Академический Документы
Профессиональный Документы
Культура Документы
Jean-Sébastien LEBEAU
In the last few years the number of piled raft foundations especially those with few piles, has
increased. Unlike the conventional piled foundation design in which the piles are designed to carry
the majority of the load, the design of a piled raft foundation allows the load to be shared between
the raft and piles and it is necessary to take the complex soil-struture interaction eects into account.
The aim of this paper is to describe a nite element analysis of deep foundations: piled and mainly
piled raft foundations. A basic parametric study is rstly presented to determine the inuence of
mesh discretisation, of materials - loose or dense sand -, of dilatancy and interface elements. Then
the behavior of piled raft foundations is analysed in more details using partial axisymmetric models
of one pile-raft.
We continue by preparing a more sophisticated 3D study to take into account the complex pile-
pile interaction which occured when the pile spacing is small. So the possibilies of employing the
embedded pile concept as implemented into Plaxis 3D foundations is investigated. Finally, some
Key words: Piled raft foundation, piles, embedded pile, volume pile, hardening soil
model
1
Acknowledgements
First of all I would like to express my gratefulness to Professor Helmut F. Schweiger for giving me
the opportunity to work on geotechnical issues at the Institute for Soil Mechanics and Foundation
This paper was made possible by the great contribution of my supervisor Dipl.-Ing Franz Tschuch-
nigg. I am indebted to him for his friendly supervision and guidance throughout the period of my
traineeship. I deeply thank him because he conveyed me a better understanding of nite element
I also would like to thank my French professor, Yvon Riou for getting me in touch with the Institute.
Finally, I would like to express my appreciation to all the people I met here who made my ve months
2
Contents
1 Introduction 6
2 Preliminary studies 7
2.1.1.1 Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1.1.4 Meshes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2 Pile-raft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2.1.1 Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2.1.4 Meshes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3
CONTENTS CONTENTS
3 Analysis of 2D models 24
3.1 Single-pile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.2 Pile-Raft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2.2 Variations of Skin friction and Normal Stresses along the pile . . . . . . . . . 29
3.2.5.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.1.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4
CONTENTS CONTENTS
4.2.1.5 Comparison of the three options: Linear, multilinear and layer de-
pendent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
5 Group eect 82
5.1.1 Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6 Conclusion 98
5
Chapter 1
Introduction
In traditional foundation design, it is customary to consider rst the use of shallow foundation such
deep foundation such as a fully piled foundation is used instead. In the last few decade, an alternative
solution has been designed: piled raft foundation. Unlike the conventional piled foundation design in
which the piles are designed to carry the majority of the load, the design of a piled raft foundation
allows the load to be shared between the raft and piles and it is necessary to take the complex
The concept of piled raft foundation was rstly proposed by Davis and Poulos in 1972 and is now
used extensively in Europe, particularly for supporting the load of high buildings or towers. The
favorable application of piled raft occurs when the raft has adequate loading capacities, but the
settlement or dierential settlement exceed allowable values. In this case, the primary purpose of
The aim of this paper is to describe a nite element analysis of deep foundations: piled and mainly
piled raft foundations. A basic parametric study is rstly presented to determine the inuence of
mesh discretisation, of materials - loose or dense sand -, of dilatancy and interface elements. Then
the behavior of piled raft foundations is analysed in more details using partial axisymmetric models
of one pile-raft.
We continue by preparing a more sophisticated 3D study to take into account the complex pile-
pile interaction which occured when the pile spacing is small. So the possibilies of employing the
embedded pile concept as implemented into Plaxis 3D foundations is investigated. Finally, some
6
Chapter 2
Preliminary studies
- 2D axisymmetric models -
In order to prepare a more sophisticated analysis a large number of calculations have been per-
formed in axisymmetric conditions. This approach oered the possibility to study with reasonable
calculation times the inuence of mesh discretisation, dilatancy and interface elements for a single
pile and a pile-raft. The dierent models and conclusions are presented in this part.
In order to analyze the behavior of the single pile, a model has been made in PLAXIS V8 using an
axisymmetric model. A working area of 20 m width and 40 m depth has been used. At the axis
of symmetry the pile has been modeled with a length of 15 m and a diameter of 0,8 m. The soil
is modeled as a single layer of sand with properties are described in 2.1.1.3). The ground water is
located at 40 m below the soil surface. In this way we did not take into account the water
inuence. Along the length of the pile an interface has been modeled. We extended this interface
to 0,5 m below the pile inside the soil body to prevent stress oscillation in this sti corner area.
1
We added two clusters close to the pile to enrich easily the mesh in this more moving area.
1
This longer interface will enhance the exibility of the nite element mesh in this area and will thus prevent
non-physical stress results. However, these elements should not introduce an unrealistic weakness in the soil according
to PLAXIS V8 manual.
7
2.1. SINGLE PILE CHAPTER 2. PRELIMINARY STUDIES
We used the standard xities PLAXIS tool to dene the boundaries conditions. Thus these bound-
Vertical geometry lines for which the x-coordinate is equal to the lowest or highest x-coordinate
Horizontal geometry lines for which the y-coordinate is equal to the lowest y-coordinate in the
Figure 2.1: Global geometry of the axisymmetric model of the single pile
The constitutive model used for the soil - sand - is the Hardening soil model. The main advantage
of this constitutive law is its ability to consider the stress path and its eect on the soil stiness and
its behavior. We used two dierent types of sand : one loose and the other dense. We also varied
8
2.1. SINGLE PILE CHAPTER 2. PRELIMINARY STUDIES
For the concrete pile, a linear elastic material set was applied.
The parameters of all this materials are summarized in the following table:
Permeability k 1 1 0 m/day
ref 3
E50 20 000 60 000 kN/m
ref 3
Stiness Eoed 20 000 60 000 3E7 kN/m
ref 3
Eur 1E5 1,8E5 kN/m
Power m 0,65 0,55
Poisson ratio νur 0,2 0,2 0,2 -
Dilatancy y 2/0 8/0 °
Friction angle f 32 38 °
Cohesion cref 0,1 0,1 kN/m
2
2.1.1.4 Meshes
To study the mesh dependency 3 analyses were performed: one with a coarse, one with a medium
and one with a very ne mesh. For each one we considered 6 models varying the interface elements.
2
This factor relates the interface strength (wall friction and adhesion) to the soil strength (friction angle and
cohesion)
9
2.1. SINGLE PILE CHAPTER 2. PRELIMINARY STUDIES
Figure 2.2: A very ne mesh for calculations with interface elements
To assign a load at the top of the pile we considered two approaches: one with prescribed displace-
With prescribed displacement we impose a certain displacement at the top of the pile whereas with
2.1.2 Results
Remark: All the following curves are plotted for the node point located at the top right side of
the pile.
10
2.1. SINGLE PILE CHAPTER 2. PRELIMINARY STUDIES
By analysing all the calculations made, we can conclude that for each material - loose or dense sand
- the curves have the same shapes for calculations performed with coarse, medium and
very ne mesh. Nevertheless, we can observe that with ner meshes, we have unphysical premature
soil body collapsing. The following gure illustrates this conclusion with some examples.
11
2.1. SINGLE PILE CHAPTER 2. PRELIMINARY STUDIES
Figure 2.4: Mesh dependency for the loose sand - ψ =2° - and dierent values for Rinter
To avoid this premature failure we decided to restart the medium and very ne calculations switching
o the arc length control procedure. But we now observed convergence problems with more or less
important oscillations.
These oscillations occurred for important displacements (from 20 cm) whatever the material, mesh
or Rinter value. However, the global shape of the load-displacement curve seems to stay realistic
12
2.1. SINGLE PILE CHAPTER 2. PRELIMINARY STUDIES
Figure 2.5: Mesh dependency for the loose sand - ψ =2° - and dierent values for Rinter
Figure 2.6: Inuence of arc length control for the loose sand - ψ =2° - and Rinter =0,7/0,1
parameters: mesh1= coarse, arc length=ON; mesh2= medium, arc length=OFF; mesh3= Very ne, arc
length=OFF;
13
2.1. SINGLE PILE CHAPTER 2. PRELIMINARY STUDIES
The gure 2.6 enables us to conrm that the mesh dependency is negligible for this model.
The previous paragraph was based on les with the load approach. We did the same calculations with
the displacement approach. By comparing these two approaches we can conclude that the shape
of the load-settlement curves is exactly the same in each case . Moreover, there are less
oscillations with prescribed displacemen t than with distributed loads. There are no stairs' ' even
with an important displacement. So because it limits this problem of big oscillations, prescribed
displacement seems to be better to study a single pile. The following picture illustrates
these conclusions.
Figure 2.7: Distributed loads and prescribed displacement, comparison for the loose sand - ψ =0°
- and Rinter =0,1/0,4/0,7
14
2.1. SINGLE PILE CHAPTER 2. PRELIMINARY STUDIES
Remark: The gure 2.7 shows us the settlement with the load [kN] whereas in the previous section
2
we plotted the settlement with the distributed load [kN/m ]. To compare the two approaches we have
to:
2
Distributed load: Multiply the distributed load [kN/m ] by the area of the pile to get the total
force (Rtot ) [kN].
Prescribed displacement: Read out the force value in Plaxis output [kN/rad] and multiply it
by 2 π to get the total force (Rtot ) [kN]
Now we can try to interpret the stairs of the load approach by comparing with the same calculations
Figure 2.8: Comparison Displ and Load approaches for the loose sand ψ =0° - No interface
medium mesh
15
2.1. SINGLE PILE CHAPTER 2. PRELIMINARY STUDIES
Figure 2.9: Comparison Displ and Load approaches for the loose sand ψ =0° - R=1 medium mesh
As we can see on gures 2.8 and 2.9, it is impossible to deduce correctly the normal shape from the
distributed load curves by interpreting the oscillations. Sometimes, the prescribed displacement
curve is under the distributed load one, sometimes it is in the middle. So we have to interpret
with caution the shape of the stairs part of the distributed load curves.
We varied the way to model the pile to sand interface by changing the Rinter value and doing a model
without interface. We also performed one calculation by drawing an interface in Plaxis input and
We can conclude that the choice of the value for Rinter is not negligible when you model
a single pile. As we can see in the table, for the same load, the settlements increase by more than
40 % between R=0,4 and 0,1, 80% between R=0,7 and 0,4 and 600 % between R=1 and 0,7 for
loose sand.
Table 2.3: Settlements of the single pile for loose sand, ψ =2° and Rtot =2000 kN
16
2.1. SINGLE PILE CHAPTER 2. PRELIMINARY STUDIES
As we could expect the load-displacement curves have almost the same shape for models with R=1
and without interface. We also noticed that unselecting the interface lead to false results with
premature failure (see red following curve) or an unrealistic behavior. So the interface drawn in
the Plaxis input must be selected in the calculations steps. The following curves sum up all these
conclusions.
Figure 2.10: Load-settlement curves for Loose sand, ψ =2° and coarse mesh
les : Geo2Load_Mesh1_Loose_Rinter0,1/0,4/0,7/1/Unselected/No_Psi=2_HS.plx
We tested two values of dilatancy ψ for both materials: (ϕ − 30) and 0°. As expected, we have less
17
2.2. PILE-RAFT CHAPTER 2. PRELIMINARY STUDIES
les : Geo2Load_Mesh2_Dense_Rinter0,4/0,7_Psi8/0_HS_ALCo.plx
2.2 Pile-raft
We performed the same calculations as we have done with the single pile model using an axisym-
As we did for the single pile, the pile has been modeled with a length of 15 m and a diameter of
0,8 m at the axis of symmetry. We added a slab in concrete with a thickness of 0,5 m. The soil is
also modeled as a single layer of sand with the same properties as the single pile. The ground water
is located at 40 m below the soil surface. In this way we did not take into account the water
inuence. Along the length of the pile an interface has been modeled. We extended this interface
to 0,5 m below the pile inside the soil body to prevent stress oscillation in this sti corner area.
18
2.2. PILE-RAFT CHAPTER 2. PRELIMINARY STUDIES
We also used for this study the standard xities PLAXIS tool (see 2.1.1.2).
The parameters of all the materials are recalled in the following table:
Permeability k 1 1 0 m/day
ref 3
E50 20 000 60 000 kN/m
ref 3
Stiness Eoed 20 000 60 000 3E7 kN/m
ref 3
Eur 1E5 1,8E5 kN/m
Power m 0,65 0,55
Poisson ratio νur 0,2 0,2 0,2 -
Dilatancy y 2/0 8/0 °
Friction angle f 32 38 °
Cohesion cref 0,1 0,1 kN/m
2
2.2.1.4 Meshes
To study the mesh dependency 3 analysis were also performed: one with a coarse, one with a medium
and one with a very ne mesh. For each one we considered 6 models varying the interface elements.
Thus we varied the Rinter coecient from 0,1 to 1. We also performed one batch of calculation with
19
2.2. PILE-RAFT CHAPTER 2. PRELIMINARY STUDIES
To assign a load at the top of the slab we considered in this case only a distributed load
Figure 2.12: Details about a pile-raft geometry with the axisymmetric model, Very ne mesh
2.2.2 Results
Remark: All the following curves are plotted for the node point A, situated at the top right side
of the pile, under the slab (see gure 2.12).
By analysing all the calculations made, we can conclude that for each material - loose or dense
sand - the curves have exactly the same shapes for calculations performed with coarse,
medium and very ne mesh.
20
2.2. PILE-RAFT CHAPTER 2. PRELIMINARY STUDIES
Figure 2.13: Example, Mesh dependency for the pile raft model with loose sand, ψ =2°, Rinter =0,7
les : Geo1_Mesh1/2/3_loose_Rinter0,7_Psi2_HS.plx
We varied the way to model the pile to sand interface by changing the Rinter value and doing a model
without interface.
As we can see in the table and on the following curve the way you model the interface has a
negligible inuence on the settlements.
Table 2.6: Settlements with dierent values of Rinter for load=1000 kN/m
2
21
2.2. PILE-RAFT CHAPTER 2. PRELIMINARY STUDIES
Table 2.7: Load-settlement curves for Loose sand, ψ =2° and coarse mesh
les : Geo1_Mesh1_loose_Rinter0,1/0,4/0,7/1/NO_Psi2_HS.plx
We tested two values of dilatancy (ψ ) for both materials: 2° and 0° for the loose sand, 8° and 0° for
the dense sand. We can conclude that the inuence of the dilatancy is negligible for this
model even for the dense sand.
Table 2.8: Settlements with dierent values of Rinter for load=1000 kN/m
2
22
2.2. PILE-RAFT CHAPTER 2. PRELIMINARY STUDIES
les : Geo1_Mesh1_dense_Rinter0,4/0,7_Psi8/0_HS.plx
23
Chapter 3
Analysis of 2D models
model of a single pile and a pile-raft. Now we present other calculations performed by taking these
In design of piled rafts, design engineers have to understand the mechanism of load transfer from
the raft to the piles and to the soil. It requires to take complex interactions into account such as:
The aim of this chapter is to have a better understanding of the pile and raft behavior and to check
the ability of the software to model such complex interactions. In this part, we only modeled a single
pile with a raft so we did not take into account the pile-pile interaction.
3.1 Single-pile
In the previous calculations we simulated an axial load test on a bored pile. We get the following
load-displacement curve:
24
3.1. SINGLE-PILE CHAPTER 3. ANALYSIS OF 2D MODELS
le : Geo2Disp_Mesh2_loose_R=0,7_Psi2_HS.plx
Now we observe the mobilisation of the skin friction (qs ) with dierent loads.
Figure 3.2: Evolution of the Skin friction with the load (Rtot)
le : Geo2Disp_Mesh2_loose_R=0,7_Psi2_HS.plx
25
3.2. PILE-RAFT CHAPTER 3. ANALYSIS OF 2D MODELS
Table 3.1: Evolution of the skin and base resistance with settlements
le : Geo2Disp_Mesh2_loose_R=0,7_Psi2_HS.plx
That shows that the maximum skin friction is already reacted when 1,0 cm settlements occur (see
3.2 Pile-Raft
Key questions that arise in the design of piled rafts concern the relative proportion of load carried
by raft and piles. It depends on the geometric parameters of the pile and of the raft.
We performed four new models based on the rst geometry described in chapter 2 to interpret the
1
The Pile-Raft I is the geometry described in details in the chapter 2
26
3.2. PILE-RAFT CHAPTER 3. ANALYSIS OF 2D MODELS
27
3.2. PILE-RAFT CHAPTER 3. ANALYSIS OF 2D MODELS
2
We tested all these geometries with the materials loose and dense sand , with and without dilatancy
and varying the value of Rinter . The outcome was that the inuence of dilatancy and of
Rinter is very limited. We also performed these calculations with 3 dierent meshes to conrm
that there is no mesh dependency. We tryed to have next the pile the same mesh coarseness in
2
See table n°2.4
28
3.2. PILE-RAFT CHAPTER 3. ANALYSIS OF 2D MODELS
each model in order to compare precisely the dierent models. The load is a distributed load applied
on the slab and the boundaries conditions are those described in chapter 2. In this study we did
not take into account the ground water.
Remark: As we did in chapter 2, all the load-displacement curves are plotted for the node point
A, situated at the top right side of the pile, under the slab (see gure 2.12).
completely dierent.
Figure 3.6: Load settlement curve for pile and pile-raft foundation
les : Geo1/1Bis/2load_Mesh1_loose_R=0,7_Psi2_HS.plx
3.2.2 Variations of Skin friction and Normal Stresses along the pile
For each model we plotted the skin friction and the normal stresses along the pile. This procedure
gave us the possibility to illustrate how the load transfer works when the load increases. All the
29
3.2. PILE-RAFT CHAPTER 3. ANALYSIS OF 2D MODELS
Remark: All these gures are plotted by selecting the interface in the Plaxis output. In order to
get something comparable from one model to an other, we subtracted the rst phase with the pile-
activation for each load steps plotted. Thus the Skin friction or Normal Stresses that we present
here are only due to the load and the weight of slab .
dpile
Figure 3.7: Evolution of the skin friction with the load for the Pile-raft I (
lraf t = 2, 5)
le : Geo1_Mesh1_Dense_R=0,7_Psi2_HS.plx
30
3.2. PILE-RAFT CHAPTER 3. ANALYSIS OF 2D MODELS
dpile
Figure 3.8: Evolution of the skin friction with the load for the Pile-raft II (
lraf t = 6, 25)
le : Geo1Bis_Mesh1_Dense_R=0,7_Psi2_HS.plx
On the previous gures we can easily see that the mobilization of skin friction of a pile in a piled-raft
foundation is completely dierent from the one of with a single pile. For the model Pile-Raft II with
d pile
a big spacing ( lraf t
= 6, 25), the slab has a strong inuence on the shear stress distribution along the
pile. We notice an increase of shear stresses at the top of the pile, just under the slab. In this case,
the slab increases locally the normal stress, so the shear stresses increase in this area provoking this
For the model Pile-Raft I, the slab does not participate to the load transmission because we do not
dpile
see such a peak in the distribution: The spacing (
lraf t = 2, 5) is too small and almost all the load
goes to the pile. Nevertheless, the slab has an inuence too because the distribution is dierent from
the one for the single pile. There is an important mobilization of skin friction in the lower part of
As we can see on the following curves the shape of the normal stresses is in compliance with these
31
3.2. PILE-RAFT CHAPTER 3. ANALYSIS OF 2D MODELS
dpile
Figure 3.9: Evolution of the normal stresses with the load for the Pile-raft I (
lraf t = 2, 5)
le : Geo1_Mesh1_Dense_R=0,7_Psi2_HS.plx
dpile
Figure 3.10: Evolution of the normal stresses with the load for the Pile-raft II ( l = 6, 25)
raf t
le : Geo1Bis_Mesh1_Dense_R=0,7_Psi2_HS.plx
32
3.2. PILE-RAFT CHAPTER 3. ANALYSIS OF 2D MODELS
We now plotted the Skin friction for the second set of calculation.
These curves plotted for the geometries with a 30 m length pile and a 1,5 m diameter conrme theses
comments.
dpile
Figure 3.11: Evolution of the skin friction with the load for the Pile-raft V (
lraf t = 3)
le : Geo1Cinq_Mesh1_Dense_R=0,7_Psi2_HS.plx
33
3.2. PILE-RAFT CHAPTER 3. ANALYSIS OF 2D MODELS
dpile
Figure 3.12: Evolution of the skin friction with the load for the Pile-raft III (
lraf t = 6)
le : Geo1Ter_Mesh1_Dense_R=0,7_Psi2_HS.plx
dpile
Figure 3.13: Evolution of the skin friction with the load for the Pile-raft IV (
lraf t = 12)
le : Geo1Quater_Mesh1_Dense_R=0,7_Psi2_HS.plx
34
3.2. PILE-RAFT CHAPTER 3. ANALYSIS OF 2D MODELS
For the model Pile-raft IV - biggest spacing - with a 1000 kN/m2 loading (gure 3-12), there is
positive shear stresses on some centimeters in the top part of the pile . This eect should be studied
in further research.
By plotting the same curves for the dierent materials -loose and dense sand- and dierent values
for ψ we can conclude both dilatancy and materials have very few inuence on the normal stresses
les : Geo1Bis_Mesh1_Dense_R=0,7_Psi0/8_HS.plx
35
3.2. PILE-RAFT CHAPTER 3. ANALYSIS OF 2D MODELS
Figure 3.15: Evolution of the skin friction with dense or loose sand
les : Geo1Bis_Mesh1_Dense/Loose_R=0,7_Psi0_HS.plx
The previous curves in the last section let us understood some aspects of the behaviour of a piled-
raft foundation. We easily saw that the bigger the spacing is the more the raft acts in the load
transmission. We are now going to describe these observations in a more precise way by calculating
In Austria and Germany a common approach consists in calculating the αKpp 4 factor.
The αKpp factor is the ratio between the load carried by the pile and the total load applied on the
piled raft foundation.Thus it gives us a precise idea of the proportion of load carried by the pile and
by the raft.
Rpile
αKpp = Rtot
with:
4
In English, Kpp (Kombinierte-Pfahl-Plattengründung) means piled-raft-foundation
36
3.2. PILE-RAFT CHAPTER 3. ANALYSIS OF 2D MODELS
Rtot =Total load =Distributed load on the slab + weigth of the slab = Rraf t + Rpile 6 [kN]
So it means that:
We will also use the (1-αKpp ) coecient which represents the proportion of load carried by the raft.
Rraf t
(1-αKpp )= R
tot
Remark: Again the weight of the pile is not taken into account.
The simplest way to calculate αKpp with Plaxis 2D consists in realizing a cross section under the
slab and reading out the normal stresses on this cross section. Then we just have to sort the normal
stresses which are into the pile and into the soil.
Remarks:
In order to get an accurate value for αKpp we need to take care of:
Making a cross section which crosses as much stress points as possible because the value is
Making a cross section not too close to the slab because the junction Slab/pile is a high stress
variation area and singularities could occur (take 10 cm to 20 cm under the slab usually leads
to accurate values).
5
Rb =Base resistance of the pile [kN]; Rs =Skin resistance of the pile [kN]
6
Rraf t =Load carried by the raft [kN]
37
3.2. PILE-RAFT CHAPTER 3. ANALYSIS OF 2D MODELS
Example: calculation of αKpp for Pile-Raft I, Load=1000 kN/m2 , Dense sand, mesh medium,Ψ=8°:
In this case, we have Rtot = 1000.area + weigth of the slab = 3180 kN/m
2
We rst made the cross section n°1 just under the slab. We get the normal stresses as we can see
38
3.2. PILE-RAFT CHAPTER 3. ANALYSIS OF 2D MODELS
From the values of this prole we calculated Rpile and Rtot . We found: Rpile = 3687 kN and
Rtot = 3704kN , thus there is an error of 16 % for Rtot . In this way we overestimate Rpile and Rtot
because of the unrealistic high normal stress value at the interface.
So we started again with the cross section n°2. This one is not directly under the slab, thus we
avoid the singular area. Moreover the soil weigth added is negligible in comparison with the load.
Here we calculate: Rpile = 3127 kN and Rtot = 3151kN. There is an error of only 1 % for Rtot .
W idthraf t
With a small spacing (
Diameterpile =2,5 or 3) it seems that the raft takes a small part of the load. In
these cases, we calculated an αKpp equal to 0,99 for all load.
W idthraf t
With a bigger spacing (
Diameterpile =6; 6,25 or 12), we can notice that:
The stress repartition between the raft and the pile evolves with the loading. The higher the
loading is, the more the stress is shared. With a load between 0 and 200 kN/m
2 everything
2
goes mostly to the pile (1 <αKpp < 0,8). From 200 kN/m the raft has a stronger inuence.
39
3.2. PILE-RAFT CHAPTER 3. ANALYSIS OF 2D MODELS
The bigger the spacing is, the more load the raft takes.
In each case the curves converge to an equilibrium state, around αKpp =0,65 for Pile-Raft III,
The pile obviously carries more load by increasing the length of the pile (compare the geome-
Figure 3.19: Inuence of geometry on αKpp for loose sand, ψ =2°, R=0,7, mesh medium.
W idthraf t
Name Lengthpile Diameterpile Widthraf t
Diameterpile
Pile-Raft I 15 m 0,8 m 2 m 2,5
Pile-Raft II 15 m 0,8 m 5 m 6,25
Pile-Raft V 30 m 1,5 m 4,5 m 3
Pile-Raft III 30 m 1,5 m 9 m 6
Pile-Raft IV 30 m 1,5 m 18 m 12
40
3.2. PILE-RAFT CHAPTER 3. ANALYSIS OF 2D MODELS
Rtot 2
[kN/m ] 25 + Slab 250 + Slab 500 + Slab 1000 + Slab
Table 3.5: Few values of αKpp for loose sand, ψ =2°, R=0,7
As we can see on the following curves, the material - loose or dense dand - and the dilatancy have
41
3.2. PILE-RAFT CHAPTER 3. ANALYSIS OF 2D MODELS
In this sub-section we present the evolution of αKpp with the load (gure 3.19) and the displacement
(gure 3.20) for dierent Rinter values. In both cases, the tendency is exactly the same. We only
added with the displacement because it is also a common presentation in the literature.
Concerning the inuence of Rinter on αKpp we can conclude that the part of the load carried by
the pile decreases when we reduce the interface strength factor. It is an expected behavior because
by reducing the Rinter value we decrease the maximum amount of mobilization of the skin friction
7
On the interface, τ ≤ Rinter .(σ n .tanφsoil +csoil )
42
3.2. PILE-RAFT CHAPTER 3. ANALYSIS OF 2D MODELS
Figure 3.22: Inuence of Rinter on the evolution of αKpp with the load, Pile-raft II, R=0,7
Figure 3.23: Inuence of Rinter on the evolution of αKpp with the displacement, Pile-raft II, R=0,7
43
3.2. PILE-RAFT CHAPTER 3. ANALYSIS OF 2D MODELS
resting to compare the settlements with and without a pile. So, we performed one new calculation
for each geometry putting just the raft without the pile. Then we calculated the β coecient.
Denition
β is the ratio between the settlements which occured without pile (Uraf t ) and with the settlements
β = Upile+raf
Uraf t
t
As expected, the evolution of β with the load has the same tendency as αKpp . When we have a
high value for αKpp the pile carries most of the load and thus acts a lot against displacements. So
On the contrary, when the raft carries a big part of the load - for example with Pile-Raft IV - the
For the model Pile-Raft II in which we have a good sharing of the load, we have a β value from 1,15
to 2,1.
Figure 3.24: Evolution of β with the load, Loose sand Ψ=2° - Mesh Medium Rinter =0,7
44
3.2. PILE-RAFT CHAPTER 3. ANALYSIS OF 2D MODELS
Rtot 2
[kN/m ] 25 200 500 1000
Figure 3.25: Comparison between the evolution of β and α for Pile-Raft II, loose, ψ =2°, R=0,7
we study in detail these two forces in order to have a better idea of the pile behavior for dierent
geometries.
The method to calculate Rb is the same as for Rtot . We made a cross section under the pile. In this
part, we considered only the contribution of the distributed load by subtracting the two
rst phase with the pile and raft activation.
On the next gure, we can see the evolution of the base resistance with the load for the model
Pile-Raft I and II. The curves are both approximatively linear. It means that in our cases the part
of the total load (Rtot ) carried by the base of the pile is approximatly constant.
45
3.2. PILE-RAFT CHAPTER 3. ANALYSIS OF 2D MODELS
Figure 3.26: Evolution of Rb with the load for Pile-Raft I and II, dense sand, ψ =8°
Now we compare the Rbase for the pile-raft I and the single-pile.
Figure 3.27: Comparison of Pile-Raft I and Single Pile, evolution of Rb with the load for Pile-Raft I
and II, dense sand, ψ =8°
The skin friction proles presented previously give us the possibility to work out the skin resistance
Rs . As we did for Rb we only considered in this section the contribution of the distributed load by
46
3.2. PILE-RAFT CHAPTER 3. ANALYSIS OF 2D MODELS
Figure 3.28: Evolution of Rs with the load for Pile-Raft I and II, dense sand, ψ =8°
3.2.5.3 Conclusions
The following curves sum up the Rbase , Rskin and Rraf t proportions for various models.
Figure 3.29: Repartition of the forces into the single-pile, dense sand, ψ =8°
47
3.2. PILE-RAFT CHAPTER 3. ANALYSIS OF 2D MODELS
Figure 3.30: Repartition of the forces into pile for Pile-Raft I, dense sand, ψ =8°
Figure 3.31: Repartition of the forces into pile for Pile-Raft II, dense sand, ψ =8°
48
Chapter 4
behavior of piles in a pile group is usually observed to be totally dierent from the behavior of
a corresponding single pile. This group eect cannot be studied with axisymmetric models and
In order to prepare the group eect analysis, we rstly tested the dierent Plaxis 3D foundation
tools to model a pile: the volume pile and a new feature, the embedded pile. These comparisons are
The previous calculations with axisymmetric models showed a negligible mesh dependency. We also
checked that 6-node coarse meshes lead to the same load-displacement behavior as 15-node ne
meshes.
Due to the bigger size of working areas in 3D models we cannot use eciently ne meshes. Thus,
we will perform calculations from coarse to medium meshes. The results should be realistic because
To create a mesh with Plaxis 3D foundation we rstly generate a 2D mesh on a horizontal work
plane. When the 2D mesh is satisfactory, the 3D mesh is generated from the 2D mesh. Since
there is no vertical renement option, badly shaped elements with a higher vertical than horizontal
dimension could occur. To get a satisfactory vertical renement, we added multiple work planes in
the input, then when the 3D mesh is generated from the 2D one, these additional planes are taken
into account and the vertical size of the elements is adapted from their spacing. In this way we get a
good medium 3D mesh with a local 3D renement under the slab and at the pile bottom (see gure
4.1).
49
4.1. VOLUME PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
3D foundation. The working area was adapted in each case to have the same raft area with 3D and
Actually the raft area with axisymmetric models is circular whereas it is a square raft in 3D. Thus
In this way, the area of the 3D raft is equal to the one in 2D: Araf t3D = Araf t2D 2
[m ]
The pile is modeled as a volume pile and we selected the massive circular pile type. Interfaces are
modeled along the pile with a Rinter = 0, 7. The soil consists of a single layer of dense sand with
the same properties as the sand we used previously. The load is modeled as a distributed load on the
slab. Two dierent meshes with dierent levels of renement were applied to the rst two geometries.
Only a medium one was used for the remaining geometries. The following tables and gures sum
W idthraf t
Name Thicknessslab Depthmodel Lengthpile Diameterpile Widthraf t
Diameterpile
Pile-Raft I 0,5 m 40 m 15 m 0,8 m 1,8 m 2,25
Pile-Raft II 0,5 m 40 m 15 m 0,8 m 4,4 m 5,5
Pile-Raft V 1 m 60 m 30 m 1,5 m 4 m 2,7
Pile-Raft III 1 m 60 m 30 m 1,5 m 8 m 5,3
Pile-Raft IV 1 m 60 m 30 m 1,5 m 16 m 10,6
50
4.1. VOLUME PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
Permeability k 1 0 m/day
ref 3
E50 60 000 kN/m
ref 3
Stiness Eoed 60 000 3E7 kN/m
ref 3
Eur 1,8E5 kN/m
Power m 0,55
Poisson ratio νur 0,2 0,2 -
Dilatancy y 8 °
Friction angle f 38 °
Cohesion cref 0,1 kN/m
2
Figure 4.2: Details about a pile-raft geometry in 3D, medium mesh (Pile-raft IV)
51
4.1. VOLUME PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
Medium Fine
Pile-Raft I 12 610 31 290
Pile-Raft II 22 230 31 464
Pile-Raft V 17 574 /
Pile-Raft III 22 134 /
Pile-Raft IV 24 186 /
4.1.2 Results
Remark: As we did for axisymmetric models all the following load-settlement curves are plotted
for the node point located at the top right side of the pile, under the slab.
We plotted the load-displacement curve for each geometry. Then we compared these curves with
the associated axisymmetric curves. In each case, we noticed a good match with the 3D volume
Moreover the gures 4.3, 4.4 and 4.5 conrm that the mesh dependency is negligible.
52
4.1. VOLUME PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
Figure 4.4: Load-displacement curves comparison for Pile-Raft I, dense sand, ψ =8°
Figure 4.5: Load-displacement curves comparison for Pile-Raft II, dense sand, ψ =8°
53
4.1. VOLUME PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
Figure 4.6: Load-displacement curves comparison for Pile-Raft IV, dense sand, ψ =8°
Remarks:
All the following gures are plotted by selecting the interface in Plaxis output. For Pile-Raft I
and II (respectively for Pile-Raft V and IV) we plotted the interface along the line - X = 0, 4
(resp. 0, 75); Y ∈ [−15;0] (respec. [−30; 0]); Z = 0 (resp. Z = 0 ) -. In order to get something
comparable from one model to an other, we subtracted the rst phase from the pile for each
load steps plotted. Thus the skin friction presented in this part are only due to the load and
Then, we compared the 3D volume pile proles with the axisymmetric ones. They are not
strictly comparable because the shape of the raft area is not the same. Nevertheless a com-
parison stays relevant as we choose the same area for every models.
Results of 3D volume pile models are in a very good aggreement with those we got with axisymmetric
calculations. We observed almost the same shape of skin friction for each Pile-Raft le.
54
4.1. VOLUME PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
Figure 4.7: Axisymmetric and 3D volume pile skin friction curves, Pile-Raft I, Dense, ψ = 8,
Rinter = 0, 7
Figure 4.8: Axisymmetric and 3D volume pile skin friction curves, Pile-Raft II, Dense, ψ = 8,
Rinter = 0, 7
55
4.1. VOLUME PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
Figure 4.9: Axisymmetric and 3D volume pile skin friction curves, Pile-Raft V, Dense, ψ = 8,
Rinter = 0, 7
Figure 4.10: Axisymmetric and 3D volume pile skin friction curves, Pile-Raft III, Dense, ψ = 8,
Rinter = 0, 7
We can also notice that there are more oscillations in the lowest part of pile with the 3D volume
pile models than with 2D axisymmetric models. These non-physical stress oscillations are due to
56
4.1. VOLUME PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
the high peaks in stresses at the bottom of the pile. As we can see on gure 4.11, we can reduce
these numerical inaccuracies by lengthening the interface at the bottom of the pile (+0,5 m).
Figure 4.11: Reduction of oscillations by lengthening the interface, Pile-raft III, Dense, ψ = 8,
Rinter = 0, 7
By analysing in more details the load repartion for each model we can conclude that not only the
57
4.1. VOLUME PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
Remarks: The following values have been calculated without subtracting the weight of the pile
and of the raft. For the volume pile, we estimated Rbase and Rskin by reading in Plaxis output the
normal force values N at the top and at the bottom of the pile. Then we considered that: Rbase =
We also varied the value of Rinter and ψ with some 3D volume pile models. We can conclude that
Figure 4.12: Load-displacement curves for Pile-Raft I for dierent values of Rinter , dense sand, ψ =8°
58
4.1. VOLUME PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
Figure 4.13: Load-displacement curves for Pile-Raft III for dierent values of ψ , dense sand, Rinter =
0, 7
Figure 4.14: Skin friction with the load for ψ=8 and 0°, Pile-Raft III, Dense sand, Rinter = 0, 7
59
4.2. EMBEDDED PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
An embedded pile is a pile composed of beam elements that can be placed in arbitrary direction
in the sub-soil (irrespective from the alignment of soil volume elements) and that interacts with
the sub-soil by means of special interface elements. The interaction may involve a skin resistance
as well as a foot resistance. Although an embedded pile does not occupy volume, a particular
volume around the pile (elastic zone) is assumed in which plastic soil behaviour is excluded. The
size of this zone is based on the (equivalent) pile diameter according to the corresponding embedded
pile material data set. This makes the pile almost behave like a volume pile. Nevertheless, when
creating embedded piles no corresponding geometry points are created. Thus, contrary to volume
pile, embedded piles do not inuence the nite element mesh as generated from the geometry model.
In contrast to what is common in the Finite Element Method, the bearing capacity of an embedded
pile is considered to be an input parameter rather than the result of the nite element calculation.
Plaxis gives us the possibility to enter the skin resistance prole in three ways:
Linear: The user enters the skin resistance at the pile top and the skin resistance at the pile
bottom. The skin resistance is dened as linear along the pile. This way of dening the pile
Multi-linear: The skin resistance is dened in a table at dierent positions along the pile.
Multi-linear can be used to take into account inhomogeneous or multiple soil layers with
Layer dependent, can be used to relate the local skin resistance to the strength properties of
the soil layer in which the pile is located, and the interface strength reduction factor Rinter ,
as dened in the material data set on the corresponding soil layer. Using this approach the
pile bearing capacity is based on the stress state in the soil, and thus unknown at the start of
We performed another set of calculations by modeling the previous geometries using embedded
piles. This study gave us the possibility to test the reliability of this new feature to model pile-raft
structures.
Raft II. We took exactly the same geometries by using embedded piles instead of volume piles. We
1
See Plaxis manual for more details about embedded piles
60
4.2. EMBEDDED PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
considered the pile to raft connection as rigid. As mentioned previously the capacity of the pile is an
input parameter for an embedded pile so we had to dene the most relevant skin friction distribution
and base resistance. This is the reason why we tested each possibility oered by Plaxis to try to
The parameters we examined for the embedded pile are the same as those desribed previously for
the volume pile. We performed calculations for only one material, the dense sand with Rinter = 0, 7.
The load is modeled as a distributed load on the slab. Two dierent meshes with dierent levels of
renement had been used. The following tables sum up the most important parameters.
W idthraf t
Name Thicknessslab Depthmodel Lengthpile Diameterpile Widthraf t
Diameterpile
Pile-Raft I 0,5 m 40 m 15 m 0,8 m 1,8 m 2,25
Pile-Raft II 0,5 m 40 m 15 m 0,8 m 4,4 m 5,5
Permeability k 1 0 m/day
ref 3
E50 60 000 kN/m
ref 3
Stiness Eoed 60 000 3E7 kN/m
ref 3
Eur 1,8E5 kN/m
Power m 0,55
Poisson ratio νur 0,2 0,2 -
Dilatancy y 8 °
Friction angle f 38 °
Cohesion cref 0,1 kN/m
2
61
4.2. EMBEDDED PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
Weight γ 5 kN/m
3
Medium Fine
Pile-Raft I 16 048 36 120
Pile-Raft II 14 300 36 800
Figure 4.15: Details about an embedded pile-raft geometry in 3D, ne mesh, pile-raft II
62
4.2. EMBEDDED PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
Remark:
As we did previously all the following load-settlement curves are plotted for the same node
point A located at the top right side of the pile, under the slab.
Concerning the skin friction proles, we read out the Tskin 2 value [kN/m] by selecting the
embedded pile in Plaxis output. Then we divided Tskin by the perimeter of the pile to get the
2
skin friction qs [kN/m ]. In order to get something comparable from one model to an other,
we subtracted the rst phase from the pile for each load step plotted. Thus the skin friction
that we present here are only due to the load and the weight of the slab.
output. Thus we compared this value with the base resistance values found with axisymmetric
models.
Table 4.9: Linear skin friction distribution n°1 for Pile-Raft I and II
2
The Skin force Tskin , expressed in the unit of force per unit of pile length, is the force related to the relative
displacement in the pile´s rst direction (axial direction)
3
The pile foot force Ff oot , expressed in the unit of force, is obtained from the relative displacement in the axial
pile direction between the foot of the pile and the surrounding soil.
63
4.2. EMBEDDED PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
Figure 4.16: Load-displacement curves for Pile-raft I, dense sand, ψ =8°, Rinter = 0, 7
Figure 4.17: Load-displacement curves for Pile-raft II, dense sand, ψ =8°, Rinter = 0, 7
64
4.2. EMBEDDED PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
2
Table 4.10: Settlements for the dierent models for 1000 kN/m , Dense sand, ψ =8°, Rinter = 0, 7
We can conclude that the mesh inuence seems to be still quite negligible.
We also notice that axisymmetric curves are not in a very good agreement with embedded pile curves.
Thus, we note a dierence of around 15% in the settlements for axisymmetric and embedded ne
Figure 4.18: Skin friction with the load for ψ = 8, Pile-Raft I, Dense sand, Rinter = 0, 7
65
4.2. EMBEDDED PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
Figure 4.19: Skin friction with the load for ψ = 8, Pile-Raft II, Dense sand, Rinter = 0, 7
When we compared these embedded pile skin friction proles with the axisymetric ones we notice
that they are very dierent. We cannot observe the increase under the slab we described previously
in the 2D analysis. Thus the mobilization of such a dened embedded pile is dierent.
So we decided to change our linear skin friction distribution using more realistic values. We dened
these values from the axisymetric skin friction proles. Thus we performed new calculations by
66
4.2. EMBEDDED PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
67
4.2. EMBEDDED PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
The load-displacement curves we got with these new parameters are almost strictly the same as
those we had with the linear skin friction n°1. However, we can observe some dierences in the
Figure 4.20: Skin friction with the load for Pile-Raft I, Dense sand,ψ = 8, Rinter = 0, 7
Figure 4.21: Skin friction with the load for Pile-Raft II, Dense sand,ψ = 8, Rinter = 0, 7
68
4.2. EMBEDDED PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
For the lowest load, the proles are exactly the same for the distribution -n°1 or n°2- . Nevertheless,
with the highest load and the input linear skin friction distribution n°2 , the skin friction reaches
To conclude we can say that neither linear skin friction distribution n°2 nor linear skin friction
distribution n°1 leads to a skin friction prole in aggrement with the realistic one.
The multilinear skin friction distribution n°1 is a quite simple but realistic multilinear distribution:
Table 4.13: Multilinear skin friction distribution n°1 for Pile-Raft I (left) and II (right)
69
4.2. EMBEDDED PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
The multilinear skin friction distribution n°2 is the same as multilinear skin friction distribution
n°1 with Tmax =0 kN/m instead of 1 in the depth equal to 15m. Finally the multilinear skin
friction distribution n°3 is a more complex multilinear distribution designed from the axisymetric
Table 4.14: Multilinear skin friction distribution n°3 for Pile-Raft I (left) and II (right)
Output:
We noticed that the behaviors observed with distributions n°1 and 2 are exactly the sames. More
precisely, the load-displacement curves and the shear stresses distributions we got with n°1 and
70
4.2. EMBEDDED PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
When we compare the n°1&2 load-displacement curve with the axisymmetric one, we see that they
do not t very well. There is a dierence of 12,5% (Pile-Raft I) and 7% (Pile-Raft II) in settlements.
Figure 4.22: Load-displacement curves for multilinear n°1/2 embedded and axisymmetric pile-raft I
Figure 4.23: Load-displacement curves for multilinear n°1/2 embedded and axisymmetric pile-raft
II
Concerning distribution n°3, the load-settlement curve is almost the same as for distribution
n°1/2.
71
4.2. EMBEDDED PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
Load=25 kN/m
2 Load=500 kN/m
2 Load=1000 kN/m
2
Finally by plotting the shear stresses distributions for each case, no multilinear skin resistance input
yields to the realistic skin friction mobilization we calculated with the axisymmetric models.
72
4.2. EMBEDDED PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
Figure 4.24: Evolution of skin friction for dierent models and loadings
update on plaxis website, when using the layer dependant skin resistance for the embedded
piles, while leaving the linear skin resistance values to their defaults, the calculation kernel will show
a "severe divergence" error message. This severe divergence is caused by the zero values for the
linear skin resistance, though they do not have any inuence on the layer dependant skin resistance.
To overcome this error, users are advised to set the linear skin resistance values to some values not
will not have an inuence on the layer dependant values for the skin resistance. 4
Input:
For the layer dependent distribution n°1 we let the default values suggested by Plaxis.
4
Plaxis website, Known issues 3D Foundation 2.1, 26-03-2008
73
4.2. EMBEDDED PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
As we explained in the introduction of this section, we did not let the default values for the linear
For the layer dependent distribution n°1bis we used the values as described in the previous table,
The layer dependent distribution n°1 and the layer dependent distribution n°1bis perfectly match.
It conrmed that these linear skin resistance values do not have an inuence on the layer
dependent results. We just need to write a value not equal to zero in linear to use correctly the
layer dependant option.
Moreover, the skin distribution prole is in a perfect agreement for the layer dependent distribution
n°1 and n°1Bis. We also have a quite good match with the axisymmetric prole.
If we calculate the dierence of skin friction at half a pile between the axisymmetric and layer
74
4.2. EMBEDDED PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
We performed another calculation with the parameters of the so called layer dependent distribution
75
4.2. EMBEDDED PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
The load-displacement curves we got for the layer dependent distribution n°1 with Rinter = 1 is close
to one with Rinter = 0, 7 but not exactly equal. In each case, they are not in good aggrement with
the axisymmetric results. We have a dierence of around 12,5 % (Pile-Raft II) in the settlements
Load=25 kN/m
2 Load=500 kN/m
2 Load=1000 kN/m
2
For each input distributions we also plotted the skin friction distributions. In each case, the skin
distribution prole is in quite good agreement with the axisymmetric prole, particularly with the
76
4.2. EMBEDDED PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
The layer dependent distribution option seems to be the best way for embedded piles to get skin
friction distributions with realistic shapes. Nevertheless further tests must be done to really determine
the inuence of the virtual value we need to input in linear skin resistance when we use the layer
dependant option.
4.2.1.5 Comparison of the three options: Linear, multilinear and layer dependent
We now compare the three approaches in order to determine which approach is the best for analysing
Load-displacement behavior:
Concerning the load-displacement behavior, the linear embedded pile raft is the closest to the ax-
isymmetric model. However for common geotechnical displacements (max. 10 cm), whatever the
input option for the embedded pile is, the load displacement curve stays quite reasonably close to
77
4.2. EMBEDDED PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
Load=25 kN/m
2 Load=500 kN/m
2 Load=1000 kN/m
2
Pile-raft behavior:
According to the skin friction distributions presented previously, we saw that the mobilization of the
embedded pile-raft foundation and the axisymetric pile-raft foundation is dierent. So we analysed
in more details the load repartion for each model. The following values have been calculated by
78
4.2. EMBEDDED PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
Table 4.19: Rraf t ,Rskin ,and Rbase repartition for dierent models of Pile-Raft II (Dense sand, ψ =8°)
79
4.2. EMBEDDED PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
Remark
For these previous gures we substracted the weight of the pile and of the raft .
80
4.2. EMBEDDED PILE CHAPTER 4. PRELIMINARY STUDIES OF 3D MODELS
Table 4.20: Rraf t ,Rskin ,and Rbase repartition for dierent models of Pile-Raft I (Dense sand, ψ =8°)
The linear and multilinear embedded pile models seem to lead to realistic values of αKpp . Actually
we calculated values of αKpp very close to the axisymmetric ones. The main problem remains
the mobilization of the base resistance because whatever the input is, we underestimate Rbase
with embedded piles in comparison with 2D models.
Moreover, the skin friction of embedded piles seems to be overestimated in comparison with chapter 2
except the calculations with the layer dependent skin resistance. But with this approach we always
To conclude we can say that with embedded pile option we did not manage to calculate
the pile raft behavior we observed with volume piles or axisymmetric models.
81
Chapter 5
Group eect
into account the pile-soil interaction, the raft-soil interaction and the pile-raft interaction but not
the pile-pile interaction. Yet when the piles spacing is small, a partial geometry with a single pile
and one section of the raft is not accurate enough. We must consider the pile-pile interaction and
In this chapter, we present some observations about the group eect in a piled raft foundation.
We studied a piled raft foundations with 6×6 piles. Thus by using symetries we modeled only 9
piles. To study the inuence of pile length and diameter as well as spacing between piles we designed
82
5.1. PRESENTATION OF CALCULATIONS CHAPTER 5. GROUP EFFECT
Figure 5.1: 6×6 piled raft foundation, we model the red delimited quarter only
5.1.1 Geometry
The following pictures present the global geometry of our models.
83
5.1. PRESENTATION OF CALCULATIONS CHAPTER 5. GROUP EFFECT
Figure 5.2: Model of one quarter of one piled raft foundation (6×6 piles)
84
5.1. PRESENTATION OF CALCULATIONS CHAPTER 5. GROUP EFFECT
The main geometric paramaters of this study are presented in this gure:
85
5.2. RESULTS CHAPTER 5. GROUP EFFECT
For each group we also performed the same calculations without the piles in order to evaluate the
mesh with around 50 000 elements (see gure 5.2). Finally, the load is a distributed load applied on
the slab.
We tested only the dense sand material, with Rinter =0,7 and ψ =8°.
Permeability k 1 0 m/day
ref 3
E50 60 000 kN/m
ref 3
Stiness Eoed 60 000 3E7 kN/m
ref 3
Eur 1,8E5 kN/m
Power m 0,55
Poisson ratio νur 0,2 0,2 -
Dilatancy y 8 °
Friction angle f 38 °
Cohesion cref 0,1 kN/m
2
5.2 Results
86
5.2. RESULTS CHAPTER 5. GROUP EFFECT
Remark: To plot the load displacement curves we selected the node point located in the center of
the upper face of each pile.
and the corresponding pile settlements for each load steps. We got the following curves:
87
5.2. RESULTS CHAPTER 5. GROUP EFFECT
For the other models, we plotted the load-settlement curves for the raft. We also compared the
behavior with the closest single pile-raft model. As these partial geometries have not exactly the
same geometric parameters (see previous chapter) they are not entirely comparable but give us some
88
5.2. RESULTS CHAPTER 5. GROUP EFFECT
Whatever group we examine the settlement behavior of each pile is not really dierent from one pile
to another. The whole behavior is very sti and we could propably improve our models by varying
to observe vertical displacements (uy ) around center, middle and edge piles. All these cross sections
89
5.2. RESULTS CHAPTER 5. GROUP EFFECT
90
5.2. RESULTS CHAPTER 5. GROUP EFFECT
91
5.2. RESULTS CHAPTER 5. GROUP EFFECT
We notice that the displacement behavior around middle and center piles for groups with a large
spacing (group 2 and 5) is quite comparable and in good agreement with the associated single pile-
raft model. On the other hand, with the small spacing groups (group 1 and 4) the settlements are
In each case the behavior of the edge pile cannot be compared with the one of the associated single
pile-raft model because the raft is larger in the area of the edge pile.
and at the bottom of the pile. Then we considered that: Rbase = Nbottom and Rskin = Ntop -Nbottom .
The value from Nbottom are not the highest value of the normal force along the pile but this value
normally occurs a bit above the pile toe (around -29,5 m).
These values give us an idea of the real base and skin resistance.
92
5.2. RESULTS CHAPTER 5. GROUP EFFECT
Group 5 Center pile Middle piles edge piles corner pile Single pile-raft III
Rskin [kN] 12 281 12 732 13 060 13 092 24 268
Rbase [kN] 23 400 22 526 23 318 22 228 21 500
Rskin + Rbase 35 681 35 258 36 378 35 320 45 768
Rskin
0,52 0,57 0,56 0,59 1,12
Rbase
Remark:
We noticed that each pile of the group seems to be mobilizated in the same way (around the same
values of Rskin and Rbase ). But the mobilization of skin resistance appears to be lower than for a
single pile-raft. The skin friction proles we see in the following pictures conrm these observations.
We represented the skin friction prole in four directions for the center, middle and corner piles of
group 5. We added the skin friction proles for the single pile-raft III (in blue).
93
5.2. RESULTS CHAPTER 5. GROUP EFFECT
2
Figure 5.14: Skin friction prole for the center pile, Load=1000 kN/m
94
5.2. RESULTS CHAPTER 5. GROUP EFFECT
Figure 5.15: Skin friction prole for a middle pile, Load=1000 kN/m
2
95
5.2. RESULTS CHAPTER 5. GROUP EFFECT
Figure 5.16: Skin friction prole for the corner pile, Load=1000 kN/m
2
96
5.2. RESULTS CHAPTER 5. GROUP EFFECT
5.2.5 Conclusion
So, we can notice that the behavior of a pile in a pile raft foundation is dierent from the single pile
raft. The pile-pile interaction seems not to be negligible even with a quite big spacing
(6 × draf t ).
Nevertheless, we did not have enough time to perform a more precise study. Our models seem to be
too sti and we can easily improve this studies by modifying the raft parameters for instance. This
section only give some clues to start a more relevant work about the pile-pile interaction.
Remark:
A last calculation has been perfomed with Group 2. We reduced the raft stiness from Eref = 3.107
kN/m
2 to E = 1.106 2
kN/m . We get the following load displacement curves:
ref
97
Chapter 6
Conclusion
In chapter 2 a parametric study has been performed. Some remarks about how to design axisym-
The importance of interface elements was shown. The input value for Rinter changes the load
settlement curves
Prescribed displacement seems to be better than load control to study a single pile
By using load control, premature failures can occur. We can prevent this problem switching
arc-length control o. But without arc-length control, oscillations can appear. We observed
In chapter 4 we tested some 3D Plaxis tools to model piles: volume pile and embedded pile. The
results observed with volume piles are in a very good agreement with the ones of axisymmetric
models. On the other hand, the use of embedded pile to model piled-raft foundations was more
dicult. We did not manage to calculate the pile raft behavior we had observed with
98
CHAPTER 6. CONCLUSION
volume piles or axisymmetric models. The main problem remains the mobilization of
the base resistance because whatever the input is, we underestimate Rbase with embedded piles
Chapter 3 and 5 were focused on the analysis of the load repartition in piled-raft foundations. The
ability of Plaxis to model the complex pile-raft-soil interaction was conrmed. We observed the
main aspects of the pile-raft behavior that are usually described in books. We noticed
that:
The stress repartition between the raft and the pile evolves with the loading. The higher the
loading is, the more the stress is shared. With a load between 0 and 200 kN/m
2 everything
2
goes mostly to the pile (1 <αKpp < 0,8). From 200 kN/m the raft has a stronger inuence.
The bigger the spacing is, the more load takes the raft.
In each case the raft and pile load sharing converges to an equilibrium state
The pile obviously carries more load by increasing the length of the pile
We started a more sophisticated analysis of the total structure with a three dimensional model to
take the pile-pile interactions into account. We noticed that the behavior of a pile in a piled raft
foundation is dierent from the single pile raft model. There is a reduction of the skin friction
mobilization due to a group eect. We did not have enougth time left to pursue in more details
the improvement of these models. This nal chapter gives some basic clues to perform a deeper
analysis of pile-pile interactions in piled-raft foundations. This eect should be studied in further
researches.
99
Bibliography
[1] Brinkgreve R.B.J (2006). Plaxis, Finite Element Code for Soil and Rock Analyses, user
manual. Delft University of Technology & Plaxis b.v, The Nederlands. Balkema, Rotterdam.
[2] Y. El-Mossalany. Single pile and pile group in overconsolidated clay, Plaxis 3D foundation /
validation Version 2. Ain shams University
[3] Y. El-Mossalany. Pile raft foundation in frankfurter clay, Plaxis 3D foundation / validation
Version 2. Ain Shams University
[4] H. K Engin. Validation of embedded piles, the Azley Bridge pile load test, Plaxis 3D foundation
/ validation Version 2. Middle-East Technical University
[5] F. Tschuchnigg and H. Schweiger. Application of the ground anchor facility, Plaxis 3D
foundation / validation Version 2. Graz University of Technology
[6] M. Wehnert and P.A. Vermeer. Numerical Analyses of Load Tests on Bored Piles. University
of Stuttgart
100