of the ultimate load than the virtual work
theory, and it is a safe approximation
 even slabs without flexural reinforcement have more capacity then is shown analytically
 due to introduction of normal compressive forces, the shear capacity is improved greatly
 even structures with a high slenderness the enhancement is noticeable
 for high reinforcement ratios and low slenderness, shear failure might become governing
 the presented model is only useable for totally horizontal restrained slabs, for partly restrained
slabs, see the next paragraph
6.2. Partly horizontal restrained model
In reality a horizontal restrained edge is, in most cases not totally, but only partly restrained so the
model will be extended to the model as shown below.
1 1
2 L 2 L
elongation of the slab
k k
k=0
k = infinite
Figure 6.9: Schematic overview of a slab, which is partly horizontal restrained
54
To convert the schematic of Figure 6.9 to a finite element model, there need to be added springs on
the left side so that it can translate. Since rotation is not a degree of freedom for 2D plane stress
elements, a beam element is introduced, which is very stiff and is prevented from rotating at the
edges, see Figure 6.10.
symmetry axis
rotation spring with infinite stiffnes distributed load
same node number shear reinforcement
very stiff CL9BE beam elements CQ16M elements

translation spring with variable stiffnes flexural reinforcement
Figure 6.10: Schematic input for the finite element model
For the beam elements, the L7BEN element can not be used since this is element has only two
nodes and a linear shape function. A quadratic beam element with three nodes is needed to get the
right connectivity. This is explained in Figure 6.11.
1,0
L7BEN element CL9BE element

CQ16M element CQ16M element
linear shapefunction (L7BEN)
1,0
node can not connect all nodes can connect

quadratic shapefunction (CL9BE)
Figure 6.11: L7BEN elements can not be used
In Figure 6.12 the variation in spring stiffness is plotted against the ultimate load, for both the the-
ory of Park and the finite element model. In the finite element model shear reinforcement is in-
cluded, so that bending failure will be governing.
55
1200
1000
ultimate load in kN/m
800
600
400 prediction by Park

total horizontal restrained
200 DIANA tension softening model
DIANA brittle model
0
0,0E+00
5,0E+05
1,0E+06
1,5E+06
2,0E+06
2,5E+06
3,0E+06
3,5E+06
4,0E+06
4,5E+06
5,0E+06
analytic solution
spring stiffness in N/mm
Figure 6.12: Ultimate load for different spring stiffnesses and models
Figure 6.13: Horizontal displacement of a horizontal unrestrained edge
In Figure 6.13 it can be seen that the restrained edge does not move horizontally until cracking load
of the slab is reached. This is in accordance with the theory, which states that compressive mem-
brane forces will be generated after cracking of the concrete.
The normal compressive force that is generated can be calculated by multiplying the maximum
horizontal displacement of the edge of the slab by the spring stiffness, which is a known parameter
since it needs to be inputted in the finite element model. These values are shown in Figure 6.14 and
Figure 6.15.
56
0,8
horzintal movement of the restrained edge in
0,7
0,6
0,5
mm
0,4
0,3
0,2
0,1
0
0,0E+00
5,0E+05
1,0E+06
1,5E+06
2,0E+06
2,5E+06
3,0E+06
3,5E+06
4,0E+06
4,5E+06
5,0E+06
springstiffness in N/m m
Figure 6.14: Maximum horizontal displacement of the edge
2000
1800
compressive normal force in kN
1600
1400
1200
1000
800
600
400
200
0
0,0E+00
5,0E+05
1,0E+06
1,5E+06
2,0E+06
2,5E+06
3,0E+06
3,5E+06
4,0E+06
4,5E+06
5,0E+06
springstiffness in N/mm
Figure 6.15: Compressive normal force in the slab
Now that the restraining edge forced can be calculated, it will be checked if this force is bigger for
unreinforced slabs, as was suggested by evaluating Figure 6.7.
57
900
800
restraining edge force in kN/m
700
600
500
400
300
200
100
0
0 100 200 300 400 500 600
load in kN/m
fully reinforced
no reinforcement
Figure 6.16: restraining edge force for a reinforced and unreinforced slab
The graph starts with a horizontal part with a value of zero. In this part the concrete is not cracked
yet and the edge does not move laterally. Then, after the cracking load is reached the restraining
force starts to develop. The unreinforced slab has indeed a large restraining force for the same load-
ing as the reinforced slab.
6.3. Comparison with test results
The results as in chapter 4.1.1 will be used. See Table 4.1 for an overview.
In Table 6.1 the finite element calculations are compared to experimental and the predicted results.
This is performed for a model with brittle cracking and a model that takes into account tension sof-
tening.
If the three slabs with the highest deviations (slab 1, 8 and 12) are excluded from the calculation of
the mean value and standard deviation, the test results correspond very well with the finite element
models. Even when these experimental results are included, the results lie in between an acceptable
range.
58
Slab Pe Pp Pfem,ts Pfem,brit Pfem,ts/Pp Pfem,brit/Pp Pfem,ts/Pe Pfem,brit/Pe

kN/m kN/m kN/m - -
1 328,1 242,3 288 232 1,19 0,96 0,88 0,71
2 218,7 236,3 282 227 1,19 0,96 1,29 1,04
3 302,8 274,1 356 320 1,30 1,17 1,18 1,06
4 298,6 264,7 348 309 1,31 1,17 1,17 1,03
4A 290,2 261,2 343 305 1,31 1,17 1,18 1,05
4B 323,9 262,7 345 309 1,31 1,18 1,07 0,95
5 412,2 315,5 414 380 1,31 1,20 1,00 0,92
6 382,7 310,4 401 376 1,29 1,21 1,05 0,98
7 134,6 117,9 184,0 140 1,56 1,19 1,37 1,04
8 96,7 117,3 182,0 132 1,55 1,13 1,88 1,36
9 168,3 137,9 215 186 1,56 1,35 1,28 1,11
9A 172,4 137,8 215 184 1,56 1,34 1,25 1,07
10 - 137,5 214 178 1,56 1,29 - -
10A - 137,3 214 173 1,56 1,26 - -
11 193,5 152,9 236 193 1,54 1,26 1,22 1,00
12 92,5 153,5 234 188 1,52 1,22 2,53 2,03
Mean value 1,42 1,19 1,31 1,10
Standard deviation 0,15 0,11 0,42 0,30
Mean value, slab 1, 8 & 12 excluded 1,19 1,02
Standard deviation, slab 1, 8 & 12 excluded 0,11 0,05
Table 6.1: Finite element calculations compared to experimental- and predicted results
Figure 6.17 and Figure 6.18 show the finite element results of a brittle cracking model and the ex-
perimental results.
450
400
350
300
failure load in kN/m
250
200
150
w0 = 0,52% fy = 344,8 N/mm2
100 w0 = 0,74% fy = 403,2 N/mm2
w0 = 1,06% fy = 403,2 N/mm2
50
ultimate load by DIANA
0 ultimate load by DIANA
28,5 29,0 29,5 30,0 30,5 31,0
fcu in N/mm2
slab 1
Figure 6.17: Test results versus finite element (brittle model) calculations for the thick slabs of 58,7 mm
59
225
200
175
failure load in kN/m
150
125
100
75 w0 = 0,52% fy = 464,2 N/mm2

w0 = 1,14% fy = 403,2 N/mm2
50
w0 = 1,47% fy = 403,2 N/mm2
25
0 ultimate load by DIANA
34,1 34,2 34,3 34,4 34,5 34,6 34,7
fcu in N/mm2
slab 8 & 12
Figure 6.18: Test results versus finite element (brittle model) calculations for the thin slabs of 41,2 mm
6.4. Brittle versus tension softening cracking model
To get a better view of the difference between the tension softening and the brittle cracking model,
Pfem
the factor of test results is plotted in Table 6.1. It can be clearly seen that the calculations with
Pe
a brittle cracking model lies in between upper and lower 15% boundaries. Further more it can be
seen that the model that includes tension softening gives results that are around 20% to high.
1,5
1,4
1,3
1,2
Pfem divided by Pe
1,1
1,0
0,9
0,8
0,7
calculation with tension softening
0,6 calculation with brittle cracking
0,5 15% lower boundary

slab number
15% upper boundary
Figure 6.19: The brittle versus the tension softening model
60
Looking at Figure 6.12 some things can be noticed. The analytical solution is determined without
the use of tension softening. This corresponds very well with the brittle finite element calculation
without any spring stiffness. The finite element model with tension softening gives a higher ulti-
mate load of approximately 30% for an unrestrained edge. For a totally horizontal restrained edge
the following factors can be determined:
q fem ,brittle
 1,19
qbendingtheory
q fem ,tensionsoftening
 1,47
qbendingtheory
q exp erimets
 1,12
qbendingtheory
The last factor is the mean value from the test data as shown in Table 4.1. Comparing those three
factors and the finite element calculations of the experimental results, it seems that the brittle finite
element model will give a better approximation of the collapse load then the finite element model
that includes tension softening.
The difference in ductility between the brittle and tension softening model is very large. Including
tension softening makes the behaviour of the material much more ductile. This can be clearly seen
in Figure 6.3, Figure 6.6 and Figure 6.7.
Another difference between the brittle and the tension softening is the calculation time. The brittle
material model needs more iterations to reach its convergence criterion in each step. This increases
the calculation time considerably. For this 2D plane stress model it is not much of an influence, the
calculation time stays within reasonable limits. For a fully 3D solids model however this will be
something to keep in mind.
.
61
Chapter 7: Finite element model for punching failure
7. FINITE ELEMENT MODEL FOR PUNCHING FAILURE
A punch cone can only occur in a 3D or an axi symmetric model (the latter is discussed in chapter
7.2). The 2D plane stress model used in chapter 6 is thus not suitable for punching analysis. First,
the 3D solid model, as presented in chapter 5.1.4 will be evaluated.
7.1. 3D solids model
The model that will be evaluated is shown in Figure 7.1. The dimensions are chosen equal to the 2D
plane stress model.
12
00 00
12
0 10
10 0
fcu = 35 N/mm2
150
fs = 435 N/mm2
w0 = 0,37%
115
150
115
100
1200
Figure 7.1: Dimensions of the 3D solid model, all the edges are clamped and totally horizontal restrained
The analytic solution for this plate is the lowest value of the maximum bending or punching load.
As  w0bd  510,6mm 2
As f s N
ns   185,1
b mm
 4A f  Nmm
mu  ns z  ns  d  0,39 s s   20438,1
 3 f 'c b  mm
See chapter 3.1.1.
63
b=l
l
yield line
load
Figure 7.2: Yield lines in a clamped square slab
The ultimate bending load is determined with the virtual work theory, as presented in a reader 7.1.
To do the calculation, a few simplifications are made:
 the load is located in one point, and not distributed over the give area
 the ultimate moment of the slab is the same in all directions
Wexternal  F
2

b
2

2b 2

Wint ernal   4b mu  2 2 2b 2 mu 
Wexternal  Wint ernal  Fu  8mu  8mu  16mu
Fu  327kN
64
The ultimate punching load is calculated as presented in chapter 3.1.2.
N
f ' c  1,4
mm 2
a1  100mm
4a1
a  127,3mm

k d  1,5  0,6d  1,431
 1  0,8 f ' c k d 3 w0  1,15
p   (a  d )  761,36mm
Fu   1 pd  100,7kN
As was to be expected Fu, punch < Fu, bending, so the punching failure mode is governing. The predicted
collapse load including compressive membrane action calculated by the method as described in
chapter 2.2 is 161 kN (calculated with maple). An overview of the found values is shown below.
Fanalytic  100,7kN
Fpredicted  161kN (including compressiv e membrane action)
In the figures below, the DIANA output of laterally restrained clamped slab is presented. The load
is distributed over an area of 100 x 100 mm. Only a quarter of the slab is modelled, since the slab
has two symmetry axes. The model will be build up just like the model presented in chapter 5.1.4,
but now with boundary conditions of the two supported sides fixed in lateral and z-direction. Both
brittle and tension softening models will compared to each other. To check whether DIANA can
calculate punch. So four models will be compared to each other:
 brittle material model, all sides simply supported

 brittle material model, all sides clamped
 tension softening material model, all sides simply supported
 tension softening material model, all sides clamped
65
300
250
200
load in kN
150
model figure (a) brittle

100
model figure (b) brittle
50 model figure (a) tension

softening
model figure (b) tension
softening
0
punch capacity according
0 0,1 0,2 0,3 0,4 to the Dutch code
central deflection in mm predicted solution
Figure 7.3: The 4 load-displacement graphs of the 3D solids model
(a) simply supported

no compressive membrane action
(b) clamped and lateral restrained

compressive membrane action
Figure 7.4: Difference between the simply supported and clamped model
As already noticed in chapter 6, there is a big difference in a model with and without tension soften-
ing. Furthermore it can be seen that the brittle material models collapse before reaching the punch-
ing shear load as calculated according the Dutch code. The clamped model with the tension soften-
ing material model comes closest to failure in punching shear. For this model, the load-
displacement graph, the displacement fields and the cracking pattern are shown in Figure 7.5 to
Figure 7.7. This model seems to give a collapse load that might be too high, as the predicted value
was more or less in agreement with experimentally found results, see chapter 4.2.
66
Figure 7.5: Load displacement graph for the 3D solid clamped tension softening model, the model fails between loadstep 23 and 24
Figure 7.6: Displacement field in Z-direction and cracking pattern of the slab just before failure (loadstep 23)
67
Figure 7.7: Displacement field in Z-direction and cracking pattern of the failed slab (loadstep 24)
The results of the 3D solids model seem not to correspond to the expected values, see Figure 7.3.
Varying parameters to see what influence they have is very time consuming, since each calculation
takes a long time. To spare time an axi symmetric model will be evaluated. Since punch is the gov-
erning failure mode, this model might give good results.
7.2. axi-symmetric model
In this model a 2D slice is modelled in DIANA, and is then rotated 360 degrees around the y-axis.
See Figure 7.8. To get a good comparison, the input is the same as the 3D solids model. This model
can only be used to model circular shaped structures. This holds also for the loaded area, which is
not square, but circular shaped. The diameter of the area is 127,3 mm.
concrete slab
loaded area
2D slice modeled in DIANA
reinforcement grid
x z
axi-symmetric model
Figure 7.8: axi-symmetric model
68
centre of the plate
support
loading
reinforcement grid
CQ16A elements
Figure 7.9: Schematic input of the axi-symetric model
This model is again evaluated for the same four cases as the 3D solids model. For the difference
between (a) and (b) see Figure 7.4.
300
250
200
load in kN
150
model figure (a) brittle
100
model figure (b) brittle
model figure (a) tension

50
softening
model figure (b) tension
softening
0
punch capacity according
0 0,2 0,4 0,6 0,8 to the Dutch code
central deflection in mm predicted punch failure
Figure 7.10: Load-displacement graphs for 4 different models of the axi-symmetric model
The difference between the brittle and tension softening material behaviour can be clearly seen
again. The tension softening material model is again much more ductile. The simply supported
variant with a brittle material model seems to fail in bending, this can be seen because it has a hori-
zontal part in load-displacement graph.
The lateral restrained model with the brittle material behaviour is almost equal to the predicted
punching load failure as described in chapter 2.2. To check whether this is a coincidence or that the
69
Pp
models really give the same results, a parameter study is done and the factor is calculated. The
Pfem
values of the steel ratio and steel strength are not taken into account, since these have little influ-
ence on the punching capacity.
d0 Pp Pfem Pa Pfem/Pa Pfem/Pp h Pp Pfem Pa Pfem/Pa Pp/Pfem

mm kN kN kN - - mm kN kN kN - -
50 78,5 76,8 74,2 1,04 0,98 50 46,3 24,8 24,9 1,00 0,54
100 161 156 100,7 1,55 0,97 100 100,6 92 61 1,51 0,91
150 228,2 222 127,1 1,75 0,97 150 161 156 100,7 1,55 0,97
200 295,3 281 153,6 1,83 0,95 200 227,5 222 166,1 1,34 0,98
250 362,5 359 180,1 1,99 0,99 250 299,4 292 249,2 1,17 0,98
Mean value 1,63 0,97 Mean value 1,31 0,87
Standard deviation 0,37 0,01 Standard deviation 0,23 0,19
fcu Pp Pfem Pa Pfem/Pa Pp/Pfem

N/mm2 kN kN kN - -
15 49,1 87 63 1,38 0,56
25 102,5 128 85 1,51 0,80
35 161 156 100,7 1,55 1,03
45 222,7 188 115,6 1,63 1,18
55 286,1 223 133,1 1,68 1,28
Table 7.1: The difference between the predicted, analytical and finite element values
Looking at Table 7.1 it can be seen that for the concrete strength fcu = 35 N/mm2 the predicted value
and the finite element value correspond very well. Only for very large slenderness (L/h = 24 in this
cases) the results do not match. The punch load according to the Dutch code for this plate with a
height of 50 mm is 24,9 kN. This value corresponds very will with the finite element result. This
indicates that for a finite element calculation in DIANA, in slabs with a large slenderness no com-
pressive membrane action is generated. The theoretical model presented in chapter 2.2 gives a result
that does include compressive membrane action for a large slenderness. The influence of the con-
crete strength differs. The finite element results show that it has less influence than the predicted
solution. To determine which model is more accurate, they have to be compared to experimental
results.
70
350
300
250
punch load in kN
200
150
100
50
0
0 10 20 30 40 50 60
fcu in N/mm2
Predicted res ults
Finite element results
Figure 7.11: Difference between the predicted and finite element value for different concrete strengths
In Figure 7.12 and Figure 7.13 the load-displacement graph and the crack patterns are shown for a
axi-symmetric model with a brittle material behaviour. It can be seen that this model is failing in
punch.
Figure 7.12: Load-displacement graph for the clamped model, with a brittle material behaviour, force controlled
71
Figure 7.13: Crack pattern just before (loadstep 16) and after (loadstep 17) failure
From now on the axi-symmetric model with a brittle material behaviour will be used. This model is
chosen above the 3D solids model based on the following points
 all tests described in chapter 4.2 failed in punch and the axi-symmetric model predicts punch
well
 the model is fast and easy to build
 the calculation time is much less than that of a 3D solids model
7.3. Enhancement factors
The enhancement factors for variation in the length of the loaded area, the height of the slab and the
concrete strength are shown in Figure 7.14.
enhancement factor P/Pa
2,00 2,00
1,60
1,60
1,20
1,20
0,80
0 50 100 150 200 250 300 0,80
f inite element 0 5 10 15 20 25 30
d0 in m m
predicted f inite element
sle nderne ss
predicted
2,00
1,60
1,20
0,80
0 10 20 30 40 50 60
finite element
fcu in N/m m 2
predicted
Figure 7.14: Enhancement factor Pfem/Pa and Pp/Pa for variation in d0, slenderness and fcu
The theory of chapter 2.2 does not take into account that compressive membrane action is not pre-
sent in slabs with a high slenderness. This can be concluded from the high enhancement factors
given by this theory for a high slenderness. These values are in contradiction with the finite element
results.
72
The reinforcement placement is varied from no reinforcement to reinforcement in tension zones to

continuos reinforcement. The un-reinforced slab has a higher collapse load then the slabs that does
include reinforcement. A similar result as already been seen for the bending model presented in
chapter 6.1.2. The higher punch resistance of the un-reinforced slab might be explained by the fact
that the compressive membrane force generated in the un-reinforced slab is higher than that of a
reinforced slab, see Figure 6.16. Unfortunately, no experimental data was found that includes dif-
ferent reinforcement layouts, so it can not be verified if these results are right at this moment.
200
180
160
140
120
load in kN
100
80
60
40 no reinforcement
20
reinforcement in tension zones only
0
0 0,1 0,2 0,3 0,4 0,5 reinforcement in tension and
compression zones
central deflection in m m
Figure 7.15: ultimate punch load for different reinforcement locations
7.4. Comparison with test results
The experimental results described in chapter 4.2.1 are compared to a finite element calculation.
The results are presented in Table 7.2.
Pfem Pfem ts Pe Pp Pfem,ts/Pe Pfem,ts/Pp

kN kN kN kN - -
46 118 101 104 1,17 1,14
35 97 118 64 0,82 1,52
44 105 149 81 0,70 1,30
25 60 49 63 1,22 0,95
23 63 70 59 0,90 1,08
16 61 68 53 0,90 1,15
46 115 116 95 0,99 1,21
24 65 42 67 1,55 0,97
27 73 69 80 1,06 0,91
46 120 99 96 1,21 1,25
25 62 43 62 1,44 0,99
28 83 63 81 1,32 1,02
Table 7.2: experimental results presented in chapter 4.2.1 compared to finite element results
The slenderness of the tested slabs are respectively 20 and 30. As already mentioned in chapter 7.2
the used finite element model does not take into account compressive membrane action for these
high values of the slenderness. This is not in comparison with the experimental results. Therefore, a
73
second finite element analysis is done, but now with the use of the tension softening model. The
result from these analyses seems to be in better accordance with the experimental result.
In Table 7.3 the experimental result presented in chapter 4.2.2 are compared to the finite element
results. Since the concrete strength is rather high in these experiments, the brittle material model
might give values that might be somewhat lower than the experimental result, see Figure 7.11.
Therefor the finite element calculation will also be done with a tension softening material model.
The tension softening material model gives results that lie in acceptable range of the experimentally
found values.
Pfem Pfem ts Pe Pp Pfem,ts/Pe Pfem,ts/Pp

kN kN kN kN - -
95 416 369,4 466,7 1,13 0,89
79 372 290,6 367,6 1,28 1,01
85 380 402,2 403,4 0,94 0,94
81 368 394,1 376,6 0,93 0,98
The results of the measurements on the bridge decks presented in chapter 4.2.4 are compared to
finite element models in Table 7.4.
measured values
slab cracking load deflection at deflection at deflection at
kN 112,5 kN in mm 330 kN in mm 430 kN in mm
C1 118 0,35 2,6 -
C2 143 0,25 - 1,2
D1 154 0,4 1,85 -
D2 143 0,25 - 1,75
finite element results (totally horizontal restrained)

slab cracking load deflection at deflection at deflection at
kN 112,5 kN in mm 330 kN in mm 430 kN in mm
C1 120 0,175 0,815 -
C2 112 0,08 - 0,5
D1 150 0,12 0,55 -
D2 112 0,08 - 0,5
The slabs A1 to B2 included reinforcement in the form of fibres, which can not (yet) be included in
the finite element model. The results of the finite element model do not match at all with the ex-
perimental found values. However, the goal of this study is to predict the ultimate load of the struc-
tures, and not an accurate value for deflections and crack widths.
7.5. Partly horizontal restrained axi-symmetric model
To make the model as described in chapter 7.2 partly horizontal restrained, springs will be added,
just like is done for the 2D plane stress model as described in chapter 6.2. However, the use of a
stiff edge beam is not a possibility anymore. These stiff beam elements will be rotated over 360
degrees, which creates a shell of revolution. This shell of revolution acts as a stiff ring element
74
which can not deform horizontally freely and is thus not useful. Instead the tying option will be
used. With this input, a group of nodes can be tied togheter, so they have the same displacement in
a chosen direction. Tying the nodes of the edge together will create the boundary condition that is
needed, an edge that can move latterly but does not rotate.
centre of the plate
loading
springs to set the grade of lateral restrainedness
nodes are tyed to have the same displacement in x-direction
reinforcement grids
For low spring stiffness the slab is failing in bending, which can be identified in the load displace-
ment curve, which ends in a horizontal part if it fails in bending. Shear or punch failure can be iden-
tified by abrupt ending of a rising curve, or by a fallback of the load. For higher spring stiffness the
slab fails in punching shear.
180
160
140
ultimate load in kN
120
100
80
60
40
20
0
1,0E+00
1,0E+01
1,0E+02
1,0E+03
1,0E+04
1,0E+05
1,0E+06
1,0E+07
1,0E+08
1,0E+09
1,0E+10
1,0E+11
1,0E+12
1,0E+13
Springs tiffness in N/m m per m m circumference

finite element results
punch load according to the

Dutch code
ultimate load for a fully
lateral restrained model
Figure 7.16: Ultimate load of the finite element calculation for different spring stiffnesses
75
With the spring stiffness per m circumference and the horizontal displacement known, the value for
the force generated by the lateral restraint can be calculated by multiplying the horizontal displace-
ment by the spring stiffness.
N  u x .S .
restraining edge force in kN/m circumference
700
600
500
400
300
200
100
0
1,0E+00
1,0E+01
1,0E+02
1,0E+03
1,0E+04
1,0E+05
1,0E+06
1,0E+07
1,0E+08
1,0E+09
1,0E+10
1,0E+11
1,0E+12
1,0E+13
springstifness in N/mm per m circumference
Figure 7.17: Nrs for an increasing spring stiffness
7.5.1. Ontario experimental results
In Ontario some experiments where done on bridge decks, see appendix D. Some small scale ex-
periments where done, at which the membrane forces where measured by the expansion of a steel
ring. The test set-up is shown in Figure 7.18.
steel ring
reinforcement 0,2 - 0,3 - 1,0 %
fcu = 35 N/mm2
fs = 310 N/mm2
32 - 38- 44
63
572 25 - 32 - 38
610
Figure 7.18: test set-up of small-scale test
76
For the calculation of the spring stiffness the linear relationship between spring and force will be
used, F = k.u.
nt
nr r
Figure 7.19: deformation in lateral direction before and after loading
In the formula r is the radius and p the perimeter of the slab.
p  2r
n .p
p  t
E. A
nt  n r .r
p  p
r  r 
2
n r .2r
2r  r 2
r  E. A  r  n r r
2 E. A
nr  S .r
EA N / mm 2
S
r 2 mm circumference
pr
For the conversion of radial into tangential stresses, the formula   , in which p is a pressure
t
and not the perimeter, is used7.2. For the input in the finite element model the found value has to be
N / mm 2
multiplied with a factor 1000, since the input is in .
m circumference
77
A finite element model will be used to calculate the restraining force and will then be compared to
the values presented in the Ontario code, see appendix D Figure 18. It is not clear in the Ontario
code what dimensions each of the samples have, so for the finite element model, the average of the
values presented in Figure 7.18 are used.
250
restraining force in kN/m circumferrence
200
150
100
50
0
0 5 10 15 20 25 30 35 40
applied load in kN initial restraing force 0 kN

initial force of 40 kN
Figure 7.20: restraining force calculated by multiplying the spring stiffness with the lateral displacement
Comparing Figure 7.20 with the one presented in appendix D, a few differences can be noticed.
 the ultimate load of the Ontario specimens is higher than the load found with the finite element
model
 the maximum restraining load of both methods is about the same
 the Ontario specimens started with a restraining load for a unload slabs, which might indicate
that the steel ring was attached tightly to the specimen, this is not modelled in the finite element
model
It is not clear why the graph given in the Ontario report does not show a horizontal part at the be-
ginning. Even when a starting with a compression force on the edge, the cracking load still has to be
reached, before the compressive membrane action starts to form.
78
Part III: A Practical example
79
Chapter 8: The deck of a common girder bridge
8. THE DECK OF A COMMON GIRDER BRIDGE
As an example, a ZIP-girder system will be evaluated, which is a commonly used system in the
Netherlands. This system has already been introduced in chapter 1. The system will be designed
according to the guidelines of the Dutch company Spanbeton. All information according to dimen-
sions is downloaded from their Internet page 8.1. The thickness of the compression layer (the con-
crete slab) is the same for every case, 230 mm. This means the span is not an influence factor for
the height of the slab. Chosen is for a ZIP 1200 system.
35 35
280
230
35
30
5
300
300
50
50
formwork
pressure layer
1225
1220
565
565
100
100
35
35
140
140
215 235 280 235 215 215 235 280 235 215
1180 1180
ZIP 1200
250
10
120
70 110
pressure layer
1200
1200
1010
145 350 140 350 145 350 140 350

985 985
TRA edge beam
Figure 8.1: Dimension of the ZIP profile and the edge beam
The system has the following quantities:
 the concrete quality of the deck is C28/35

 the top reinforcement is Ø12 – 150 and the bottom reinforcement is Ø16 -100
 the concrete cover is 35 mm (according to the code 5.1)
 the effective height is 187 mm
 the area over which the load is spread is 350 x 600 mm
81
edge beam
in-situ concrete
top layer
compression layer
formwork
1200 1200
vehicle wheel
600
225
1200 230
850
920
Figure 8.2: Cross-section of the bridge and one span of the compression layer zoomed in to
Since punch failure is governing in the most cases, only load configuration 2 (see Figure 1.2) with
the high axle loads will be discussed here.
82
8.1. Analytical solutions
The ultimate load for both bending and shear will be calculated. Some assumptions that are made
for these calculations are:
 the top layer does not contribute to the bearing capacity

 the form work does not contribute to the bearing capacity
 the length of the mechanical system is the distance between the centre lines
8.1.1. Bending
600
variable wheel load
permanent load
600 600
model
12 - 150
230
187
950
16 - 100
cross section
Figure 8.3: Model for the calculation of the bending capacity

300
350
950
45°
300
600
1200
Figure 8.4: Width over which the wheel load has influence
83
950
As ,top  . .62  716,3mm2
150
950
As ,bottom  . .82  1910,1mm2
100
 4.716,3.435 
M u , top  716,3.435.187  0,39.   55,7kNm
 3.21.950 
 4.1910,1.435 
M u , top  1910,1.435.187  0,39.   137, 4kNm
 3.21.950 
F = 0,5 x Fwheel max F = 0,5 x F wheel max

F = 5,75 x 0,6 = 3,45 kN
F = 5,75 x 0,6 = 3,45 kN
300 150 150
Figure 8.5: load positions to calculate the virtual work
Wextern  2.3,45.0,3  2.0,5 Fwheel , max 0,45

Wint ern  2 M u ,top  2M u , ,bottom
Wextern  Wint ern  Fwheel , max  853,6kN
The analytic value is higher than the 200 kN wheel load which is prescribed in the code.
8.1.2. Punch
The l/b ratio of the loaded area is less then 2, which means the slab will fail in pure punch, and not a
combination of punch and shear.
A
 o, top  s ,top  0,40%
b.d
A
 o, bottom  s , top  1,08%
b.d
2.600  350
a  604,8mm

84
p   (604,8  187)  2487,5mm

kd  1,5  0,6.0,187  1,39
 1  0,8.1,4.1,39.3 0,74  1,4.0,8   1  1,41
Fwheel , max  1,41.2487,5.187  655,9kN
The slab will fail in punching shear, and not in bending, at least for the analytic solution.
8.2. solutions including compressive membrane action
The example structure will be calculated with the bending and punch theory described in chapter 2
and with an axial symmetric finite element model. The example will be calculated according to the
New Zealand code, which takes into account compressive membrane action by using a restrained
factor (see appendix E).
8.2.1. Bending
The theory predicts the ultimate distributed load, which is 1436,4 kN/m (calculated with the excel
sheet presented in appendix A). To calculate the maximum wheel load, it is assumed tat the maxi-
mum bending moment in the middle will be the same for the total distributed load, and the system
with the more concentrated wheel load, see Figure 8.6.
qmax wheel = 1716,8 kN/m
q = 1940,4 kN/m
q = 5,75 kN/m
M = 121 x 1940,4 x 1,2 x 1,2 = 232,8 kNm M = (1,725 + 0,3q) x 0,3 + 0,5 x (1,725 + 0,3q) x 0,3 +
0,5 x (1,725) x 0,3 = 232,8 kNm
V = 0,3 x 5,75 + 0,3 x q wheel max

V = 0,6 x 5,75 + 0,3 x q wheel max
Figure 8.6: conversion from a full distributed to a wheel load
Fwheel , max  1716,8.0,6  1030,1N

1030,1
Using the compressive membrane action for bending gives an enhancement factor of  1,57 .
655,9
85
8.2.2. Punch
To calculate the ultimate punch load with the theory presented in chapter 2.2, the loaded area has to
be square. To calculate the length of the square, it is assumed that the square and rectangle have the
same area.
2
600.350  l square  l square  458,3 mm
The maximum wheel load calculated with punching shear theory is 1012,8 kN (calculated with ma-
1012,8
ple, see appendix B). The enhancement factor now becomes  1,54
655,9
8.2.3. Finite element model
The axi-symmetric finite element model will be used. The brittle material model should give rea-
sonable results, as the concrete strength is 35 N/mm2 and the slenderness is 5,2 (see Figure 7.14).
In the axi-symmetric model, only circular loaded areas can be created. The diameter of the loaded
4.458,3
area is  583,5mm . The fully restrained model will be used.

910
The ultimate load of the finite element model is 900 kN. The enhancement factor is  1,38 .
655,9
8.2.4. New Zealand code
The New Zealand code is one of the first international codes that takes into account compressive
membrane action in bridge decks (see Appendix E, section 6.5). The empirical method described in
this code may be used if the following conditions are met:
 the supporting beams are steel or concrete

 the diaphragms are continuos and present at all supports for pre-stressed concrete beams
 the slenderness does not exceed 20
 the span length does not exceed 4,5 meter
 the concrete strength f’c is not less then 20 N/mm2
 the minimum slab thickness is 150 mm
 the overhang of the outer beam is at least 80 mm
The example bridge deck meets all of the above requirements, and may thus be analysed with the
empirical method.
The maximum wheel load in kg can be calculated with the following empirical method. Since no
safety factors are taken into account so far, the value for γL will be assumed 1,0 to get a fair com-
parison of the values.
86
  0,6 Ri  
  .40.I .8200
 L 
N
f 'c  21
mm 2
L  1200mm
w0  0,74% (called q in the New Zealand code)
Ri  1575kN (figure 6.4 appendix E)
  0,90 D (table 6.6 appendix E)
 D  0,5
 L  1,9 (table 6.3 appendix E)
I  1,3
 0, 45.0,6.1575 
 1,0.40.1,3 .8200  67058,7 kg
 
67058,7
Fwheel , max   684,3kN
98
684,3
The enhancement factor is  1,04
655,9
8.3. Overview of results
For the analytical solution, which does not take into account compressive membrane action into
account, the collapse load is 655,9 kN and the deck fails in punch.
The enhancement factors of the four methods to take compressive membrane action into account
from low to high are:
 New Zealand code (appendix E) 1,04

 finite element model (chapter 7.5) 1,38
 punch theory (chapter 2.2) 1,54
 bending theory (chapter 2.1) 1,57
87
The minimum concrete strength for the deck subscribed by Spanbeton is C28/35, but in practice
C50/60 is commonly used. If this concrete strength is used in the calculations, the bearing capacity
and enhancement factors are:
 bending analytic 887,0 kN

 punch analytic 888,5 kN
 New Zealand code 899,3 kN 1,01
 finite element model 1360,0 kN 1,53
 bending including compressive membrane action 1415,4 kN 1,60
 punch including compressive membrane action 1794,7 kN 2,02
8.4. Existing bridges
Existing bridges are a point of discussion in the Netherlands, as already mentioned in the article in
chapter 1. To see whether these bridges meet the bearing capacity as prescribed in the NEN-EN
1991-2, the minimum enhancement factor will be determined and then checked if it can hold the
wheel load of 200 kN (load configuration 2, see Figure 1.2). The above used dimensions are only
for new to build bridges, for this calculation commonly used dimensions in the 1970’s will be used.
These dimensions of are given in Figure 8.7. A recent study showed that the concrete quality of the
decks of this existing bridges is at least C50/60. For more information on this study Dr.ir. C. van
der Veen from Delft University of Technology can be contacted.
fcu = 55 N/mm2
fs = 350 N/mm2
d = 117 mm
1200
160
117
10 - 150
12 - 150
Figure 8.7: Commonly encountered dimensions of existing bridge decks
88
The bearing capacity and enhancement factors are:
 bending analytic 220,3 kN

 punch analytic 477,6 kN
 New Zealand code 488,8 kN 2,22
 bending including compressive membrane action 597,4 kN 2,71
 finite element model 600,0 kN 2,72
 punch including compressive membrane action 1225,4 kN 5,56
It can be concluded that a 200 kN wheel load, as prescribed in the EN-NEN 1991-2 can be carried if
compressive membrane action is taken into account.
The enhancement factor found by the bending theory lies again between 2,2 and 2,7. Looking at the
similarity of the bending enhancement factors and taking into account that this method gives a
lower bound vale, it can be concluded that the enhancement factor for these kind of bridge decks
and load configuration 2 (see Figure 1.2) is at least 2,7 and increases for higher concrete strengths.
The presented calculation does not include any safety factors for the loading. The load factor for
live loads on bridges is 1,35 according to the European code 8.2. This means that the deck must have
the capacity to withstand an wheel load of 1,35 x 200 = 270 kN. The lower-bound value found by
the theory’s including compressive membrane action is significant higher then this prescribed load.
It can be concluded that the deck can withstand the 270 kN wheel load if compressive membrane
action is taken into account.
Not taking into account the load safety factor for the does not chance the value of the enhancement
factors, since the load safety factors are equal for both calculations.
89
Chapter 9: Conclusions and recommendations
9. CONCLUSIONS AND RECOMMENDATIONS
9.1. Conclusions
The goals as described in chapter 1.1 are met. This rapport describes finite element models, for
bending and punch failure, to predict the bearing capacity of latterly restrained concrete slabs.
From this rapport the following things can be concluded:
 if a concrete slab is laterally restrained, the bearing capacity is higher than can be shown with a
linear analytic calculation
 the finite element program DIANA takes into account compressive membrane action and gives
acceptable results (compared to experimental data) if
 the calculation method used is non-linear
 the material is modelled as total strain rotating brittle cracking model
 the 2D plane stress model is used for bending failure, or the axi symmetric model is used for
punch or a combination of punch and bending failure
 the concrete strength fcu lies around 35 N/mm2
 he slenderness is less than 15
 for lower values of the concrete strength the found values are to high, but compression mem-
brane action is still taken into account. For higher values of the concrete strength the found val-
ues are to low, but compression membrane action is still taken into account
 for higher values of the slenderness DIANA does not take predict compressive membrane ac-
tion, while experimental results show that the phenomenon is still present
 the enhancement factor, which is the ultimate load from the finite element solution divided by
the ultimate load of the analytic solution, differs for different input parameters
 for a higher concrete strength the enhancement factor will increases
 for a higher reinforcement percentage the enhancement factor will decrease
 for a higher reinforcement yielding strength the enhancement factor will decrease
 for a higher slenderness the enhancement factor will decrease
 for a low slenderness failure of the trust relieving arch might become governing (the value τ2
gets exceeded), in which case the enhancement factor will be lower
 if a concentrated load is spread over a larger area the enhancement factor increases
 the finite element model can include partly laterally restrained edges, this lowers the enhance-
ment factor
 the decks of the in the Netherlands commonly used ZIP girder structures have a minimum en-
hancement factor of approximately 1,5
 using compressive membrane action to determine the bearing capacity of existing decks of ZIP
girder bridges shows that these decks can carry the loads prescribed by the newest Euro codes
91
9.2. Recommendations
Recommendations for further research on the following subjects are done.
9.2.1. Tension softening
Using a tension softening instead of a brittle material behaviour, the results seem to be too ductile.
The model does not fail in punch shear anymore, but fails in bending. Test specimens do fail in
punch. Research has to be done to verify the different parameters such as the tensile fracture energy
and the crack bandwidth, to come to a more accurate model.
9.2.2. Reinforcement layout
Finite element results show that an un-reinforced slab has a higher punch capacity than a reinforced
slab provided that only flexural and no shear reinforcement is included. This might be due to the
fact that in an un-reinforced slab the generated compression membrane force is larger, which makes
the enhancement of the punch capacity larger. However, no test results where found to confirm this,
so it is recommended to do testing on slabs with and without reinforcement to see if the results of
the finite element model can be verified.
9.2.3. 3D solids model
Doing research on a 3D solids model is still very time consuming due to convergence difficulties
and is therefore not continued in this thesis. However, this model might give very accurate results if
the model is build-up in accordance with the specimen (which is not the case in an axi-symmetric
model, since only circular structures can be modelled with this model). Research has to be done, to
see if the 3D solids model can give an accurate value for the collapse load. If that is the case, this
model can be extended for more complex structures, including irregular edges or holes for example.
9.2.4. Serviceability limit state
This study only checks the bearing capacity for bridge decks in the ultimate limit state. The service-
ability limit state has to be checked to see if the decks fulfil all the requirements of the latest Codes.
92
Chapter 10: References
10. REFERENCES
1.1 P. de Waard, De volkskrant, 4 oktober 2007
2.1 A. Hon, G. Teplin and R.S. Al-Mahaidi, Structural journal June 2005 “Strength of rein-
forced concrete bridge decks under compressive membrane action”
2.2 R. Park and W. L. Gamble, “Reinforced concrete slabs, 2nd edition” 2000, chapter 12
2.3 W. Salim and W. M. Sebastian, Structural journal July 2003 “Punching shear failure in re-
inforced concrete slabs with compressive membrane action”
2.4 J. S. Kuang and C. T. Morley, International journal of mechanical sciences 1993 “A plastic-
ity model for punching shear of laterally restrained slabs”
2.5 B. Mullin, The structural engineer 2003 “Arching action in concrete slabs with novel rein-
forcement”
3.1 C. Hartsuijker and H. Welleman, “Toegepaste mechanica deel 3: Statisch onbepaalde con-
structies en bezwijkanalyse” module 2chapter 5.3
3.2 R. Sagel and A. J. van Dongen, “Constructeileer gewapend beton” 2000, chapter 11
3.3 “NEN 6706, traffic loads on bridges” 2007, chapter 7.2
4.1 L. K. Guice and E. J. Rhomberg, Structural journal July 1988 “Membrane action in par-
tially restrained slabs”
4.2 J.S. Kuang and C.T. Morley, Structural journal January 1992 “Punching shear behaviour of
restrained reinforced concrete slabs”
4.3 S.E. Talor, B. Rankin, D.J. Cleland and J. Kirkpatrick, Structural journal January 2007
“Serviceability of bridge deck slabs with arching action”
5.1 “NEN 6720 Voorschriften beton, constructieve eisen en rekenmethoden”
7.1 A.C.W.M. Vrouwenvelder and Witteveen, “The plastic behaviour and the calculation of
plates subjected to bending”
7.2 J.M. Gere and S.P. Timoshenko, “Mechanics of materials”
8.1 www.spanbeton.nl
8.2 NEN 6702 “Technische grondslagen voor bouwconstrcuties –TGB 1990– Belastingen en
vervormingen”
93
Appendix A: Excel sheet for bending capacity
I
Appendix A
General input F
L 1200 mm S
b 1200 mm
h 150 mm
d0 127,3 mm
l/h 8 -
c (concrete cover) 27 mm q
F steel top 8 mm
F steel bottom 8 mm
number of bars top 13 -
number of bars bottom 13 -
d 115 mm L /2 L /2
fck 35 N/mm2
fbm 2,8 N/mm2
fc 1,4 N/mm2 qL 2 Shear failure is governing
W load , q  
f'c 21 N/mm2 4 Shear failure is NOT governing
T2 4,2 N/mm2 c s
FL 
Ec 31000 N/mm2 W load ,F   0,375 h d  0,625h
Es 210000 N/mm2 2
As,top 653,4512719 mm2 W  2 M p , sup port   2 M  d  0,625h 
int p , field  s  0,002625
As,bot 653,4512719 mm2 0,375h
fs 435 N/mm2 8M p , sup port 8M p , field
q max   AE A E ( d  0,625h )
S 1,00E+11 N/mm/mm width L 2
L 2
C s  s s s  0,002625. s c
wo 0,36% passes b 0,375bh
wmin 0,18%
4M p , sup port 4 M p , field
F max  
wmax 1,94% L L
Standaard virtual work theory Including compressive mebrane action (Park's theory) q q/qmax sagg/h
#DEEL/0! #DEEL/0! 0
Ns,sup 284251,3033 N B1 0,85 - 816,6282 2,369098 0,001667
xu,sup 15,0397515 mm T 236,88 N/mm width 837,1984 2,428774 0,003333
z,sup 109,1344969 mm T' 236,88 N/mm width 844,0778 2,448732 0,005
Mu,sup 31021622,98 Nmm Cs 113,40 N/mm width 845,8533 2,453882 0,006667
Ns,field 284251,3033 N Cs' 113,40 N/mm width 845,4427 2,452691 0,008333
xu,field 15,0397515 mm 844,0147 2,448549 0,01
z,field 109,1344969 mm 842,0917 2,44297 0,011667
Mu,field 31021622,98 Nmm 839,9261 2,436687 0,013333
qmax 344,7 kN/m maxium qpark/qmax 2,45 - 837,6468 2,430075 0,015
Fmax 207 kN 835,3215 2,423329 0,016667
qpark max 845,9 kN/m 832,9869 2,416556 0,018333
Punching sagg at qpark max 1 mm 830,6628 2,409814 0,02
wo,bottom 0,36% 828,3595 2,403132 0,021667
kd 1,431 - 826,0821 2,396525 0,023333
t1 1,12 N/mm2 823,8328 2,389999 0,025
P 761,2079 mm 821,6117 2,383556 0,026667
Fmax 98043,57752 N 819,4185 2,377193 0,028333
Fmax 98 kN 817,2519 2,370908 0,03
Fshear maple kN 815,1105 2,364695 0,031667
812,993 2,358552 0,033333
810,8977 2,352474 0,035
Shear (no shear reinforcement included) 808,8234 2,346456 0,036667
Md max 31021622,98 Nmm 806,7686 2,340495 0,038333
Vd max 206820 N 804,732 2,334586 0,04
λv 1,940869565 - 802,7124 2,328728 0,041667
A0 138000 mm2 800,7089 2,322915 0,043333
gλ 4,77 - 798,7203 2,317146 0,045
kλ 2,52 - 796,7458 2,311418 0,046667
kh 1,45 - 794,7844 2,305728 0,048333
ω0 0,36 % 792,8356 2,300074 0,05
t1,increased 1,46 N/mm2 790,8985 2,294454 0,051667
t1 (0,4fb) 0,56 N/mm2 788,9724 2,288867 0,053333
qmax 335 kN/m 787,057 2,28331 0,055
qmax (t2) 966 kN/m 785,1514 2,277782 0,056667
III
Appendix B: Maple sheet for punch capacity
V
Appendix B
The maple sheet in which the ultimate punching shear load is calculated looks as follows:
> restart;
This sheet works as follows:
(1) fill in the variables (blue fields) in N and mm
(2) run the sheet and fill in d1 as a smaller value than d1start
(3) run again
(4) change the value 'angle' to the value 'nextangle' (3 decimals)
(5) run again and repeat step (4) till dP is approximately zero
(5) write down the value of P (this value is shown in kN!)
(6) slightly decrease d1 and repeat step 3 till 6 untill a minimum is reached for P
(7) this minimum is the value P at which the plate will fail in shear mode (mind that d1 can reach d0 closely)
> L:=2250;
L := 2250
> h:=150;
h := 150
> d:=135;
d := 135
> d0:=300;
d0 := 300
> fcu:=35;
fcu := 35
> fy:=435;
fy := 435
> W0:=0.0075;
W0 := 0.0075
> fca:=0.85*fcu;
fca := 29.75
> fc:=0.85*fca;
fc := 25.2875
> fta:=0.7*(1.05+0.05*fcu);
fta := 1.960
> ft:=fc/400;
ft := 0.06321875000
> Ec:=evalf((4730*sqrt(fca)));
Ec := 25799.10415
> Es:=210000;
Es := 210000
> ck:=sqrt(1+fc/fta)-1;
ck := 2.728509851
VII
Maple sheet for punch capacity
> d1start:=solve((d1/d0)^d1=exp(ck*h), d1);

d1start := 596.0873953
> d1:=350;
d1 := 350
> angle:=4.988;
angle := 4.988
> beta:=convert(angle*degrees,radians);
b := 0.02771111111 p
> tb:=evalf(tan(beta));
tb := 0.08727762478
> a:=L/2;
a := 1125
> R:=L/2;
R := 1125
> As:=evalf(W0*L*d);
As := 2278.1250
> x:=1.76*d*W0*(fy/fcu);
x := 22.14771428
> S:=1/((a*R)/(0.8*Ec*L*d+Es*As)+a/(0.5*Ec*(x+h)));
S := 1440.547813
> phi:=(a*fc)/(2*h*S);:
f := 0.06582782200
> wi:=0.03*h;
wi := 4.50
> w0:=0.5*h;
w0 := 75.0
> N0:=0.5*h*fc-(As/L)*fy;
N0 := 1456.125000
> n0:=(N0/(h*fc));
n0 := 0.3838853188
> k:=(0.5*n0+0.25+0.25*phi-0.25*(wi/h))*exp(wi/(h*phi));
k := 0.7112180663
> na:=-1*(k*exp(-(w0/h)/phi)-0.5*(n0+0.5+0.5*phi)+0.25*(w0/h));
na := 0.3330421055
> B:= fsolve(d1/2=(d0/2+tan(beta)/B)*exp(B*h)-tan(beta)/B,B);

B := 0.0004890077716
VIII
Appendix B
> A:=d0/2+tb/B;
A := 328.4790137
> d0/2=A-tb/B;
150 = 150.0000000
> d1/2=A*exp(B*h)-tb/B;
175 = 175.0000000
> r:=A*exp(B*y)-tan(alpha)/B;
(0.0004890077716 y) 2044.957275 tan ( )
r := 328.4790137 e - a
> dr:=diff(r,y);
(0.0004890077716 y)
dr := 0.1606287905 e
> nr:=na+(w0/(2*h))*(1-(r/a));
(0.0004890077716 y)
nr := 0.5830421055 - 0.07299533640 e + 0.4544349500 tan (a)
> Nin:=int(nr*2*Pi*r,y=0..h);
5 5 5 2
Nin := 1.629455225 10 - 8.317644672 10 tan(a ) - 8.758446696 10 tan(a )
> Nrscont:=fc*Nin;
6 7 7 2
Nrscont := 4.120484900 10 - 2.103324396 10 tan (a ) - 2.214792208 10 tan(a )
> Nrs:=2*Pi*fc*(na+(w0/(2*h)))*((A/B)*(exp(B*h)-1)-(h/B)*tan(alpha))-
Pi*fc*(w0/(a*h))*((A^2/(B*2))*(exp(2*B*h)-1)-((2*A)/B^2)*(exp(B*h)-
1)*tan(alpha)+(h/B^2)*(tan(alpha))^2);
5
Nrs := 29.48735448 p (51123.93190 - 3.067435912 10 tan(a ))
7 8 8 2
- 0.01123888889 p (1.743218788 10 - 2.090925128 10 tan(a ) + 6.272775384 10 tan (a ) )
> diff(Nrs,alpha);
5 5 2 8
29.48735448 p (-3.067435912 10 - 3.067435912 10 tan(a ) ) - 0.01123888889 p (-2.090925128 10
8 2 9 2
- 2.090925128 10 tan(a ) + 1.254555077 10 tan(a ) (1 + tan(a ) ))
> F:=r*(dr-tb+(ck^2/4)*((1+dr*tb))^2/(dr-tb));
æ
(0.0004890077716 y) 2044.957275 tan( ) ç (0.0004890077716 y)
F := (328.4790137 e - a ) ç 0.1606287905 e
ç
è
2
(0.0004890077716 y) ö
1.861191502 (1 + 0.01401929931 e ) ÷
- 0.08727762478 + ÷
(0.0004890077716 y)
0.1606287905 e - 0.08727762478 ÷ø
> Pin:=int(F,y=0..h);
6 6
Pin := -7.440824738 10 tan(a ) + 1.238993219 10
IX
> P:=2*Pi*ft*((A^2/2)*(exp(2*B*h)-1)-((2*A)/B)*(exp(B*h)-
1)*tan(alpha)+(h/B)*(tan(alpha))^2)+Pi*ft*(ck^2/2)*(h/B+((2*A)/B)*(exp(B*h)-
1)*tan(alpha)+(A^2/2)*(exp(2*h*B)-1)*(tan(alpha))^2)+Nrs*tan(alpha);
5 5 2
P := 0.1264375000 p (8524.475350 - 1.022478638 10 tan(a ) + 3.067435912 10 tan(a ) )
5 5 2
+ 0.2353244005 p (3.067435912 10 + 1.022478638 10 tan(a ) + 8524.475350 tan(a ) ) + (29.487354\
5
48 p (51123.93190 - 3.067435912 10 tan(a ))
7 8 8 2
- 0.01123888889 p (1.743218788 10 - 2.090925128 10 tan(a ) + 6.272775384 10 tan(a ) )) tan (a )
> Pcontr:=evalf(2*Pi*ft*Pin+Nrs*tb);
6 5 6 2
Pcontr := -4.791339674 10 tan(a ) + 8.517729766 10 - 1.933018034 10 tan(a )
> dP:=(diff(P,alpha))/(1+(tan(alpha))^2);
1 5 5 2
dP := (0.1264375000 p (-1.022478638 10 - 1.022478638 10 tan (a)
2
1 + tan(a)
5 2
+ 6.134871824 10 tan(a ) (1 + tan(a ) ))
5 5 2 2
+ 0.2353244005 p (1.022478638 10 + 1.022478638 10 tan(a ) + 17048.95070 tan(a ) (1 + tan(a ) )) +
5 5 2 8
(29.48735448 p (-3.067435912 10 - 3.067435912 10 tan(a ) ) - 0.01123888889 p (-2.090925128 10
8 2 9 2
- 2.090925128 10 tan(a ) + 1.254555077 10 tan(a ) (1 + tan(a ) ))) tan(a ) + (29.48735448 p
5
(51123.93190 - 3.067435912 10 tan(a))
7 8 8 2
- 0.01123888889 p (1.743218788 10 - 2.090925128 10 tan(a ) + 6.272775384 10 tan (a ) )) (1
2
+ tan(a ) ))
> rad:=max(solve(dP=0,alpha));
rad := 0.08706146960
> nextagle:=evalf(convert(rad,degrees));
nextagle := 4.988254766 degrees
> alpha:=beta;
a := 0.02771111111 p
> dPcontr:=Pi*ft*(ck^2/2)*(((2*A)/B)*(exp(B*h)-1)+A^2*(exp(2*B*h)-1)*tan(alpha))-
2*Pi*ft*(((2*A)/B)*(exp(B*h)-1)-((2*h)/B)*tan(alpha))-
2*Pi*fc*(na+(w0/(2*h)))*(h/B)*tan(alpha)+Pi*fc*(w0/(a*h))*(((2*A)/B^2)*(exp(B*h)-1)-
((2*h)/B^2)*tan(alpha))*tan(alpha)+Nrs;
5
dPcontr := 0.2353244005 p (1.022478638 10 + 17048.95070 tan(0.02771111111 p ))
5 5
- 0.1264375000 p (1.022478638 10 - 6.134871824 10 tan(0.02771111111 p ))
6
- 9.045057010 10 p tan(0.02771111111 p )
X
Appendix B
8 9
+ 0.01123888889 p (2.090925128 10 - 1.254555077 10 tan(0.02771111111 p )) tan(0.02771111111 p )
5
+ 29.48735448 p (51123.93190 - 3.067435912 10 tan(0.02771111111 p )) - 0.01123888889 p
7 8 8 2
(1.743218788 10 - 2.090925128 10 tan(0.02771111111 p ) + 6.272775384 10 tan(0.02771111111 p ) )
> evalf(Pcontr)/1000;
418.8716901
> evalf(P)/1000;
418.8716902
> evalf(dP)/1000;
0.2392902344
> evalf(dPcontr)/1000;
0.2392895000
> evalf(Nrs)/1000;
2116.044104
> evalf(Nrscont)/1000;
2116.044103
XI
Appendix C: Some DIANA input files
XIII
Appendix C
The command file for the non-linear DIANA calculations
*FILOS
INITIA
*INPUT
READ
*NONLIN
BEGIN EXECUT
BEGIN LOAD
LOADNR=1
BEGIN STEPS
BEGIN ITERAT
ARCLEN
GAMMA=0.25
MAXSIZ=10
NSTEPS=500
END ITERAT
END STEPS
END LOAD
ITERAT MAXITE=500
END EXECUT
BEGIN OUTPUT
DISPLA TOTAL TRANSL
DISPLA INCREM TRANSL
FORCE REACTI TRANSL
STRESS TOTAL CAUCHY GLOBAL
STRESS TOTAL CAUCHY LOCAL
STRAIN CRACK
FORCE EXTERN TRANSL
END OUTPUT
*END
The material properties in the data file
'MATERIALS'
1 YOUNG 3.100000E+04
POISON 2.000000E-01
TOTCRK ROTATE
TENCRV BRITTL
TENSTR 1.400000E+00
COMCRV CONSTA
COMSTR 2.100000E+01
2 YOUNG 2.100000E+05
YIELD VMISES
YLDVAL 4.350000E+02
XV
Command file for a phased calculation
*FILOS
INITIA
*INPUT
*PHASE
BEGIN ACTIVE
ELEMEN 1 CIRCLE
END ACTIVE
*NONLIN
BEGIN EXECUT
BEGIN LOAD
LOADNR=2
BEGIN STEPS
EXPLIC SIZES 1 (1)
END STEPS
END LOAD
ITERAT MAXITE=500
ITERAT LINESE
END EXECUT
BEGIN OUTPUT FEMVIE BINARY

file="ph1"
DISPLA TOTAL TRANSL
FORCE REACTI TRANSL
STRAIN CRACK
FORCE EXTERN TRANSL
END OUTPUT
*PHASE
BEGIN ACTIVE
ELEMEN 1 CIRCLE 2 ELASTI
END ACTIVE
*NONLIN
BEGIN EXECUT
BEGIN LOAD
LOADNR=2
BEGIN STEPS
EXPLIC SIZES 1 (1)
END STEPS
END LOAD
ITERAT MAXITE=500
ITERAT LINESE
END EXECUT
BEGIN EXECUT
BEGIN LOAD
LOADNR=1
BEGIN STEPS
XVI
Appendix C
EXPLIC SIZES 0.5 (5000)

END STEPS
END LOAD
ITERAT MAXITE=500
ITERAT LINESE
END EXECUT
BEGIN OUTPUT TABULA

BEGIN SELECT
NODES 1106
END SELECT
displace
END OUTPUT
BEGIN OUTPUT FEMVIE BINARY

file="ph2"
DISPLA TOTAL TRANSL
FORCE REACTI TRANSL
STRAIN CRACK
FORCE EXTERN TRANSL
END OUTPUT
*END
XVII
Appendix D: Experiments based on the Ontario highway bridge de-
sign code
XIX
Appendix D
XXI
Experiments based on the Ontario highway bridge design code
XXII
Appendix D
XXIII
XXIV
Appendix D
XXV
XXVI
Appendix D
XXVII
XXVIII
Appendix D
XXIX
XXX
Appendix D
XXXI
XXXII
Appendix D
XXXIII
XXXIV
Appendix D
XXXV
XXXVI
Appendix D
XXXVII
XXXVIII
Appendix D
XXXIX
XL
Appendix D
XLI
XLII
Appendix D
XLIII
XLIV
Appendix D
XLV
XLVI
Appendix D
XLVII
XLVIII
Appendix D
XLIX
L
Appendix D
LI
LII
Appendix D
LIII
LIV
Appendix D
LV
LVI
Appendix D
LVII
LVIII
Appendix D
LIX
LX
Appendix D
LXI
LXII
Appendix D
LXIII
LXIV
Appendix D
LXV
LXVI
Appendix D
LXVII
LXVIII
Appendix D
LXIX
Appendix E: New Zealand code
LXXI
Appendix E
LXXIII
New Zealand code
LXXIV
Appendix E
LXXV
New Zealand code
LXXVI
Appendix E
LXXVII
New Zealand code
LXXVIII
Appendix E
LXXIX
New Zealand code
LXXX
Appendix E
LXXXI
New Zealand code
LXXXII
Appendix E
LXXXIII
New Zealand code
LXXXIV
Appendix E
LXXXV
New Zealand code
LXXXVI
Appendix E
LXXXVII
New Zealand code
LXXXVIII
Appendix E
LXXXIX
New Zealand code
XC
Appendix E
XCI
New Zealand code
XCII
Appendix E
XCIII
New Zealand code
XCIV
Appendix E
XCV
New Zealand code
XCVI
Appendix E
XCVII

Reportbakker PDF

Uploaded by

Copyright:

Available Formats

Reportbakker PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Reportbakker PDF

Uploaded by

Copyright:

Available Formats

What is the paper about?

What is the paper about?

What phenomenon is studied in the paper?

What phenomenon is studied in the paper?

A finite element model for the deck of plate-girder bridges in-

cluding compressive membrane action, which predicts

Gert Jan Bakker

Gert Jan Bakker

August 25, 2008

Gert Jan Bakker

Delft University of Technology

List of symbols ........................................................................................................................vii

2. Theory of compressive membrane action.........................................................................6

5. Finite element models...................................................................................................... 35

8. The deck of a common girder bridge ............................................................................. 81

Appendix A: Excel sheet for bending capacity ........................................................................ I

Appendix B: Maple sheet for punch capacity......................................................................... V

Appendix C: Some DIANA input files ................................................................................XIII

Appendix E: New Zealand code ....................................................................................... LXXI

A cross-sectional area [mm2]

 angle between yield surface and displacement rate vector [rad]

L change in length [mm]

Dutch bridges will not collapse yet1.1

De Volkskrant, 4 October 2007

Twelve steel bridges need urgent maintenance.

Figure 1.1: Typical Dutch bridge build-up

DAF stands for

q = 4 x DAF kN/m2 q = 2 2,5 kN/m2

q = 0,8 x 4 x DAF kN/m2

q = 0,8 x 4 x DAF kN/m2 rectangle 350 x 600 mm

Left over area

1.1. The goal of this study

This study will be build up of three parts:

Part I: The theory

Part II: The numerical development of the finite element model

 the creation of a numerical finite element model

Part III: A practical example

 the application of the in part II found model on a practical bridge deck

The study will finish with conclusions and recommendations.

2. THEORY OF COMPRESSIVE MEMBRANE ACTION

Figure 2.1: Compressive membrane action in a latterly restrained slab

2.1. Enhancement of the bending strength by compressive membrane forces

Figure 2.3: Starting point for Park’s compressive membrane theory

Assumptions in his model are

 the tension steel in the plastic hinges is yielding

Figure 2.4: Section 1-2 in the deformed state

L  L cos   L cos   0,5 (1  2  ) L  t

The second equation is formed using horizontal force equilibrium:

C ' s C ' c T '  C s  C c  T (3)

With this information equation (3) forms into:

With this figure, the quantities ns and ms can be calculated.

nu  C c  C s  T  0,85 f c' 1c  C s  T (7)

mu'  mu  nu (11)

variation in the length of the slab

0 0,02 0,04 0,06 0,08 0,1 0,12 0,14

variation in the thickness of the slab

predicted load by Park in kN/m

0 0,02 0,04 0,06 500

variation in the reinforcem ent percentage