Engineering Analysis with Boundary Elements 27 (2003) 919–933

www.elsevier.com/locate/enganabound

An elastostatic FEM/BEM analysis of vertically loaded raft

and piled raft foundation
A.V. Mendonçaa, J.B. Paivab,*
a
Department of Structures/CTU, State University of Londrina, P.O. Box 6001, CEP 86051-990 Londrina, PR, Brazil
b
Department of Structures, São Carlos Engineering School, Av. Trabalhador Sãocarlense, 400, CEP 13566-590 São Carlos, SP, Brazil
Received 26 February 2002; revised 4 March 2003; accepted 6 March 2003

Abstract
In this paper, a static analysis of vertically loaded raft and piled raft foundations in smooth and continuous contact with the supporting soil
is presented. In this approach the finite element method (FEM) and the boundary element method (BEM) are coupled: the bending plate is
assumed to have linear elastic properties and is modelled by FEM while the soil is considered as an elastic half-space in the BEM. The pile is
represented by a single element and the shear force along the shaft is interpolated by a quadratic function. The plate– soil interface is divided
into triangular boundary elements (soil) also called cells and finite elements (plate) and the subgrade reaction is linearly interpolated across
each cell. The subgrade tractions are eliminated from the FEM and BEM algebraic systems of equations, resulting in the governing system of
equations for plate – pile– soil interaction problems. Numerical results are presented and they are close to those resulting from much more
elaborate analyses.
q 2003 Elsevier Ltd. All rights reserved.
Keywords: Boundary elements; Finite elements; Piled rafts; Foundation; FEM/BEM coupling

1. Introduction the interaction of two piles were initially obtained and

then applied to the study of larger groups of piles by
In a large number of publications that analyse plate – soil superposing the results for all piles in the group. Although
interactions, the supporting medium is represented as a elastic superposition is valid only when the piles are located
linear-elastic homogeneous half-space. In Refs. [1,2], around a circumference and submitted to the same load,
analytical solutions have been developed for a uniformly these load-transfer curves were applied to the analysis of
loaded circular plate resting on a continuum, using an general capped-pile groups. In Refs. [9,10], variational
approach where the displacements (approximated by power approaches are presented in which the soil is modelled with
series expansions) are inserted into the total potential energy the use of load-transfer curves and as a half-space,
functional in which the strain energy (related to the plate respectively. In Ref. [10], the tractions of soil – pile
and the half-space) and the work done by the external interaction are approximated by finite series and the soil
loading are taken into account. The unknowns are contributions are obtained with the help of Mindlin’s
determined by minimising the total potential energy solution [11]. In Ref. [12] a formulation is presented for
functional. Other formulations have used only the finite pile –soil interaction in which each pile is considered as a
element method [3,4]. In Ref. [5], the analysis made use of a 3D element when calculating the interface tractions and as a
combination of FEM and BEM, where a quadrilateral finite linear element for its internal tractions.
element called ACM (Adini, Clough and Melosh) [6] was In Refs. [13 – 15] a general formulation is proposed for
used to model the bending plate and a BEM was applied to the analysis of capped-pile groups in which the soil is
the soil. modelled with the use of load-transfer curves and in Ref.
Pile – soil interactions have been analysed by many [16] the interactions between the capped-pile groups and
authors [7,8], where load – displacement curves for the soil (assumed to be a half-space) are derived by an
extension of the variational approach described in Ref.
* Corresponding author. [9]. However, as in Refs. [7,8,13 – 15], the cap is
E-mail address: [email protected] (J.B. Paiva). supposed to be rigid. Only in a few published analyses
0955-7997/03/$ - see front matter q 2003 Elsevier Ltd. All rights reserved.
doi:10.1016/S0955-7997(03)00061-4
920 A.V. Mendonça, J.B. Paiva / Engineering Analysis with Boundary Elements 27 (2003) 919–933

is the plate assumed to be flexible. This is the case, for the nodes at the top of the piles, and then the subgrade
instance, in Ref. [17], where a flexible footing supported tractions are eliminated from the two algebraic systems of
on the soil and piles is analysed by classical beam equations, resulting in a system for plate – pile – soil
theory, which limits the range of applications. A interaction problems. Thus the final number of nodal points
formulation is presented in Ref. [18] where the plate is is reduced to those at the plate nodes alone. Numerical
analysed by BEM and the piles are represented by results are presented and compared with those from other
springs. The stiffness of the springs is initially deter- formulations.
mined by a computer program for the analysis of pile –
soil interaction, but no account is taken for the reaction
between the plate and the soil, thus resulting in a poor 2. Plate– pile – soil equations
representation of the problem. In Ref. [19], the plate is
analysed by FEM and the contributions of the piles are In Fig. 1, the general layout of a piled raft and the
given by load-transfer curves taken from Refs. [7,8]. tractions applied to the cap, soil, pile shafts and pile
Again, elastic superposition is used in case of piled rafts, bases are shown. The triangular elements on the plate –
although the requirements for it are not fulfilled. In Ref. soil interface, also called cells, have nodes defined at
[20], as in Ref. [18], the piles are represented by springs their corners and the tractions on them at different
and once more a computer program is used to evaluate points are approximated by a linear function. Each pile
their stiffness. This formulation is implemented in an is represented by one element and a quadratic function
finite difference method program and load –displacement is assumed to represent the tractions along its shaft.
curves are also used to obtain the vertical displacement There are three nodes along the pile shaft and one at
of the soil at the junctions of the piles with the plate. In the tip of the pile with tractions uniformly distributed
a recent formulation [21], both the soil and the plate are across it.
represented by integral equations and all the interactions The tractions p3 on cells are approximated by a linear
among all the structural elements involved are function within their domains and can be written as
considered.
The purpose of this paper is to introduce a coupled p 3 ¼ F pn ð1Þ
~~
BEM/FEM formulation for the static analysis of piled where F is the matrix of domain functions and pn the nodal
rafts, in which all the interactions between the plate, the traction~ vector. Therefore ~
pile and the soil are simultaneously taken into account. 8 9
In this approach the soil and piles are represented in a >
> pi >
similar way to that reported in Ref. [21], in which the < > =
p3 ¼ ½ F1 F2 F3  pj ð2Þ
following assumptions are adopted: the soil is treated as >
> >
: > ;
a linear-elastic homogeneous half-space, represented by pk
integral equations, using Mindlin’s fundamental solution
described in Ref. [11]; each pile is represented by a
single element with three nodal points and the traction
along it is approximated by a quadratic function. At the
bottom of the pile, the traction is assumed to be constant
over its cross-section and one more nodal point is located
there. With this approximation, only a few unknowns are
associated with each pile. As an alternative to the
application of a BEM technique in the raft analysis [21],
in the present approach the raft is analysed by FEM
using two finite elements. In addition, in the analysis of
the plate –soil interaction, the interface is divided into
triangular elements that coincide with the finite element
mesh on the plate. The subgrade reaction is assumed to
vary linearly across each element and the elastic half-
space is treated as medium that is not disturbed by the
piles. It is also supposed that the plate does not detach
from the soil and that the raft domain is only subjected
to vertical static loads.
The systems of equations from the BEM (soil and piles)
and FEM (bending plate) are coupled through matrix
operations, so that the contributions from the pile nodes
(located within the half-space) are written as functions of Fig. 1. Plate–pile–soil system.
 A.V. Mendonça, J.B. Paiva / Engineering Analysis with Boundary Elements 27 (2003) 919–933 921

where the domain functions ½Fn  in rectangular co-ordinates the shear modulus and ns is the Poisson ratio of the half-
ðx1 ; x2 Þ are given by space.
By writing Eq. (6) for all nodes and performing all the
F1 ¼ ½ðx2j 2 x2k Þx1 þ ðx1k 2 x2j Þx2 þ ðx1j x2k 2 x1k x2j Þ=Dt numerical integrations involved, the following set of linear
ð3aÞ equations can be obtained
Us ¼ LP ð8Þ
F2 ¼ ½ðx2k 2 x2i Þx1 þ ðx1i 2 x2k Þx2 þ ðx1k x2i 2 x1i x2k Þ=Dt ~ ~~
ð3bÞ where Us is the vector that contains the nodal displacements
and P ~is the vector of nodal tractions. L is the influence
F3 ¼ ½ðx2i 2x2j Þx1 þðx1j 2x1i Þx2 þðx1i x2j 2x1j x2i Þ=Dt ð3cÞ ~ of vector P:
matrix ~
~
The axial strain at generic point z located along the pile
Dt ¼ x1i ðx2j 2x2k Þþx1j ðx2k 2x2i Þþx1k ðx2i 2x2j Þ ð3dÞ
shaft can be written as
where xab ða ¼ 1;2; b ¼ i;j;kÞ are the co-ordinates of the 2NðzÞ
nodal points of a cell. u3;z ¼ ð9Þ
Ap E p
The tractions along the shaft can be written as
where Ap and Ep are, respectively, the cross-sectional area
tp ¼ f tn ð4Þ
~~ and Young’s modulus of the pile. NðzÞ is the axial force
where f and fn are, respectively, the matrices of domain acting in the pile at depth z; given by
~ and ~the vector of nodal tractions along the pile.
functions ðz
The expression (4) can be written alternatively as: NðzÞ ¼ ti dGpi þ sc Ap ; i ¼ 1; 2; 3 ð10Þ
L
8 9
>
> tl >
< > = or in the discretized form by
tp ðsÞ ¼ ½ f1 f2 f3  tm ð5Þ  ð1  ð1 
>
> >
: > ; NðzÞ ¼2pRp up1 lJldh tp1 þ up2 lJldh tp2
tn h h
ð1  
where f1 ðjÞ ¼ 1=2ð9j2 2 9j þ 2Þ; f2 ðjÞ ¼ 29j2 þ 6j; þ up3 lJldh tp3 þ sc Ap ð11Þ
f3 ðjÞ ¼ 1=2ð9j2 2 3jÞ with j ¼ ðx3 ðsÞ=LÞ and range 0 # h
j # 1; L is the length of the pile. where Rp is the pile radius and J is the Jacobian.
The integral equation giving the vertical displacement u3 After integrating Eq. (9) and imposing the boundary
at a generic point s in the half-space (for the case in which condition u3 ¼ up at z ¼ 0; the expression for axial
the contributions from the plate – pile – soil interfaces are displacement is obtained. Next, the displacements of all
taken into account) is four nodal points along the pile can be rearranged as
ð Np ( ð follows
X
u3 ðpÞ ¼ p
u33 ðp; sÞp3 ðsÞdGs þ up33 ðp; sÞtp ðsÞdGpi 8 9 8 9 8 9
Gs Gpi >
> up >
> >
> 1>> > > f1 >
>
i¼1 >
> >
> >
> > > > > > >
) > >
< up1 = >
<1= >
> < f2 >
=
ð
þ up33 ðp; sÞsc ðsÞdGbi ð6Þ >
¼ up þ ð12Þ
Gbi >
> up2 >
>
>
>
>
> 1>>
>
>
>
> f3 >
>
>
>
> >
> >
> > > > > > >
: up3 ; : ; : ;
where tp ðsÞ is the shear force along the pile shaft, sc ðsÞ is the 1 f4
traction at the tip of the pile and NP is the number of piles. In
where up ; up1 ; up2 ; up3 are the displacements at the top, in
Eq. (6), up33 ðp; sÞ is Mindlin’s fundamental solution (for
the shaft and at the pile base. The functions fi are given
direction x3 at load and field points) given in Ref. [11];
by
Gs ; Gpi ; Gbi are the domains of the plate/soil interface, the 8 9 2 38 9
shaft and the base of ith pile, respectively >
> f1 >
> b11 b12 b13 b14 > > tp1 >
>
( >
> > > 6 7>> >
>
>
< f2 >
= 6 7 > >
1 r32 ð3 2 4ns Þ 6czR23 6 b21 b22 b23 b24 7 tp2 = <
p
u33 ¼ þ þ ¼66 7 ð13Þ
> 7
16pGs ð1 2 ns Þ r 3 r R5 >
> f3 >
>
> 6 b31 b22 b33 b34 7>
4 5 >
> tp3 >
>
>
) >
> > > >
8ð1 2 ns Þ2 2 ð3 2 4ns Þ ð3 2 4ns ÞR23 2 2cz : > ; >
: >
;
þ þ ð7Þ f4 b41 b42 b43 b44 sb
R R3
pffiffiffiffiffiffi where bj1 ¼ Kf ð 14 hj 2 12 h2j þ 34 h3j 2 38 h4j Þ; bj2 ¼
where Ri ¼ xi ðpÞ 2 xi ðs0 Þ; ri ¼ xi ðpÞ 2 xi ðsÞ; R ¼ Ri Ri ; Kf ð 4 hj 2 h3j Þ; bj3 ¼ Kf ð 34 hj þ 14 h3j 2 38 h4j Þ; bj4 ¼ Kb hj ;
3 4
pffiffiffiffi
r ¼ ri ri and the index i ¼ 1; 2; 3: c ¼ x3 ðsÞ . 0; z ¼ Kf ¼ ð2L2 =Ep Rb Þ; Kb ¼ ðL=Ep Þthe dimensionless co-ordi-
x3 ðqÞ . 0; where xi ðpÞ and xi ðsÞ are the co-ordinates of the nate hj being assigned the values 0, 1/3, 2/3, 1 when the
load and field points, respectively. Point s0 is the image of source point stands at the three shaft nodes and the base
s through the half-space surface ðx3 ¼ 0Þ: In Eq. (7), Gs is node, i.e. j ¼ 1; 2; 3 and 4; respectively.
922 A.V. Mendonça, J.B. Paiva / Engineering Analysis with Boundary Elements 27 (2003) 919–933

In Eq. (8), the rows of matrix L assigned to pile k can be z ¼ 0; by the cross-sectional area of pile Ap
written as follows: ~
8 9 2 38 9 NT
> .. > .. .. .. > .. > sT ¼ ¼ C1 t1 þ C2 t2 þ C3 t3 þ C4 sc ð17Þ
>
> . > >
> 6 . . . >
> . > > Ap
>
> 7> >
> >
> 6 7>> >
>
>
> umi > > 6
> 6 ·· · bmi ·· · bmp bm;p1 bm;p2 bm;p3 ·· · 7> 7 >
> pmi > >
>
> > 6 7 > >
> where C1 ¼ L=2Rp ; C2 ¼ 0; C3 ¼ 3L=2Rp and C4 ¼ 1:
>
> >
> 6 7 >
> >
> Each row of the matrix assigned to pile k in Eq. (16) as
>
> .. >> 6 . . . 7 >
> . >
>
>
> . >
> 6 .
. .
. .
. 7 >
> .
. >
> well as the corresponding entries in the vector on the left
>
> >
> 6 7 >
> >
>
> = > 66 7 > > side of Eq. (16) can be multiplied by the constants C1 ; C2 ; C3
< 7 < =
up 6 bqi bqp bq;p1 bq;p2 bq;p3 7 t1
>
¼6
> 6
7
7> > and C4 ; respectively. After making a linear combination of
>
> up1 > > 6 bri brp br;p1 br;p2 br;p3 7>> t2 > > these four rows, a new system is obtained
>
> > 6 7 > >
>
>
>
>
> 6 7>>
>
>
>
> 8 9 2 38 9
>
> u >
> 6 7 >
> t >
>
> p2 > 6 ·· · b si ·· · b sp b s;p1 b s;p2 b s;p3 ·· · 7 > 3 > > .. > .. > .. >
>
> >
> 6 7 >
> >
> >
> . >> 6 . 7>> . >>
>
> >
> 6 7 >
> > >
< = 6 > 7>< > =
> u p3 > 6
6 b b b b b 7
7 > s c >
> 6 7
>
> >
> 4
ti tp t;p1 t;p2 t;p3
5>>
> >
> s
p
¼ 6 · · · aqp · · · 7 uT ð18Þ
> > >
> > >
: . > >
> > > 6 7>> >
T
: . ; . . . . . ; > > 4 5 > >
>
. . . >
. . . . . : .. > ; .. >
: .. >
;
. . .
ð14Þ
By imposing the compatibility of displacements between When this process is repeated for all piles, the system of
the piles and soil at the nodes along the shaft, a system of equations in Eq. (16) becomes
equations similar to Eq. (14) is obtained, now including the 8 1 9 2 38 1 9
>
> p > p
· ·· ap1p > > wcel >
flexibility of the piles > cel >
> > 6 a11 a12 ·· · a1n a1k
> 7 >
>
>
>
>
>
> > >
p 7> w2 >
8 9 2 38 9 >
> p2cel > > 6a a
6 21 22 a2n ap2k a2k 7> >
>
.. .. .. > 7> >
.. > cel
>
> > >
> .. > > >
> >
> 6 >
> >
>
>
> . > > 6 . . . 7>> . > > >
> .. > > 6 7 > >
> > 6 7> > > 6 .. .. .. 7> > . >
>
> >
> 6 ·· · b ·· · b 7>> >
> >
> . > >
> 6 . . . 7 > . >
>
. >
>
>
> u mi >
> 6 b b b ·· · 7 >
> p >
> >
> > 6 7> >
>
> >
> 6 mi mp m;p1 m;p2 m;p3
7>> mi >
> >
>
>
> 6 7>
p 7>>
>
>
< pcel > = 6 >
f f
>
> >
> 6 7 >
> >
>
p
6 af 1 af 2 ·· · afn afk <
· ·· afk 7 cel =
w
>
> . >
. >
> 66 . . . 7 >
> . >. > 6 7
> . . . . 7> . > ¼6 ð19Þ
.. 7
> > 6 . . . 7 > > >
> > 7> > > .. > > 6 .. 7>> .. >
6 . >
> > > ..
<u = 6     7< t > = >
>
>
. >
> 6
> 6 . . . 7
7>
>
> >
>
>
p 6 bqi bqp bq;p1 bq;p2 bq;p3 7 > > 6 7>> >
¼6 7 1
>
> k >
> 6 p 7> k >>
>
> > 6 7> > > sT > akn apkk akp 7> u >
> up > 6 7> t2 > > > 6 ak1 ak2 > T >
>
>
>
>
> 6 bri b rp b r;p1 b r;p2 b r;p3 7>>
>
>
>
>
>
>
>
>
> 6 7>
7>> >
>
>
> >
> 6 7>> >
> >
> .. >> 66 7>> . >
>
>
>
> u p
>
>
>
6
6 ·· · b ·· · b 
b 
b b ·· · 7
7 >
>
> t >
>
>
>
> . >>
> 4 5>> . >
. >
>
> > 6 si sp s;p1 s;p2 s;p3
7 >
3
> >
> >
> >
> >
>
>
> >
> 6 7 >
> >
> : p; ap1 ap2 ·· · apn appk · ·· app : up ;
p
>
> u >
p > 6 b b 
b 
b 
b 7 >
> s c >
> sT
> >
> > 46 ti tp t;p1 t;p2 t;p3 7 >
> >
>
T
> > 5> >
>
: . ;. > .. .. .. >
: . >. ;
. . . . . In Eq. (19), the equations that referred to nodal points
located along the pile are now transformed into equivalent
ð15Þ ones referring to nodal points located on the top of the piles,
When system (15) is inverted the set of equations referring algebraically represented by
to the pile k can be written as
8 9 2 38 9 P ¼ HUs ð20Þ
> .. > .. .. .. > .. > ~ ~~
>
> > > >
> . > 6 . . . 7> . >
>
>
>
>
> 6 7>>
>
>
>
>
where P and Us are vectors, the former composed of the
>
> pmi >
>
> 6
> 6 ·· · ami ·· · amp am;p1 am;p2 am;p3 ·· · 7> 7 >
> umi >
>
> tractions~ at the~cells nodes and top of the piles, the latter of
>
> >
> 6 7 >
> >
>
>
> . > >
> 66 7 >
> . > > the corresponding displacements.
>
> > . . . 7 >
> >
>
>
.
. > >
6
6
.
. .
. .
. 7
7 >
> . >
. >
>
>
> > 6 7> >
< > = 6 7><u > =
t1 6 aqi aqp aq;p1 aq;p2 aq;p3 7 p 3. Plate bending analysis
¼6 7
>
> > 6
> 7>
7>
>
>
>
> t2 >
> > 6
> 6 ari arp ar;p1 ar;p2 ar;p3 7>> up >
> >
>
>
> >
> 6 7 >
> >
> The bending plate is analysed by FEM, using the flat
>
> > 6 7 > >
> t 3 >> 6 ·· · a si ·· · a sp a s;p1 a s;p2 a s;p3 ·· · 7>> u p >
> triangular elements DKT (discrete Kirchhoff Theory) and
>
> >
> 6 7 >
> >
>
>
> >
> 6 7 >
> >
> the HSM (hybrid stress model), whose formulations have
> >s > 66 a a a a a 7 > u >
>
>
c
>
ti tp t;p1 t;p2 t;p3 7 >
>
p
>
>
> > 4 5 > > been thoroughly discussed elsewhere [22]. The final nodal
>
: . ;. > . . . >
: . >
. ;
. .
. .
. .
. . variables used in the formulation of both elements are
represented by two rotations and one translation in each
ð16Þ
corner of the triangle as shown in Fig. 2.
The normal stress at the pile top, sT ; can be calculated by The loads and interface tractions in the plate – pile – soil
dividing the total transferred load NT ; expression (10) with system are indicated in Fig. 1.
 A.V. Mendonça, J.B. Paiva / Engineering Analysis with Boundary Elements 27 (2003) 919–933 923

Fig. 2. Final nodal variables of finite element DKT and HSM.

The potential energy of the plate – soil system is given by element i; expressed as
ð ð 2 3
2 1 1
p ¼ pp 2 gðsÞu3 ðsÞdVg þ pðsÞu3 ðsÞdVel
A 661
7
17
Vg Vel
Q¼ 4 2 5 ð24Þ
~ 12
Np ð
X 1 1 2
þ siT ðsÞwi ðsÞdVT ð21Þ
i¼1 VT
where A is the ith element area.
The dimensions of vectors U  c and P c are smaller than
where pp is the strain energy of the plate;gðsÞ and Vg are the ~ the last ~ two vectors the
external loading and its zone of action, respectively; pðsÞ is those of Uc and Fc because in
~ ~
rotations are also included as nodal values. In order to
the subgrade reaction and Vel its corresponding domain; sT
is the plate – pile contact stress and VT its loading domain; harmonise the dimensions of these vectors, columns and
u3 is the vertical displacement. rows of zeroes are inserted into matrix Q so that it becomes
 Thus, the vector U
Q:  c can be written as ~the vector Qc and in
The discretized form of the energy functional (21) can be
~ analogous way P~ becomes P ::
an ~
written as follows ~
c
~
c

Nel ð
X Nel ð
X UTc ¼ { Wi f1i f2i Wj f1j f2j Wk f1k f2k }
p ¼pp 2 i
g ðsÞwi ðsÞdVg þ i
p ðsÞwi ðsÞdVel ~
Vgi Vel ð25Þ
i¼1 i¼1

XðNp
þ siT ðsÞwi ðsÞdVT ð22Þ PTc ¼ { pi 0 0 pj 0 0 pk 0 0} ð26Þ
i¼1 VT ~

where Nel and Np are the number of planar finite elements After expanding the dimensions of matrix Q; Eq. (23) can be
expressed as ~
and piles, respectively; gi ; pi and siT are nodal values for the
external loading, the subgrade reaction and plate – pile
contact stress related to the finite element i: p1 ¼ 1
2 UTc Kc Uc 2 UTc Fc þ U
 Tc Q
 Pc ð27Þ
~ ~ ~ ~ ~ ~ ~~
The mesh of finite elements on the plate and mesh of
boundary elements on the surface of the half-space coincide. When the contributions of all elements are computed, an
The total energy of one finite element of the plate – soil algebraic system can be obtained from the total energy
system is given by
KU ¼ F 2 RP ð28Þ
p1 ¼ 1
UTc Kc Uc 2 UTc Fc þ UTc QPc ð23Þ ~~ ~ ~~
2
~ ~ ~ ~ ~ ~ ~~
where U; F are vectors containing the displacements and
where Uc is the vector that contains displacements and ~ ~nodal forces for all finite element nodes, P is the
equivalent
rotations~ of one finite element, Kc is the stiffness matrix for ~
vector that contains the traction at all boundary element
one finite element, Fc is the nodal ~ vector formed from the nodes, K is the stiffness matrix of the plate and R is a matrix
external loading. U ~
 c is the displacement vector of one with all~ the contributions of the boundary ~ elements
~
boundary element assigned to the plate – soil and plate – pile contributions.
interaction surfaces; P c is the analogous vector for the In the plate –pile– soil analysis, an analogous procedure
~
plate – pile contact tractions. Q is a matrix that transforms to that used with the plate –soil problem can be adopted,
the traction into an equivalent ~ nodal force assigned to except that the transformation matrix Q has the following
~
924 A.V. Mendonça, J.B. Paiva / Engineering Analysis with Boundary Elements 27 (2003) 919–933

where a is the angular co-ordinate relative to the axis of

each pile shown in Fig. 3.

4. BEM/FEM coupling

The systems of equations for FEM and BEM are

coupled by eliminating the unknown plate – soil subgrade
traction. In the system of Eq. (20), columns of zeroes are
inserted into H to raise its dimension from that of Us to
that of U : ~ ~
~
P¼H U ð30Þ
~ ~~
where H  is matrix H after the expansion.
Fig. 3. Central angle of pile. ~
Substituting this ~expression for P in Eq. (28), the
configuration following equation can be written ~
0 1
R2p U
K U ¼ F 2 RH ð31Þ
B 6a A 1 1C ~~ ~ ~~~
AB C
Q¼ B C ð29Þ
~ 2B
@ 0 2 1C
A
or

0 1 2 K2 U ¼ F ð32Þ
~ ~ ~

Fig. 4. Mesh used in the analysis.

Fig. 5. Vertical displacements at points along line AB.

 A.V. Mendonça, J.B. Paiva / Engineering Analysis with Boundary Elements 27 (2003) 919–933 925

Fig. 6. Bending moments mx along AB.

where K2 is the stiffness matrix of the plate –pile– soil 5. Numerical results
system, ~obtained by combining FEM and BEM and given
by: In this section two groups of soil –structure interaction
systems are presented. The first refers to plate – soil
 interaction, and two examples are shown. In the second
K2 ¼ K þ RH ð33Þ
~ ~ ~~ group, various configurations of piled rafts are analysed.

After the solution of the final system of Eq. (32), the 5.1. Plate –soil interaction
unknown vector U containing the transverse displacements
and slopes of the~ raft at the finite elements nodes and the 5.1.1. Square plate
pile-top deflections, is determined. With this vector, the Consider a square plate submitted to a uniform transverse
raft – soil and pile – raft interaction forces can be evaluated load g and resting on a linear-elastic half-space. The length
from Eq. (19) and the interaction shear stresses on shaft and of the side of the plate is L ¼ 12 m and its thickness is
normal stress at the tip of the piles can also be determined t ¼ 0:1 m: Young’s modulus for the half-space is Es ¼
from Eqs. (14) and (16). 0:26 £ 106 kN=m2 and its Poisson’s ratio is ns ¼ 0:3:

Fig. 7. Bending moments my along AB.

926 A.V. Mendonça, J.B. Paiva / Engineering Analysis with Boundary Elements 27 (2003) 919–933

Fig. 8. Circular plate simply supported on soil.

The numerical results obtained by the present FEM/BEM approach reported in Messafer and Coates [5] (in which the
approach with a discretization of the plate – soil interface plate is modelled with the quadrilateral finite element ACM
into 128 finite and boundary elements, as shown in Fig. 4 are [6] and boundary elements with constant interpolation of
compared with solutions of a BEM technique described in quantities are used to discretize the surface of the semi-
Paiva [24] and with results from the coupled FEM/BEM infinite medium). The calculated displacements at points

Fig. 9. Displacements versus plate rigidity.

Fig. 10. Displacements in function the radius plate.

A.V. Mendonça, J.B. Paiva / Engineering Analysis with Boundary Elements 27 (2003) 919–933 927

Fig. 11. Piled circular plate and mesh used in the analysis.

Fig. 12. Displacements along AB.

Fig. 13. Bending moments mx along CD.

928 A.V. Mendonça, J.B. Paiva / Engineering Analysis with Boundary Elements 27 (2003) 919–933

Fig. 14. Bending moments along CD.

along line AB are shown in Fig. 5 and bending moments the circular plate are radius a ¼ 8:0 m and Poisson’s ratio
along the same line are plotted in Figs. 6 and 7, for a rigidity n ¼ 0:3: In Fig. 9, the transverse displacement of point A is
factor log X ¼ 1:08 in which the argument X is described in plotted versus the logarithm of the relative stiffness, while
Ref. [23] as X ¼ 180pðEs =Ep Þða=tÞ3 with a ¼ ðL=6Þ: the variation of transverse displacement with distance from
There is good agreement in the results for both the the centre of the plate is shown in Fig. 10.
displacements and the bending moments between the A good consistency can be noted between the FEM/BEM
present approach and the full BEM procedure, however, coupling technique used in the present work and the
there is poor agreement, mainly for the bending moments, analytical solutions reported in Ref. [1].
when the finite element ACM is used to model the plate.
These weak results are partially explained by the neglect of
the C1 continuity requirements in the development of the 5.2. Plate– pile – soil interaction
ACM formulation. In addition, Messafer and Coates [5]
used a poorer interpolation than in the present FEM/BEM In this section, the interaction between the piled raft and
model for the contact stresses at the plate – soil interface. soil is analysed for particular cases of raft geometry and
pile-groups configurations.
5.1.2. Circular plate
A circular plate of radius a and thickness t; lying on the
surface of a semi-infinite medium, is analysed in this
example. The entire surface of the plate is assumed under a
uniform vertical loading g and its Young’s modulus is
represented by E: The half-space has Young’s modulus Es
and Poisson ratio ns : A plate – soil relative stiffness Kp is
described in Ref. [1] as Kp ¼ ð1 2 n2s ÞðE=Es Þðt=aÞ3 and the
displacements are expressed in dimensionless form as
vðrÞ ¼ ðEs =gað1 2 n2s ÞÞw in which w and v are the
dimensional and dimensionless displacements; r ¼ r=a is
the dimensionless form of a generic position r in the range
½0; a; 0 # r # a:
The plate is discretized into a mesh of 184 triangular
elements (DKT and HSM), as shown in Fig. 8, and analysed
for the relative stiffness values of 0.1, 1.0 and 10.
The results of the present approach are compared with
those reported in Zaman et al. [1] in which the solutions
were obtained analytically, based on an energy approach
and transverse displacements were written in terms of a
power series. The geometrical and mechanical properties of Fig. 15. Strip on four piles and the strip–soil interface discretization.
 A.V. Mendonça, J.B. Paiva / Engineering Analysis with Boundary Elements 27 (2003) 919–933 929

Fig. 16. Displacements at points along AB (rigid piles).

5.2.1. Circular raft on piles displacements and bending moments are plotted at points
Consider a circular plate supported on piles and a half- along the lines AB and CD, respectively.
space as shown in Fig. 11. The geometric and mechanical In Fig. 12, it can be seen that there is a good level of
properties of the plate are diameter D ¼ 16:0 m; thickness agreement in the results for transverse displacements in the
t ¼ 0:1 m; Young’s modulus E ¼ 2:0 £ 1010 kN=m2 and two finite element models. The displacements are more
Poisson’s ratio n ¼ 0:2: For the soil, Young’s modulus is affected by flexibility of piles at the junctions of the raft and
Es ¼ 2:0 £ 106 kN=m2 and Poisson’s ratio is ns ¼ 0:5: All the tops of the piles. As expected, with a rigid pile the
piles have the same length L ¼ 15:0 m and diameter Dp ¼ bending moments have higher values than those correspond-
0:30 m: When the piles are assumed to be flexible, their ing to a flexible pile. In addition, it can be seen in Figs. 13
Young’s modulus is Ep ¼ 4:0 £ 109 kN=m2 : The mesh of and 14 that the values of bending moments from the DKT
184 triangular elements is shown in Fig. 11. The authors analysis are generally lower than those from the HSM
proposed this example to study the influence of the rigidity model. The authors believe this is due to the differences
of the piles on the behaviour of a flexible raft using the finite in the hypotheses and numerical treatment used in the
elements DKT and HSM. In Figs. 12 –14, the results for formulations of these finite elements. For instance, in

Fig. 17. Displacements at points along AB (flexible piles).

930 A.V. Mendonça, J.B. Paiva / Engineering Analysis with Boundary Elements 27 (2003) 919–933

Young’s modulus E ¼ 2:0 £ 1010 N=m2 and its Poisson’s

ratio n ¼ 0:3: The soil has a Young’s modulus Es ¼
2:0 £ 106 N=m2 and Poisson’s ratio ns ¼ 0:5: The mesh
consists of 96 finite elements modelling the plate and an
equal number of boundary elements to discretize the
plate –soil interface, as shown in Fig. 15. A uniform unit
load is applied across the whole surface of the plate. In
Fig. 16, the results of displacements (in the case of rigid
piles) obtained by the present FEM/BEM coupling
technique are compared with BEM solutions reported in
Paiva [24], in which the strip – soil interface was
discretized into 96 triangular boundary elements and
the boundary of the raft was divided into 36 straight
boundary elements with linear interpolation for boundary
quantities of the bending plate. In Fig. 17, the piles are
flexible and the results from the present FEM/BEM
approach performed by DKT and HSM models are
shown. In Figs. 16 and 17, a very good agreement
between the analyses is observed.

Fig. 18. Layout for the raft and piles. 5.2.3. Rectangular raft
In this example a piled raft with dimensions 30:0
m £ 15:0 m £ 0:25 m; Young’s modulus E ¼ 2:0 £ 107
the HSM model it is assumed that the bending moments kN=m2 and Poisson’s ratio n ¼ 0:2; is supported on soil
have a linear distribution in the domain of the element while and on five or nine piles, as shown in Fig. 18. In the raft
in DKT the bending moments are obtained from the stiffened by five piles, the set of piles consist of groups
interpolation of quantities approximated along the sides of P1 and P4 : The mechanical properties of the soil are
the element [22]. Young’s modulus Es ¼ 2:0 £ 105 kN=m2 ; and Poisson’s
ratio ns ¼ 0:5: All the piles are of length 15 m and
5.2.2. Piled strip diameter Dp ¼ 0:3 m and, when they are considered to be
In this example, a rectangular raft with a large flexible, their Young’s modulus is Ep ¼ 2:0 £ 107 kN=m2 :
length/width ratio and supported on piles is analysed. The mesh used is composed of 256 triangular finite
This long piled strip is shown in Fig. 15. The elements modelling the plate and an equal number of
displacements calculated at points along the plate, triangular boundary elements for the plate – soil interface,
stiffened with four piles, are shown in Fig. 16. The as shown in Fig. 19. A uniform load g ¼ 30:0 kN=m2 is
dimensions of the strip are 25 m £ 2:5 m £ 0:079 m; its applied across the whole surface of the plate. The results

Fig. 19. Raft –soil interface discretization.

 A.V. Mendonça, J.B. Paiva / Engineering Analysis with Boundary Elements 27 (2003) 919–933 931

Fig. 20. Displacements at the points along AB.

of the present analysis are compared with solutions displacements on the line AB. In Fig. 21, the influence of the
obtained from other formulations (when they are rigidity of the piles on the behaviour of the flexible raft is
available in the literature). In Figs. 20 and 21, transverse shown for pile groups composed of five or nine piles. The
displacements at points along the lines AB and CD is displacements are strongly affected by increasing the
shown, respectively. The plots in Fig. 20 were obtained Young’s modulus of the piles. The responses obtained
for the raft supported on nine rigid piles and the present with the DKT and HSM models are close. In Figs. 22 and 23,
results are compared with BEM solutions reported in the bending moments at points along line EF are obtained
Ref. [24] in which 256 triangular boundary elements for the cases in which the piles are assumed to be rigid or
were used for the raft –soil interface and the boundary of flexible for pile group configurations with five and nine
the plate was discretized into 48 straight boundary piles.
elements with linear interpolation of boundary quantities From Figs. 20 and 21, it can observed, as
of the bending plate. expected, that increasing the number of piles in a
A very good agreement between the FEM/BEM model group in general leads to a decrease in the values of
and the BEM approach can be observed in Fig. 20 for the transverse displacements, i.e. the structural system

Fig. 21. Displacements at the points along CD.

932 A.V. Mendonça, J.B. Paiva / Engineering Analysis with Boundary Elements 27 (2003) 919–933

Fig. 22. Bending moments mx along EF (rigid piles).

Fig. 23. Bending moments mx along EF (flexible piles).

becomes more rigid. Also it can be noted, in Figs. 22 6. Conclusions

and 23, that the differences between the DKT and
HSM models in their responses for bending are more A coupled BEM/FEM formulation for the analysis of
marked in the rigid pile case. The authors believe flexible piled rafts has been presented in which all the
these differences are due to the concentration of stress interactions between the plate, the pile and the soil are
at certain regions of the raft (mainly near the pile – considered. In this paper two types of problem were
raft connections) which is more pronounced when the analysed. The first concerns raft – soil interactions in
piles are assumed to be rigid. The large variation of which the rafts have circular and rectangular plan form.
bending moments across these critical regions is taken The results obtained here are close to numerical and
into account in different ways by the mathematical analytical solutions reported in the literature. The second
approaches of the HSM and DKT models. case analysed by the present FEM/BEM approach is that of
 A.V. Mendonça, J.B. Paiva / Engineering Analysis with Boundary Elements 27 (2003) 919–933 933

piled raft foundations. In these analyses the transverse [9] Shen WY, Chow YK, Yong KY. A variational approach for vertical
displacements had very close values in both finite element deformation analysis pile groups. Int J Numer Anal Meth Geomech
1997;21(11):741–52.
models, but this performance was not observed in the
[10] Shen WY, Chow YK, Yong KY. A variational approach for vertically
bending moment distribution, specially in the cases where loaded pile groups in an elastic half-space. Géotechnique 1999;49(2):
the pile is assumed to be rigid. The authors believe the rigid 199–213.
piles provoked higher stress concentration in the plate than [11] Mindlin RD. A force at the interior point of a semi-infinite solid.
flexible piles and that the methodological differences in the Physics 1936;7:195 –202.
mathematics used in DKT and HSM models to represent the [12] Xu KJ, Poulos HG. General elastic analysis of piles and pile groups.
Int J Numer Anal Meth Geomech 2000;24:1109–38.
variation of bending moments may be one of the main
[13] Butterfield R, Banerjee PK. The elastic analysis of compressible piles
reasons for the discrepancies observed in the bending and piles groups. Géotechnique 1971;21(1):43–60.
moment values in the cases studied. [14] Butterfield R, Banerjee PK. The problem of pile group-pile cap
interaction. Géotechnique 1971;21(2):135–42.
[15] Kuwabara E. An elastic analysis for piled raft foundations in a
References homogeneous soil. Soils Found 1989;29(1):82–92.
[16] Shen WY, Chow YK, Yong KY. A variational approach for the
analysis of pile group–pile cap interaction. Géotechnique 2000;50(4):
[1] Zaman MM, Issa A, Kukreti AR. Analysis of circular plate-elastic
349–57.
half-space interaction using an energy approach. Appl Math Model
[17] Brown PT, Weisner TJ. The behaviour of uniformly loaded piled strip.
1988;12:285– 92.
Footings. Soils Found 1975;15(4):13– 21.
[2] Kukreti AR, Zaman MM, Issa A. Analysis of fluid storage tanks
[18] Fatemi-Ardakani B. A contribution to the analysis of pile-supported
including foundation–superstructure interaction. Appl Math Model
raft foundations. PhD Thesis, University of Southampton, South-
1993;17:618– 31.
ampton, UK, 1987.
[3] Hemsley JA. Elastic solutions for large matrix problems in
foundation interaction analysis. Proc Inst Civil Engnr, Part 2 [19] Hain SJ, Lee IK. The analysis of flexible pile-raft systems.
1993;89:471 –94. Géotechnique 1978;28(1):65–83.
[4] Hemsley JA. Application of large matrix interaction analysis to raft [20] Poulos HG. An approximate numerical analysis of pile-raft inter-
foundations. Proc Inst Civil Engnr, Part 2 1993;89:495–526. actions. Int J Numer Anal Meth Geomech 1994;18:73–94.
[5] Messafer T, Coates LE. An application of FEM/BEM coupling to [21] Mendonça AV, Paiva JB. A boundary element method for the static
foundation analysis. In: Brebbia, Connor, editors. Advances in analysis of raft foundations on piles. Engng Anal Bound Elem 2000;
boundary methods. Southampton: CMP; 1993. p. 211 –21. 24:237–47.
[6] Clough RW, Tocher JL. Finite element stiffness matrices for [22] Batoz JL, Bathe KJ, Ho LW. A study of three-node triangular plate
analysis of plate bending. Proceedings of Conference on Matrix bending elements. Int J Numer Meth Engng 1980;15:1771 –812.
Methods in Structural Mechanics, Ohio, USA: WPAFB; 1965. [23] Cheung Y, Zienkiewicz OC. Plate and tanks on elastic foundation-an
p. 515– 45. application of Finite Element Method. Int J Solids Struct 1965;1:
[7] Poulos HG, Davis EH. The settlement Behaviour of simple axially- 451–61.
loaded incompressible piles and piers. Géotechnique 1968;18: [24] Paiva JB. Formulação do método dos elementos de contorno para
351–71. análise de interação solo-estrutura. Associate Professor Qualifying
[8] Poulos HG, Mates NS. Settlement and load distributions analysis of Manuscript, Brazil: Sao Carlos Engineering School, Sao Paulo
pile groups. Aust Geomech J 1971;G1(1):18– 28. University; 1993. [in Portuguese].