You are on page 1of 7

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/265683850

Pile installation using FLAC


Conference Paper January 2003

CITATIONS

READS

119

2 authors:
Assaf Klar

Itai Einav

Technion - Israel Institute of Technology

University of Sydney

98 PUBLICATIONS 759 CITATIONS

110 PUBLICATIONS 1,161 CITATIONS

SEE PROFILE

All in-text references underlined in blue are linked to publications on ResearchGate,


letting you access and read them immediately.

SEE PROFILE

Available from: Assaf Klar


Retrieved on: 09 September 2016

Pile installation using FLAC


A. Klar

Technion Israel Institute of Technology, Haifa, Israel

I. Einav

Centre for Offshore Foundation Systems, UWA, Australia

ABSTRACT: This paper presents a numerical simulation of pile installation using FLAC. A new contact formulation between rigid and deformable bodies is employed. This formulation utilizes equations of motion to
describe the behavior of the deformable nodal point along the contact surface. Unlike FLACs own embedded
interface formulation, the new formulation does not encounter discontinuities problem along nonlinear or
piecewise linear surfaces.
1 INTRODUCTION

2 REVIEW OF METHODS

The evaluation of pile installation has great significant in design, for two main reasons:
1 In saturated clay soils, a considerable change in
pore pressure takes place due to the pile installation. This change of pore pressure and its subsequent dissipation process affect the pile capacity.
2 Simulation of pile installation allows for more
accurate evaluation of the end bearing capacity.
Over the last three decades, the problem of pile
installation has been extensively researched by different analytical/numerical methods. These methods
can be, generally, categorized into five groups:
1 Limit analysis approach,
2 cavity expansion solution,
3 strain path method,
4 Eulerian large strain finite element analysis, and
5 Lagrangian large strain analysis.
The problem of pile installation is strongly related to
the problem of cone penetration. As a result, advances in understanding were, and still are, strongly
connected to research of cone penetrations.
In the present work, simulation of pile installation
is presented using the Lagrangian large strain analysis code FLAC. To understand the importance of using Lagrangian analysis, the following section overviews the different methods and their limitations.
Since the problem of pile installation involves interaction between two bodies (pile and soil), there is
a need to employ some kind of interface formulation. FLACs own interface formulation is associated
with undesirable behaviors, especially along nonlinear convex surfaces. As a result, an alternative and
simple approach is suggested.

Generally, it may be said that five numerical/analytical methods exist for the evaluation of
pore-pressure generation and/or end bearing capacity:
2.1 Limit Analysis Approach
This group of methods includes lower bound solution (or slip line method), upper bound solution and
limit equilibrium analysis. Results obtained by this
approach correspond to collapse mechanism. In
these methods only the strength parameters of the
soil are introduced; i.e. the stiffness of the soil has
no influence on the results. One may refer to Durgunoglu & Mitchell (1975) for some examples of
failure mechanisms.
2.2 Cavity Expansion Solutions
In this group, the stresses along the pile are related
to solutions of cavity expansion. It is commonly assumed that the solution of cylindrical cavity expansion may represent the deformations and stresses
along points, which are far from the end of the pile,
and that solution of spherical cavity expansion may
be used for approximation of field quantities near
the pile lower tip. Figure 1 shows the different
zones. Zone II and III are the ones that may be represented by the cavity expansion solutions. The behavior in Zone I is highly affected by the soil surface. One of the purposes of the work presented here
is to evaluate the surface effect on the solution, and
to discover the required depth for which the evolution of the end bearing capacity factor Nc is redun-

dant. To learn more about the use of cavity expansions in the solution of pile and cone penetration,
one may refer to the excellent book by Yu (2000).

ever, that this method is associated with numerous


numerical problems, when simulating pile or cone
penetrations, due to the great deformations involved
in the problem.
3 PENETRATION DIFFCULTIES IN FLAC

Figure 1. Cavity expansion zones along the pile.

2.3 Strain path method


The strain path method was first suggested by
Baligh (1985). In this method, a flow field of soil is
assumed to exist around the pile. From this assumed
flow field, strains are derived, while stresses can
then be determined according to a particular constitutive relation. Two main drawbacks are associated
with this method: (a) equilibrium will not necessarily exist, and (b) the effect of the surface and/or
changing properties with depth cannot be included,
i.e. it can represent only deep steady penetrations.
The first limitation may be partly overcome by the
use of the iterative procedures suggested by Teh &
Houlsby (1988).

At this current stage, if one chooses to utilize


FLACs embedded interface formulation to simulate
the soil-pile contact, he should acknowledge that on
top of the problems due to the great deformations, he
introduces new problem. As in many other codes, in
FLAC'
s interface formulation the two bodies are
prevented from crossing each other. This leads to
discontinuities in the contact between the bodies
(Itasca 2000), if nonlinear or piecewise linear surfaces are involved. Figure 2 shows an example of
the problem for piecewise linear rigid contact surface. In Figure 2a the contact formulation corresponds to that of FLAC; i.e. it does not allow for the
deformable body (represented by the quadrilateral
elements) to overlap the rigid body (represented by
the thick black line), and therefore gaps between
these two are developed near discontinuity points
along of the rigid body. There are two kinds of gaps
that may develop between the rigid body and the deformable one. The first kind, (noted as type I in Fig.
2), is a gap that will always result when the deformable body is in contact with a concave surface of a
rigid body, and is fictional because the lamped grid
points are still in contact. The second kind (noted as
type II in Fig. 2) is true gap associated with zero
forces acting on the grid points.

2.4 Large strain Eulerian finite element analysis


In the Eulerian large strain finite element analysis,
the material (soil) streams through fixed points in
space [e.g. van den Berg 1994]. This method possesses the same limitation of the last methods; i.e.
inability to model the soil free surface.
2.5 Large strain Lagrangian finite element analysis
This method is the one employed in FLAC, in which
the mesh is updated throughout the pile penetration
process. This method is the only one that in theory
can capture the installation process as it is, while including the influence of both soil surface and changing properties with depth. It should be noted, how-

Figure 2. Contact problems along piecewise linear surface.

This behavior, shown in Fig. 2(a) for piecewise linear surface, will be more pronounced in nonlinear
surfaces, since, at least, every second grid point will

be out of contact with the rigid body in a convex


surfaces. To overcome this problem of undesirable
gaps, the contact formulation must allow the deformable body to overlap the rigid body, as shown in
Figure 2b. In this overlapping formulation, the
grid points, from which the strain increments are derived, travel along the rigid body surface, unless
some sort of contact logic that allows separation is
included. In the following section, such a contact
formulation is presented, and is used later for the
analysis of pile installation.
4 NEW CONTACT FORMULATION FOR FLAC
The following contact formulation is applicable to
the interaction between rigid and deformable bodies.
In general, both the rigid and the deformable bodies
are free to move in space. In the present formulation
the rigid body motion is prescribed. However, it can
easily be extended to a more general case where the
motion of the rigid body is determined by the solution of its motion equations; this feature is studied
these days and is being employed for the simulation
of anchor installation. In the most degenerate way,
the present formulation can also be used to create
roller fixing in an inclined angle; an option that is
absent from FLAC.
4.1 Formulation
A body may be defined as rigid if the distance between any two points of it is constant with time. The
motion of a non-rotating rigid body can be described
by two components, a velocity vector, vR and an acceleration vector aR. The motion of each lumped
mass located on a grid point that represents the deformable body can also be described by two vectors,
vD and aD for velocity and acceleration respectively.
vD and aD are here defined in the fixed coordinate
system. If considering the moving coordinate system
then the deformable body motion is defined by a velocity vector of vL = vD - vR and acceleration vector
of aL = aD - aR. Note the rigid body is stationary in
the moving coordinate system. Figure 3 shows velocity and acceleration diagrams of a grid point located on the rigid body. This body can be represented by a shape function x = f (z) (x and z are
the coordinates of the moving system).
Since the deformable body cannot enter the rigid
body nor departure from it (unless tensile failure is
considered as will be discussed later) the motion of
it can only be tangential to the rigid body; i.e. only
the tangential components of both velocity vector
and acceleration need to be introduced in to the
equations of motion. If we consider an explicit time
marching numerical scheme, the following expression can be written:

vLx ' t dt

(vLx ')2 (vLz ') 2 cos

dt (aLx ') 2 (aLz ') 2 cos


vLz ' t dt

sin

sin

(vLx ') 2 (vLz ') 2 cos

dt (aLx ') 2 (aLz ') 2 cos

cos

cos

arctan

vLx '
vLz '

arctan

vDx vRx
vDz vRz

arctan

aLx '
aLz '

arctan

aDx
aDz

(1)

aRx
aRz

where vLx ', vLz ' and aLx ', aLz 'are the components of velocity and acceleration vectors in the moving coordinate system (x,z) at time t, vDx , vDz , aDx , aDz , vRx ,
vRz , aRx , aRz are the components of the motion vectors of the deformable and rigid body in the fixed
coordinate system (x,z) at time t. If the motion of the
rigid body is prescribed (i.e. know a priori) then the
motion of the deformable body in the fixed coordinate system is as follows:

vDx (t dt ) vLx '(t dt ) vRx (t dt )


vDz (t dt ) vLz '(t dt ) vRz (t dt )

(2)

Figure 3. Velocity and acceleration diagrams.

Note, that in explicit numerical scheme it is assumed that state variable are frozen at each step (dt);
i.e. for each time step the rigid and deformable body
are fixed in space, and therefore, all values in the
right side of Equation 1 are known. The value of aD
in Equation 1 is obtained from the assumption that

the deformable body is not in contact with the rigid


body; i.e. the acceleration is obtained from the
forces acting on the grid point due to the deformation of the deformable body. To introduce some friction between the rigid and the deformable body one
can add frictional force in the tangential direction
and re-obtain from it the value of aD. If a rigidplastic tangential interface is desired then one can
define the friction forces direction simply according
to the relative velocity between the soil and the pile.
If an elastic plastic tangential interface is desired a
slightly more complicated formulation is required.
Since in the current paper only smooth piles are considered, this kind of formulation is not presented, although written and verified by the writers.
To consider possible separation between the rigid
body and the deformable body, a contact logic must
be introduced. If, for example, the contact logic considers zero tensile forces between the rigid and the
deformable bodies as condition for separation, then
it will occur once 180 + a > > a. Whenever this
condition is satisfied, the grid point is solved according to aD; i.e. D(t + dt) = D(t) + aD(t)dt. If during
one of the following steps the grid point comes in
contact with the rigid body, Equations 1 & 2 are applied. Some small changes need to be introduced
into Equation 1 if it is desired to apply one of
FLACs damping schemes which operates on grid
points mass. If damping results only from the constitutive model, then Equation 1 is satisfactory. One
may refer to Einav & Klar (2003) where the above
formulation is extended to a more general case of
three-dimensional rotating rigid-deformable bodies
in space.
The described procedure is easily implemented in
FLAC using a FISH function, which is called during
each of the calculation cycles. Generally, since the
motion of the contact grid points is solved independently (i.e. using Equations 1 & 2 rather than by
FLAC), they need to be fixed in both directions.
Quantities related to aD are extracted from FLACs
gridpoint variable xforce and yforce. Quantities related to velocities, both readable and writeable, are
manipulated using FLACs gridpoint variable xvel
and yvel.
5 NUMERICAL ANALYSIS
5.1 Pile shape
In the present numerical analysis the pile tip is represented by a continuous nonlinear shape function
according to the following expression:

x ' r0 1 Exp

bc
z'

(3)

where r0 is the radius of the pile and and bc are parameters that define the shape of the pile tip. Figure
4 shows three examples of tip shapes once is set to

give 95% of r0 at z=3r0. Due to space limitation, in


the current paper we present results only for pile tip
with bc equal to 2.
5.2 Assumptions
The numerical analysis was conducted under the following assumptions:
1 The material behaves elastic perfectly plastic and
satisfies the von-Mises failure criteria. Since the
analysis is associated with undrained loading, and
the volumetric stresses are decoupled from the
deviatoric ones in the considered constitutive
model, it is possible to perform a Dry simulation; i.e. to obtain the excess pore pressure value
using Skemptons parameter, B = 1 1/(1 +
(Kw/n)/Ks), utilizing the formula u = B( ii
0
ii )/3, where the superscript 0 denotes initial
state.
2 The undrained strength of the soil is defined according to the relation Cu = 0.25 0OCR0.95, 0
is the initial vertical effective stress and OCR is
the over consolidation ratio.
3 The shear modulus is taken proportional to the
undrained strength Cu, and the bulk modulus was
high enough so the material can be considered as
incompressible.
4 The analysis presented herein considered a constant OCR with value that equals 2.
5 Initial stress condition corresponded to K0 of 0.7.
6 To avoid the kinematic constrains of a fixed
boundary, and to allow approximate simulation of
an infinite soil layer, a prescribed boundary condition was applied to the outer radius of the grid.
The boundary condition corresponded to the cylindrical cavity expansion solution of an incompressible elastic material; i.e. the external pressure acting on the mesh was defined by the
analytical solution of the internal pressure of cavity expansion with an identical radius to that of
the outer boundary. It was verified that the plastic
zone did not reach the outer boundary, thus the
elastic cavity expansion solution was suited.
6 RESULTS
Figure 5 shows typical distortion of the mesh associated with steady penetration of the pile. It should be
noted that analyses with much coarser mesh resulted
with almost identical stress distributions and response, and that it was impossible for the pile to
penetrate the surface without causing a bad geometry, unless more coarser mesh, than that shown in
Fig. 5, was prescribed near the soil surface. As a
rough rule of a thumb, it was found that a contact
soil element near the surface should have a radial
dimension of about one pile radius, and this can be
rapidly become finer as elements are deeper. Figure

6 shows normalized excess pore pressure associated


with the state of Figure 5.

failure is reached quicker when the points are closer


to the axis.

Figure 4. Different pile tip shapes.


Figure 6. Normalized excess pore pressures (G/Cu = 100).

10

Points aligned
with cone tip

xi/r0=0.31
xi/r0=1.0
xi/r0=1.65
xi/r0=3.15
xi/r0=5.5
xi/r0=10.5

u/Cu

6
4
2
0
-2
0

10

20

30

40

Penetration/r0

Figure 7. Development of excess pore pressure (G/Cu = 100).


Figure 5. Distorted mesh in steady penetration (G/Cu = 100).

Figure 7 shows the development of excess pore


pressure at depth of 25 radiuses for different radial
distances (xi is the initial distance from the axissymmetric line). Figure 8 shows the changes of the
second invariant of the stress during the installation
of the pile for points located a depth of 25 radiuses.
The y-axis is normalized such that it gives maximum
value of one, in accordance to the von-Mises yield
surface radius. Initial K0 conditions create initial
value that is different than zero. Clearly, as the tip
advances towards the checkpoints, the value of the
second invariant increases until failure is reached;

As discussed in section 2, only Lagrangian large


strain analyses can simulate the penetration of the
pile through the surface, and therefore are the only
ones that can evaluate the influence of the surface on
the cone penetration. Figure 9 shows the cone resistance factor, Nc, for different rigidity indexes
(G/Cu). The Nc factor was calculated according to:

Nc Zc

Fnet
Cu Zc

r02

Ftot
r02
v Zc
Cu Zc
r02

(4)

where, Ftot is the total vertical force acting on the


pile (considering a smooth pile), v(Zc) is the total
vertical stress at depth Zc, and defined as the distance between the surface and the middle of the
piles tip (and in our shapes, 1.5r0 above its tip).

Note that in Figure 9 the horizontal axis Zc/D does


not start at zero. This is due to the fact that the cone
must be completely positioned inside the soil in order for the Nc factor to have a proper meaning, if it
is obtained from the net vertical force,

Fnet

Nc Cu

1.2

Points aligned
with cone tip

xi/r0=0.31
xi/r0=1.0
xi/r0=1.65
xi/r0=3.15
xi/r0=5.5
xi/r0=10.5

J2D /(2/3 Cu)

0.8

7 SUMMARY AND CONCLUSIONS

0.5

0.5

r02 .

piles slenderness ratios. For example, for G/Cu =


1118 the pile slenderness ratio must be greater than
25 in order for the soil surface to have no effect. In
case the pile slenderness is smaller than that value,
the soil surface influences the Nc factor. In such case
it is not legitimate to use cavity expansion solutions
for obtaining the Nc factors, as they assume that
there is no influence from surface. The same may be
regarded to the solutions based on the strain path
method, which also does not consider the soil surface.

0.4

0
0

10

20
Penetration/r 0

30

40

Nc

Figure. 8. Shear behavior (G/Cu) = 100.

15
14
13
12
11
10
9
8
7
6
5
4

G/Cu=1118
G/Cu=500
G/Cu=223
G/Cu=100
G/Cu=44.7
G/Cu=20
Steady state front
1 4

7 10 13 16 19 22 25 28 31 34 37 40
Zc/D

Figure 9. Development of Nc factor with depth.

As can be seen from Figure 9, as the rigidity index increases, both the Nc value and depth in which
it becomes constant increase. The dashed line in
Figure 9 represents a required depth to obtain 95%
of the maximum Nc values. This depth is referred
herein as the depth of the steady state front, Zss. It is
obvious from Figure 9 that Zone III (see Fig. 3) can
be associated with spherical cavity expansion solution, as was suggested by Yu (2000), only if some
minimal pile slenderness ratio (L/D) is satisfied. It
seems that for piles installed in soil with rigidity
G/Cu < 100 the requirements for minimum slenderness are irrelevant, since the normalized steady state
penetration depth is smaller than any reasonable
slenderness ratio (L/D) associated with piles foundations. However, for G/Cu > 100 the normalized
steady state depth becomes in the same order of the

A contact formulation for interaction between rigid


and deformable bodies is presented. This approach
overcomes some of the problems associated with
FLACs own built-in interface formulation when it is
applied to nonlinear or piecewise linear surfaces.
The proposed formulation can easily be used to create rolling fixing along any line inclination, an option that is currently absent from FLAC. The contact
formulation is employed in the large strain simulation of pile installation. A study on the generation of
pore pressures and on surface effects is presented.
There is a strong indication from the analysis results,
that the use of cavity expansion or strain path
method solutions should be carefully examined before employed in the estimation of end bearing capacity, specially for low slenderness driven piles in
soils with high rigidity index.
REFERENCES
Baligh M.M. 1985. Strain path method, J. Soil Mech. and
Found. Div., ASCE, 111(9): 1108-1136
Durgunoglu, H. T. & Mitchell, J. K. 1975. Static Penetration
Resistance of soil, I: Analysis. Proc. ASCE Spec. Conf. on
In Situ Measurement of Soil Properties, New York, Vol. 1
151-171.
Einav, I and Klar, A. 2003. An approach for nonlinear contact
surface analysis and application to pile installation. BGA
Int. Conf. On Foundations: Innovations, Observations,
Design and Practice, Dundee, Scotland, Sept 2003.
Itasca Consulting Group, Inc. 2000. FLAC (Fast Lagrangian
Analysis of Continua) Ver. 4.0 User'
s Manual, Minneapolis
Minnesota: Itasca.
Teh, C.I. & Houlsby, G.T. 1988. Analysis of the Cone Penetration Test by the Strain Path Method. Proceedings of the 6th
International Conference on Numerical and Analytical
Methods in Geomechanics, Innsbruck, April, Vol. 1, ISBN
90-6191-810-3, pp 397-402
van den Berg P. 1994. Analysis of soil penetration. Ph.D. thesis. The Netherlands: Delft University Press.
Yu, H.S. 2000. Cavity Expansion Methods in Geomechanics.
London: Kluwer Academic Publisher.

You might also like