BEM-FEM coupling model for the dynamic

analysis of piles and pile groups

L.A. Padron, J.J. Aznarez, O. Maeso

Instituto Universitario de Sistemas Inteligentes y Aplicaciones Numericas en

Ingeniera (IUSIANI), Universidad de Las Palmas de Gran Canaria, Edicio
Central del Parque Cientco y Tecnologico del Campus Universitario de Tara,
35017 Las Palmas de Gran Canaria, Spain
Abstract
This paper shows a BEM-FEM coupling model for the time harmonic dynamic
analysis of piles and pile groups embedded in an elastic half-space. Piles are modelled
using Finite Elements (FEM) as a beam according to the Bernoulli hypothesis, while
the soil is modelled using Boundary Elements (BEM) as a continuum, semi-innite,
isotropic, homogeneous or zoned homogeneous, linear, viscoelastic medium. It is
assumed that the soil continuity is not altered by the presence of the piles, and
the tractions at the pile-soil interface are considered as a load applied within the
half-space. The formulation is exposed in detail. In order to validate the model,
selected numerical results of time harmonic impedances of dierent pile groups
congurations are evaluated and contrasted with other reference values taken from
the literature.
Key words: boundary elements, nite elements, BEM-FEM coupling, piles, pile
groups, pile-soil interaction, dynamic impedances
1 Introduction
The frequency domain dynamic analysis of piles and pile groups embedded
in a half-space has been treated numerically by several authors using the
frequency domain BEM (Boundary Element Method) in conjunction with dif-
ferent half-space Greens functions formulations, for the soil; and FEM (Fi-
nite Element Method) for the piles, considered as monodimensional beam

Corresponding author. Tel.: +34 928451920 Fax: +34 928451921 E-mail adress:
omaeso@iusiani.ulpgc.es
Preprint submitted to Elsevier Science October 2006
elements [112]. Also, the coupling of framed structures approached by FEM
with three-dimensional bodies represented by BEM in time domain has been
presented in [13,14], where piles would be approximated using a special cylin-
drical boundary element. A good review of presented techniques in the period
1986-1996 concerning the dynamic analysis of piles and pile groups can be
found in [15].
Using BEM for both soil and piles, more versatile and rigorous numerical
models have been developed: vibration isolation by a row of piles has been
analyzed in [16,17] and dynamic impedances of pile groups has been studied
by two of the authors in [18,19]. High computational cost is a disadvantage of
these models.
The present paper, in order to reduce the number of degrees of freedom in
the problem, presents a dierent BEM-FEM model for the time harmonic
dynamic analysis of piles and pile groups embedded in an elastic half-space,
taking advantage of the particular characteristics of each one of the methods.
This model is based on the idea of a previous static model developed in [2022]
where it is assumed that the continuity of the soil is not altered by the presence
of the piles and where the tractions in the pile-soil interface are considered as
a load applied within the half-space in the boundary integral representation
of the soil.
Piles are modelled using FEM as beams according to the Bernoulli hypothesis,
while soil is modelled using BEM as a continuum, semi-innite, isotropic, ho-
mogeneous or zoned homogeneous, linear, viscoelastic medium. Welded bound-
ary contact condition at the pile-soil interfaces are assumed. The formulation
allows the analysis of problems including soil strata, rigid rocky beds and
any topography for the soil surface. Furthermore, as the pile boundary does
not need to be discretized, low computing times and memory requirements
are needed. First, the formulation will be exposed in detail. After, in order
to validate the model, numerical results for vertical, horizontal and rocking
impedances are evaluated and compared to other reference values taken from
the literature.
2 Pile FE equations
The behaviour of a pile submitted to dynamic loads can be described by the
following dierential equation
M u(t) +C u(t) +Ku(t) = f(t) (1)
2
Fig. 1. Finite element denition
where M, C and K are the mass, damping and stiness matrices of the pile,
u(t) is the vector of nodal displacements and f(t) the vector of nodal forces
over the pile.
It will be assumed now that the pile is subjected to a harmonically vary-
ing load. In this case, the vectors of nodal displacements and forces can be
expressed as
u(t) = u
p
e
it
; f(t) = Fe
it
(2)
where u
p
is the vector of nodal translations and rotations amplitudes, F is the
vector of nodal forces amplitudes, the circular frequency of the excitation
and i =

1. Then, and considering a pile with zero internal damping, Eq.(1)

becomes
(K
2
M) u
p
= F (3)
Piles are modelled by FEM as vertical beams according to the Bernoulli hy-
pothesis, and are discretized using a three-nodes element, shown in Fig. 1,
that has been dened in order to be able to adjust the deformed shape ac-
curately with a scarce number of elements. There are 13 degrees of freedom
dened on it: two lateral displacements and a vertical displacement on each
node, and two rotations on each one of the extreme nodes, one about x
1
axis
and another one about x
2
.
3
The lateral displacements u
1
and u
2
along the element are approximated by
a set of fourth degree shape functions, while vertical displacements u
3
are
approximated by one of second degree. Thus
u
i
=
1
u
k
i
+
2

k
i
+
3
u
l
i
+
4
u
m
i
+
5

m
i
; i = 1, 2 (4)
u
3
=
1
u
k
3
+
2
u
l
3
+
3
u
m
3
(5)
where

1
= (
3
4
+ +
1
4

1
2

3
)

2
=
1
4
(1 + +
2

3
)

3
= 1 2
2
+
4
(6)

4
= (
3
4
+
1
4

1
2

3
)

5
=
1
4
(1 +
2
+
3
)
and

1
=
1
2
( 1)

2
= 1
2
(7)

3
=
1
2
( + 1)
being the elemental dimensionless coordinate varying from 1 to +1.
The stiness sub-matrices for the lateral and axial behaviour of this element
(denoted by l and a respectively), can be obtained using the principle of virtual
displacements and the shape functions dened above (see e.g. reference [23]).
These matrices can be written as
4
K
l
=
EI
5L
_

_
316
L
2
94
L
512
L
2
196
L
2
34
L
94
L
36
128
L
34
L
6
512
L
2
128
L
1024
L
2
512
L
2
128
L
196
L
2
34
L
512
L
2
316
L
2
94
L
34
L
6
128
L
94
L
36
_

_
; K
a
=
EA
3L
_

_
7 8 1
8 16 8
1 8 7
_

_
(8)
where E is the Youngs Modulus for the pile, A and I are the area and the
moment of inertia of the section of the pile and L is the element length.
The consistent-mass matrix for an element can be evaluated in a similar way.
Considering a beam with uniformly distributed mass m, the matrices obtained
for the lateral and axial behaviours are, respectively
M
l
= L m
_

_
13
63
L
63
4
63
23
630
L
180
L
63
L
2
630
2L
315
L
180
L
2
1260
4
63
2L
315
128
315
4
63
2L
315
23
630
L
180
4
63
13
63
L
63
L
180
L
2
1260
2L
315
L
63
L
2
630
_

_
; M
a
=
L m
15
_

_
2 1
1
2
1 8 1
1
2
1 2
_

_
(9)
The considered forces acting over the pile are punctual forces and moments
applied at the top of the pile and distributed forces arising from pile-soil
interaction. An axial force at the tip of the pile has been taken into account
also. So, the vector of nodal forces F can be decomposed as
F = F
ext
+F
eq
= F
top
+F
p
+F
eq
(10)
where F
ext
include the forces at the top F
top
and the axial force at the tip
of the pile F
p
; and F
eq
is the vector of the equivalent nodal forces from the
pile-soil interaction, that can be calculated as F
eq
= Q q
p
, where Q is the
matrix that transforms nodal force components to equivalent nodal forces.
The external forces dened over the generic element are schematized in Fig. 2.
The tractions q
p
along the pile-soil interface are approximated by the set of
5
Fig. 2. External punctual forces (left) and tractions along the pile-soil interface
dened on the generic element
shape functions dened by Eq.(7) as
q
i
=
1
q
k
i
+
2
q
l
i
+
3
q
m
i
; i = 1, 2, 3 (11)
Again, using the principle of virtual displacements, one can obtain the follow-
ing matrices for lateral and axial equivalent nodal forces respectively
Q
l
=
_

_
23L
140
11L
105
L
28
L
2
84
L
2
105
L
2
210
4L
105
16L
35
4L
105
L
28
11L
105
23L
140
L
2
210
L
2
105
L
2
84
_

_
; Q
a
=
L
30
_

_
4 2 1
2 16 2
1 2 4
_

_
(12)
Once all elemental matrices have been obtained for the whole pile, one can
write, for each one, the following expression

Ku
p
= F
ext
+Qq
p
(13)
where

K = K
2
M. As each pile will be discretized using as many elements
as necessary to follow its deformed shape accurately, matrices

K and Q are
6
global matrices, obtained as usual from the elemental ones.
3 Soil BE equations
The soil is modelled by BEM as a linear homogeneous isotropic viscoelastic
unbounded region with complex valued shear modulus of the type =
Re[](1 + 2i), where is the damping coecient. The boundary integral
equation for a time-harmonic elastodynamic state dened in the domain
with boundary can be written in a condensed and general form as
c
k
u
k
+
_

ud =
_

pd +
_

Xd (14)
where c
k
is the local free term matrix at collocation point k, which is a
diagonal matrix with a 1 on rows corresponding to internal points and a 0.5
on rows corresponding to boundary points where the boundary is smooth;
X are the body forces in the domain ; u and p are the displacement and
traction vectors; and u

and p

are the elastodynamic fundamental solution

tensors on the boundary due to a time-harmonic concentrated load at point
k:
u =
_

_
u
1
u
2
u
3
_

_
; p =
_

_
p
1
p
2
p
3
_

_
; u

=
_

_
u

11
u

12
u

13
u

21
u

22
u

23
u

31
u

32
u

33
_

_
; p

=
_

_
p

11
p

12
p

13
p

21
p

22
p

23
p

31
p

32
p

33
_

_
(15)
The fundamental solution used in this work corresponds to the complete space
because there is no closed form expression for the half-space fundamental so-
lution, which depends on boundless integrals and would require approximate
procedures for its evaluation. Therefore, the soil free surface should be dis-
cretized. In practice, however, only a small region around the foundation has
to be included in the model to reach accurate results.
Generally, body forces X are considered to be zero in most of elastodynamic
problems. Nevertheless, in this work, pile-soil interaction is replaced by inter-
nal punctual forces at the piles tip and load-lines placed along the piles axis,
while it is assumed that the soil continuity is not altered by the presence of the
piles, which is the main hypothesis that supports this model. The load-lines
q
s
j
within the soil and its relationship with tractions q
p
j
are represented in
Fig. 3 together with the internal punctual forces F
p
j
at the tip of the piles.
7
Fig. 3. Load-lines representation.
Then, Eq.(14) can be written as
c
k
u
k
+
_

ud =
_

pd +
np

j=1
_
_
p
j
u

q
s
j
d
p
j

j
k
F
p
j
_
(16)
where
p
j
is the pile-soil interface of pile j, n
p
is the total number of piles and

j
k
is a 3 components vector that represents the contribution of the axial force
F
p
j
at the tip of the j
th
pile, when the concentrated load is applied on point
k.
The boundary surface is discretized into quadratic elements of triangular and
quadrilateral shapes with six and nine nodes, respectively (see reference [24]).
Once the boundary has been discretized and Eq. (16) has been written for all
nodes in , the equation can be expressed in matrix form as
H
ss
u
s
= G
ss
p +
np

j=1
G
sp
j
q
s
j

np

j=1

sj
F
p
j
(17)
where u
s
is the vector of nodal displacements on the surface, H
ss
and G
ss
are
matrices obtained by integration over of the 3-D elastodynamic fundamental
solution times the shape functions of the boundary elements, and G
sp
j
is the
matrix obtained by integration over
p
j
of the 3-D elastodynamic fundamental
solution times the interpolation functions dened in (7), when the unit load
is applied over . Assuming free traction surface (p = 0), Eq. (17) becomes
H
ss
u
s

np

j=1
G
sp
j
q
s
j
+
np

j=1

sj
F
p
j
= 0 (18)
Furthermore, Eq. (16) will be also applied on the piles internal nodes, so, for
a certain pile i, one can write
8
u
p
i
k
+H
p
i
s
u
s

np

j=1
G
p
i
p
j
q
s
j
+
np

j=1

p
i
j
F
p
j
= 0 (19)
where u
p
i
k
is the vector of nodal displacements at the node k of the pile i where
the unit load is applied, H
p
i
s
is the matrix obtained by integration over of
the 3-D elastodynamic fundamental solution times the shape functions of the
boundary elements, and G
p
i
p
j
is the matrix obtained by integration over
p
j
of
the 3-D elastodynamic fundamental solution times the interpolation functions
dened in (7), when the unit load is applied over a pile i.
Besides, as an axial force at the pile tip is considered, an extra equation needs
to be written. To do so, the unit load must be applied in the x
3
direction
at any non-nodal point. The point with elemental dimensionless coordinate
= 0.5 in the bottom element of the pile has been chosen because of its
nearness to the tip. This way, the extra equation is
1
8
_
3u
b
k
3
+ 6u
b
l
3
u
bm
3
_
+
_

ud
s
=
np

j=1
_
_
p
j
u

q
s
j
d
p
j

_

j
b
_
3
F
p
j
_
(20)
where u
b
k
3
,u
b
l
3
and u
bm
3
are the vertical displacements of nodes k,l and m of
the bottom element, p

= {p

31
, p

32
, p

33
} and u

= {u

31
, u

32
, u

33
}. In a matrix
form, Eq. (20) can be written as
D
T
u
p
i
b
+H
p
i
s
e
u
s

np

j=1
G
p
i
p
j
e
q
s
j
+
np

j=1

p
i
j
b
3
F
p
j
= 0 (21)
where u
p
i
b
is the vector of nodal displacements at the bottom element nodes
of the pile i where the unit load is applied, H
p
i
s
e
is a vector obtained by
integration over of the 3-D elastodynamic fundamental solution times the
shape functions of the boundary elements, and G
p
i
p
j
e
is a vector obtained by
integration over
p
j
of the 3-D elastodynamic fundamental solution times the
interpolation functions dened in (7), when the unit load is applied on the
extra point of the pile i. D
T
is 1/8{0, 0, 3, 0, 0, 6, 0, 0, 1}.
The integrals of u

q
s
j
in (16) and u

q
s
j
in (20) over
p
j
are calculated as
a monodimensional integral extended to a load-line, dened by the pile axis,
when the collocation point is outside pile j. However, this integrals have a
singularity at the collocation point when applied on the integrated pile. In
this case, in order to avoid this singularity, the integrals are evaluated over
a cylinder which radius R
p
is
_
/A. This way, let us consider the pile-soil
9
interface
p
(whatever the kind of section) as a cylinder of radius R
p
where
tractions
ps
are applied. The last term of Eq. (14) includes integrals of the
type
_
p

ps
u

d
p
=
_
p
q
s
2R
p
u

d
p
=
1
2R
p
Ne

e=1

i=k,l,m
q
s
i
_
pe

i
u

d
p
(22)
where N
e
is the number of elements in which the load-line has been discretized
and Eq. (11) has been used to express q
s
along each element. The elastody-
namic fundamental solution, that gives the displacement in a k direction when
the load is applied in the l direction, can be written as
u

lk
=
1
4
[
lk
r
,k
r
,l
]
=
_
c
2
c
1
_
2
_
1
z
2
1
r
2

1
z
1
r
_
e
z
1
r
r
+
_
1
z
2
2
r
2

1
z
2
r
+ 1
_
e
z
2
r
r
(23)
=
_
c
2
c
1
_
2
_
3
z
2
1
r
2

3
z
1
r
+ 1
_
e
z
1
r
r
+
_
3
z
2
2
r
2

3
z
2
r
+ 1
_
e
z
2
r
r
where is the shear modulus,
lk
is the Kronecker delta function, c
1
and c
2
are the wave velocities for P and S waves and z
j
= ik
j
, being k
1
and k
2
the
wave numbers for P and S waves, respectively.
Then, the integrals in the last term of Eq. (22) can be evaluated, in cylindrical
coordinates (see Fig. 4), as
_
pe

i
u

d
p
=
_
x
r
3
_

1
4
[
lk
r
,k
r
,l
]
i
R
p
d dx
3
=
R
p
4
_
x
r
3

i
_
2
lk

r
2
R
lk
_
dx
3
(24)
where x
r
3
= x
3
x
k
3
and
R
lk
=
_
_
_
_
_
_
_
R
2
p
0 0
0 R
2
p
0
0 0 2(x
r
3
)
2
_
_
_
_
_
_
_
(25)
10
Fig. 4. Integration over pile-soil interface when the collocation point belongs to the
pile
Now, the integral of Eq. (22) can be written as
_
p

ps
u

d
p
=
Ne

e=1
R
p
L
e
8

i=k,l,m
q
s
j
_
1
1

i
_
2
lk

r
2
R
lk
_
d (26)
Fig. 5. Non-nodal collocation strategy
Nevertheless, to compute integrals over
p
j
from the same pile, a non-nodal
collocation strategy could also be carried out. This will lead to a procedure
that allows the reinterpretation of the previous equation. This way, in order
to avoid breaking the problem symmetries, at least four collocation points,
symmetrically placed around the pile, should be chosen (see Fig. 5). A single
11
equation can be obtained by adding these four equations divided by four, so
that the arising coecients are of the same order. In this case, integrating over
the load-line, the last term of Eq. (14) becomes
_
p
q
s
u

d
p
=
1
4
4

n=1
c.p.
Ne

e=1

i=k,l,m
q
s
i
4
_
x
r
3

i
[
lk
r
,k
r
,l
] dx
r
3
=
1
4
4

n=1
c.p.
Ne

e=1

i=k,l,m
q
s
i
4
_
x
r
3

i
_

lk

r
2
R
n
lk
_
dx
r
3
(27)
where
R
n
lk
n=1,2
=
_
_
_
_
_
_
_
0 0 0
0 R
2
p
(1)
n
R
p
x
r
3
0 (1)
n
R
p
x
r
3
(x
r
3
)
2
_
_
_
_
_
_
_
R
n
lk
n=3,4
=
_
_
_
_
_
_
_
R
2
p
0 (1)
n
R
p
x
r
3
0 0 0
(1)
n
R
p
x
r
3
0 (x
r
3
)
2
_
_
_
_
_
_
_
(28)
It can be seen that this procedure leads to an equation in each direction
completely equivalent to Eq. (26), i.e., both methods are analogous.
On the other hand, when the collocation point k is placed outside the pile j,
one can write

j
k
= {u

13
u

23
u

33
}
k
(29)
However, when the collocation point is placed at the bottom node of pile j,

j
k
F
p
j
becomes singular. In this case, to avoid this singularity, the axial force
at the tip can be considered as a constant pressure
b
over a circular surface
with radius R
p
=
_
/A. In this work, this strategy has been used whenever
the collocation point is placed on the axis of pile j. According to this, one can
write

j
k
F
p
j
=
_

b
u

i3

b
d
b
(30)
where
b
is the tip surface and
b
= F
p
/A.
Then, in cylindrical coordinates (see Fig. 6), Eq. (30) becomes
12
Fig. 6. Integration over pile tip surface

j
k
=
1
4A
_
2
0
_
Rp
0
[
i3
r
,i
r
,3
] a da d =
1
2A
_
Rp
0
_

i3
(x
r
3
)
2

r
2
_
a da
(31)
where the integral is regular and can be evaluated numerically.
Finally, all the integrals over are also evaluated numerically. Integrals of
O(
1
r
) are evaluated according to Li et al. [25]. Strongly-singular integrals are
evaluated by identifying the O(
1
r
2
) terms and splitting them up into a regular
surface integral and a linear integral over the element perimeter. Then, singu-
larities get cancelled with the contribution of adjacent elements. (see reference
[26])
4 BEM-FEM coupling equation
Now, a global system of equations must be built using the expressions dened
above. The links between piles and soil that will allow to do the coupling are
the tractions q
sj
= q
p
j
along the pile-soil interface and the displacements
u
p
j
along the pile j.
As it is assumed that the soil continuity is not altered by the presence of the
pile, the value of distributed mass assigned to the pile should be modied as
m = A(
p

s
) so as not to overestimate the total mass introduced in the
model, being
p
and
s
the pile and soil densities. A similar consideration was
assumed in [1,2,10].
Eq. (13) becomes, for the pile j
13

K
p
j
u
p
j
F
p
j
+Qq
sj
= F
j
top
(32)
Using equilibrium and compatibility conditions along the interface, and as-
suming the tractions q
s
as positive, Eqs. (18), (19), (21) and (32) can be
rearranged as
_

_
H
ss
G
sp

H
ps
G
pp

p
C

H
ps
e
G
pp
e

p
b
3
D

Q I

K
_

_
_

_
u
s
q
s
F
p
u
p
_

_
= B (33)
where H
ss
is the 3N 3N matrix obtained by integration of the 3-D time
harmonic traction fundamental solution times the shape functions, over the
boundary elements, when applying on the boundaries; N is the number of
nodes on the boundary; D

is a matrix made up by vectors D; C

contains
the free terms corresponding to the collocation over the pile nodes; I

is a zero
matrix with a 1 in terms corresponding to the unknown F
p
j
;
14
G
pp
=
_

_
G
p
1
p
1
G
p
1
p
2
G
p
1
pn
G
p
2
p
1
G
p
2
p
2
G
p
2
pn
.
.
.
.
.
.
.
.
.
.
.
.
G
pnp
1
G
pnp
2
G
pnpn
_

_
; G
pp
e
=
_

_
G
p
1
p
1
e
G
p
1
p
2
e
G
p
1
pn
e
G
p
2
p
1
e
G
p
2
p
2
e
G
p
2
pn
e
.
.
.
.
.
.
.
.
.
.
.
.
G
pnp
1
e
G
pnp
2
e
G
pnpn
e
_

p
=
_

p
1
1

p
1
2

p
1
n

p
2
1

p
2
2

p
2
n
.
.
.
.
.
.
.
.
.
.
.
.

pn1

pn2

pnn
_

_
;
p
b
3
=
_

p
1
1
b
3

p
1
2
b
3

p
1
n
b
3

p
2
1
b
3

p
2
2
b
3

p
2
n
b
3
.
.
.
.
.
.
.
.
.
.
.
.

pn1
b
3

pn2
b
3

pnn
b
3
_

_
G
sp
=
_
G
sp
1
G
sp
2
G
spn
_
;
s
=
_

s1

s2

sn
_
H
ps
=
_
H
p
1
s
H
p
2
s
H
pns
_
T
; H
ps
e
=
_
H
p
1
s
e
H
p
2
s
e
H
pns
e
_
T
Q =
_

_
Q
p
1

Q
p
2

.
.
.
.
.
.
.
.
.
.
.
.
Q
pn
_

_
;

K =
_

K
p
1

K
p
2

.
.
.
.
.
.
.
.
.
.
.
.


K
pn
_

_
(34)
B is the right-hand vector when all external conditions have been imposed,
and the vector of unknowns is
x = {u
s
, q
s
1
, q
s
2
, . . . , q
sn
, F
p
1
, F
p
2
, . . . , F
pn
, u
p
1
, u
p
2
, . . . , u
pn
}
T
(35)
5 Validation results
In order to validate the BEM-FEM coupling model, several results of impedances
of pile groups have been computed and compared to other reference values
taken from the literature.
The dynamic stiness matrix K
ij
of a pile relates the vector of forces (and
moments) applied at the pile top to the resulting vector of displacements (and
15
rotations) at the same point. For a group of piles, it is assumed that the
pile heads are constrained by a rigid pile-cap, and the foundation stiness is
the addition of the contributions of each pile. Fig. 7 illustrates the approached
problem for a usual conguration, where L and d are used to denote the length
and diameter of the piles, and s refers to the distance between adjacent piles.
Fig. 7. 2 2 pile group embedded in a half-space. Problem geometry denition.
The dynamic stiness terms for a time harmonic excitation are functions of
frequency and they are usually written as
K
ij
= k
ij
+ ia
o
c
ij
(36)
where k
ij
and c
ij
are the frequency dependent dynamic stiness and damping
coecients, respectively, a
o
is the dimensionless frequency
a
o
=
d
c
s
(37)
and c
s
is the soil shear-wave velocity.
Once the problem has been solved for a unitary displacement applied at the
cap, and the vectors u
s
and q
s
have been computed, the dynamic stiness at
a certain frequency can be easily obtained by applying Eq. (13) at the top
element of each pile.
Fig. 8 shows a sketch of the discretizations used to obtain the stiness of
dierent pile groups embedded in a viscoelastic half-space, where boundary
elements for the soil and mono-dimensional nite elements for the piles were
used. As the developed software incorporates symmetry properties, only a
quarter of the total geometry of the problem has to be discretized. Rectangular
quadratic nine-nodes elements were used for the free surface. The length of free
surface needed is, approximately, three times the pile length, and the element
size is chosen in such a way that its main dimension is always shorter than the
half of the wave length. On the other hand, the three-nodes elements dened
16
Fig. 8. 3x3 Pile group BEM-FEM discretization for horizontal problems
above were used on the pile. Three equal elements for vertical problems and six
elements, of which three concentrated on the top, for horizontal and rocking
problems were used. It should be noted that the model allows to change the
number and position of piles without modifying the free surface discretization.
Three sets of results are presented in this section. In the rst block, several
results are compared with those of Kaynia [1], obtained from an analysis of
single piles and pile groups considering piles as linear elastic prismatic mem-
bers and soil as semi-innite viscoelastic media by constructing the requisite
Greens function using a discrete layer matrix approach. The following prop-
erties are taken from Kaynia [1]: piles (in the sequel denoted by sub-index
p) are assumed to be elastic beams and surrounding soil a uniform viscoelas-
tic media with internal damping coecient = 0.05; ratio between densities

s
/
p
= 0.7; and Poisson ratios
s
= 0.4 (for the soil) and
p
= 0.25 (for
the pile, but not taken into account in the proposed technique). Two dier-
ent soils with ratios between the Young modulus E
p
/E
s
= 10
3
(soft soil) and
E
p
/E
s
= 10
2
(sti soil) have been considered.
Lateral, vertical and rocking impedances (stiness and damping coecients) of
3x3 pile groups embedded in a homogeneous isotropic viscoelastic half-space,
obtained by proposed technique (noted by BEM-FEM in the Figs.), are shown
from Figs. 9 to 11, all for s/d = 2, 5, 10, and soft soil. Lateral, vertical and
rocking impedances of 2x2, 3x3 and 4x4 pile groups embedded in sti soils are
shown from Figs. 12 to 14. The piles aspect ratio is L/d = 15. The vertical
and horizontal impedance functions have been normalized with respect to the
respective single pile static stiness (k
s
) times the number (N) of piles in
the group. The rocking impedances have been normalized with respect to the
sum of the products of the respective single pile static stiness (k
s
zz
) times the
square of the distance to the rotation axis (x
i
). All results are plotted versus
the dimensionless frequency parameter dened by Eq. (37).
17
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0.5
0
0.5
1
1.5
2
2.5
a
o
k
Gx
x
/
(
N

k
sx
x
o
)
s/d=2
s/d=5
s/d=10
Kaynia
BEMFEM
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
1
2
3
4
5
a
o
c
Gx
x
/
(
N

k
sx
x
o
)
s/d=2
s/d=10
s/d=5
Kaynia
BEMFEM
Fig. 9. Horizontal impedances of 3x3 pile groups in soft soil. Comparison with
Kaynias solution.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
2
1
0
1
2
3
4
5
a
o
k
Gz
z
/
(
N

k
sz
z
)
Kaynia
BEMFEM
s/d=2
s/d=5
s/d=10
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
2
4
6
8
10
12
14
a
o
c
Gz
z
/
(
N

k
sz
z
)
Kaynia
BEMFEM
Fig. 10. Vertical impedances of 3x3 pile groups in soft soil. Comparison with Kay-
nias solution.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
0.5
1
1.5
2
2.5
3
3.5
4
a
o
k
G

/
(

x
i 2

k
sz
z
o
)
s/d=2
s/d=10
s/d=5
Kaynia
BEMFEM
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
1
2
3
4
5
6
7
8
a
o
c
G

/
(

x
i 2

k
sz
z
o
)
s/d=2
s/d=5
s/d=10
Kaynia
BEMFEM
Fig. 11. Rocking impedances of 3x3 pile groups in soft soil. Comparison with Kay-
nias solution.
In the second set, several results for more slender piles (L/d = 20) are com-
pared with those of Miura et al. [8], where a 3-D formulation based on Greens
functions of cylindrical loads in layered semi-innite media is employed. The
properties used are: soil internal damping coecient = 0.05; ratio between
the material modulae E
p
/E
s
= 10
2
; ratio between densities
s
/
p
= 0.7; and
18
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
0.5
1
1.5
2
a
o
k
Gx
x
/
(
N

k
sx
x
)
Kaynia
BEMFEM
2 x 2
3 x 3
4 x 4
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0.5
1
1.5
2
2.5
3
a
o
c
Gx
x
/
(
N

k
sx
x
)
Kaynia
BEMFEM
3 x 3
4 x 4
2 x 2
Fig. 12. Horizontal impedances dierent pile groups embedded in a sti soil with
s/d=5. Comparison with Kaynias solution.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
0.5
1
1.5
2
2.5
3
a
o
k
Gz
z
/
(
N

k
sz
z
)
Kaynia
BEMFEM
4 x 4
3 x 3
2 x 2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
1
2
3
4
5
6
7
a
o
c
Gz
z
/
(
N

k
sz
z
)
Kaynia
BEMFEM
4 x 4
3 x 3
2 x 2
Fig. 13. Vertical impedances dierent pile groups embedded in a sti soil with
s/d=5. Comparison with Kaynias solution.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
1
2
3
4
5
a
o
c
G

/
(

x
i 2

k
sz
z
)
Kaynia
BEMFEM
4 x 4
3 x 3
2 x 2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0.5
1
1.5
2
2.5
a
o
k
G

/
(

x
i 2

k
sz
z
)
Kaynia
BEMFEM
4 x 4
2 x 2
3 x 3
Fig. 14. Rocking impedances dierent pile groups embedded in a sti soil with
s/d=5. Comparison with Kaynias solution.
Poissons ratios
s
= 0.4 (for the soil) and
p
= 0.25. The piles ratio between
pile separation and diameter is s/d = 5.
Lateral and vertical stiness and damping coecients of 2x2 and 4x4 pile
groups embedded in a homogeneous isotropic viscoelastic half-space are shown
19
in Figs. 15 and 16.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
0.5
1
1.5
2
a
o
k
Gx
x
/
(
N

k
sx
x
)
Miura et al.
BEMFEM
4 x 4
2 x 2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
0.5
1
1.5
2
2.5
3
3.5
4
a
o
c
Gx
x
/
(
N

k
sx
x
)
Miura et al.
BEMFEM
4 x 4
2 x 2
Fig. 15. Horizontal impedances of 2x2 and 4x4 pile groups. Comparison with Miuras
solution.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
0.5
1
1.5
2
2.5
a
o
k
Gz
z
/
(
N

k
sz
z
)
Miura et al.
BEMFEM
4 x 4
2 x 2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
1
2
3
4
5
6
a
o
c
Gz
z
/
(
N

k
sz
z
)
Miura et al.
BEMFEM
4 x 4
2 x 2
Fig. 16. Vertical impedances of 2x2 and 4x4 pile groups. Comparison with Miuras
solution.
Miura et al. also analyze the load distribution among the pile caps in 44 pile
groups. In Fig. 17, the distribution of shear and axial forces due to horizontal
and vertical excitation respectively are shown and compared with those given
in [8].
All results from Kaynia and Miura et al. have been taken directly from [1] and
[8] and may contain small errors. It can be seen that the computed values are
in very good agreement with those presented in the respective cites. It should
be noted that the results computed using the BEM-FEM model proposed here
are also very close to others (not shown in this paper for the sake of brevity)
obtained using a more rigorous 3D model in which both soil and piles are
discretized using BEM [19].
20
(a) Shear forces due to horizontal excitation (b) Axial forces due to vertical excitation
Fig. 17. Forces on pile caps a 4x4 pile group
6 Conclusion
In this paper, a three-dimensional BEM-FEM coupling model for the time-
harmonic dynamic analysis of piles and pile groups embedded in homoge-
neous isotropic viscoelastic soils has been presented. Piles are modelled us-
ing Finite Elements (FEM) as a beam according to the Bernoulli hypothesis,
while the soil is modelled using Boundary Elements (BEM) as a continuum,
semi-innite, isotropic, homogeneous, linear, viscoelastic medium. In the soil
equations, the presence of piles is replaced by internal punctual forces at the
tip and load-lines placed along the pile axis, while it is assumed that the soil
continuity is not altered by the presence of the piles. Then, the associated
unknowns at internal points (displacements and tractions) are linked to piles
variables through equilibrium and compatibility equations.
The main advantage of this model is the capacity of computing dynamic be-
haviour of piles with low computing times and low memory requirements in
comparison to other methods that need to discretize the pile surface or volume.
This way, pile groups with a big number of members can be analyzed without
diculty. Besides, once the surface (not necessarily at) has been discretized,
it does not need to be changed to analyze dierent sets of piles.
Several results of stiness have been presented and compared to well known
values taken from the literature, obtaining an excellent agreement. More cases
than these presented in this work have been tested, all of them with the
favourable conclusions. Although all results have been given for homogeneous
soils, dierent congurations including soil strata and rigid rocky beds can be
easily taken into account. Furthermore, other variables such as displacements
or internal axial and shear forces and bending moments along the pile can be
obtained without diculty.
21
Acknowledgements
The writers thank the reviewers for their valuable comments. This work was
supported by the Ministry of Education and Science of Spain through research
project BIA2004-03955-C02-02. L.A. Padron is recipient of the FPU research
fellowship AP-2004-4858 from the Ministry of Education and Science of Spain.
Authors are grateful for this support.
References
[1] A.M. Kaynia, Dynamic stiness and seismic response of pile groups, Research
Report R83-03. Massachusetts Institute of Technology. Cambridge, Mass. 1982.
[2] R.S.Y. Pak, P.C. Jennings, Elastodynamic response of the pile under transverse
excitation, J Eng Mech, ASCE 1987;113(7):1101-16.
[3] R.K.N.D. Rajapakse, A.H. Shah, On the longitudinal harmonic motion of an
elastic bar embedded in an elastic half-space, Int J Solids Struct 1987;23(2):267-
85.
[4] R.K.N.D. Rajapakse, A.H. Shah, On the lateral harmonic motion of an elastic
bar embedded in an elastic half-space, Int J Solids Struct 1987;23(2):287-303.
[5] A.M. Kaynia, E. Kausel, Dynamics of piles and pile groups in layered soil
media, Soil Dyn Earthq Eng 1991;10:386-401
[6] G. Gazetas, K. Fan, A.M. Kaynia, E. Kausel Dynamic interaction factors for
oating pile groups, J Geotech Eng, ASCE 1991;117:1531-48.
[7] A.M. Kaynia, M. Novak, Response of pile foundations to Rayleigh waves and
obliquely incident body waves, Earthq Eng Struct Dyn 1992;21:303-18.
[8] K. Miura, A.M. Kaynia, K. Masuda, E. Kitamura, Y. Seto, Dynamic behaviour
of pile foundations in homogeneous and non-homogeneous media, Earthq Eng
Struct Dyn 1994;23:183-92.
[9] A.M. Kaynia, S. Mahzooni, Forces in pile foundations under seismic loading,
J Engng Mech, ASCE 1996;122:46-53.
[10] R. Sen, T.G. Davies, P.K. Banerjee, Dynamic analysis of piles and pile groups
embedded in homogeneous soils, Earthq Eng Struc Dyn 1985;13:53-65.
[11] T.G. Davies, R.Sen, P.K. Banerjee, Dynamic behavior of pile groups in
inhomogeneous soil, J Geotech Eng 1985;111(12):1365-1379.
[12] S.M. Mamoon, A.M. Kaynia, P.K. Banerjee, Frequency domain dynamic
analysis of piles and pile groups, J Eng Mech ASCE 1990;116(10):2237-2257.
22
[13] H.B. Coda, W.S. Venturini, M.H. Aliabadi A general 3D BEM/FEM coupling
applied to elastodynamic continua/frame structures interaction analysis, Int J
Numer Meth Eng 1999;46:695-712.
[14] H.B. Coda, W.S. Venturini, On the coupling of 3D BEM and FEM frame model
applied to elastodynamic analysis, Int J Solids Struct 1999;36:4789-804.
[15] D.E. Beskos Boundary element methods in dynamic analysis: Part II (1986-
1996), Applied Mech Review 1997;50:149-197.
[16] S.E. Kattis, D. Polyzos, D.E. Beskos, Vibration isolation by a row of piles
using a 3-D Frequency domain BEM, Int J Numer Meth Eng 1999;46:713-28.
[17] S.E. Kattis, D. Polyzos, D.E. Beskos, Modelling of pile wave barriers by
eective trenches and their sreening eectiveness, Soil Dyn Earthq Eng
1999;18:1-10.
[18] F. Vinciprova, J.J. Aznarez, O. Maeso, G. Oliveto, Interaction of BEM analysis
and experimental testing on pile-soil systems., in Problems in structural
identication and diagnostic: General aspects and applications., C. Davini,
E. Viola (Editors), Springer-Verlag, 195-227, 2003.
[19] O. Maeso, J.J. Aznarez, F. Garca, Dynamic impedances of piles and groups
of piles in saturated soils, Comput Struct 2005;83:769-82.
[20] A.V. Mendonca, J.B. de Paiva, A boundary element method for the static
analysis of raft foundations on piles, Eng Anal Boundary Elem 2000;24:237-
47.
[21] A.V. Mendonca, J.B. Paiva, An elastostatic FEM/BEM analysis of vertically
loaded raft and piled raft foundation, Eng Anal Boundary Elem 2003;27:919-33.
[22] R. Matos Filho, A.V. Mendonca, J.B. Paiva, Static boundary element analysis
of piles submitted to horizontal and vertical loads, Eng Anal Boundary Elem
2005;29:195-203.
[23] R.W. Clough, J.Penzien, Dynamics of structures, McGraw-Hill, 1982.
[24] J. Domnguez, Boundary elements in dynamics, Southampton, New York,
Computational Mechanics Publications & Elsevier Applied Science; 1993.
[25] H.B. Li, G.M. Han, H.A. Mang, A new method for evaluation singular integrals
in stress analysis of solids by the Direct Boundary Element Method, Int J
Numer Meth Eng 1985;21:2071-98.
[26] F. Chirino, O. Maeso, J.J. Aznarez, Una tecnica simple para el calculo de las
integrales en el sentido del valor principal en el MEC 3D, Revista Internacional
de Metodos Numericos para Calculo y Dise no en Ingeniera 2000;16(1):77-95.
23