You are on page 1of 12

Doherty, J. P. & Deeks, A. J. (2003). Géotechnique 53, No.

8, 703–714

Elastic response of circular footings embedded in a non-homogeneous


half-space
J. P. D O H E RT Y  a n d A . J. D E E K S 

This study uses a semi-analytical technique (the scaled Cette étude utilise une technique semi-analytique (méth-
boundary finite-element method) to evaluate the load– ode d’élément fini à limite proportionnelle) pour évaluer
displacement response of rigid circular footings em- la réponse au déplacement de charge d’assises circulaires
bedded in a non-homogeneous elastic half-space. Results rigides enfouies dans un espace simi-infini élastique et
are presented in terms of dimensionless elastic stiffness non homogène. Nous présentons les résultats en termes
coefficients. The footings are subjected to vertical, hori- de coefficients adimensionnels de rigidité élastique. Les
zontal, moment and torsional loads. The cross-coupling assises sont soumises à des charges verticales, horizon-
between the horizontal and moment load case is also tales, de moment et de torsion. Nous évaluons également
evaluated. The variation of soil shear modulus with depth le couplage transversal entre les cas de charge horizon-
may correspond to normally consolidated sand or clay, tale et de moment. La variation du module de cisaille-
and the footing geometries considered cover a range of ment du sol en fonction de la profondeur peut
footings commonly encountered in engineering practice. correspondre à du sable ou à de l’argile à consolidation
normale et les géométries d’assises étudiées couvrent
KEYWORDS: elasticity; footing/foundations; offshore engineer- toute une gamme d’assises rencontrées couramment dans
ing; soil/structure interaction; stiffness; torsion la pratique industrielle.

INTRODUCTION O O
Foundation design principally consists of two main tasks. z d
First, an estimate of the ultimate bearing capacity is made
to obtain a safe footing design at the likely maximum load, O O
and second, the deflections of the footing under working R
loads are determined for serviceability purposes or for
soil–structure interaction studies. Accounting for the inher- (a) (b) (c) (d)
ent non-linear mechanical behaviour of soil in predicting the
load–displacement response of footings is an onerous task
Fig. 1. Footing geometry for three embedment conditions: (a)
(Selvadurai, 1979). Consequently a simplified linear elastic case 1; (b) case 2; (c) case 3; (d) case 4
analysis is often preferred, as it can provide reasonable
estimates within the range of working loads (Schofield &
Wroth, 1968): that is, loads that are remote from those are all measured relative to the reference point O, as shown
required to cause foundation failure (Martin, 1994). It is in Fig. 1. For cases 1 and 2 this point corresponds to the
most convenient if a previously derived elastic solution can centreline of the footing at depth d, whereas for cases 3 and
be applied that simply requires substitution of the site- 4 this point is located at the intersection of the ground
specific elastic parameters, shear modulus (G) and Poisson’s surface (z ¼ 0) and the centreline of the footing. Fig. 2
ratio (). This can be done with confidence only if an shows the positive loading convention adopted and the
existing elastic solution reflects the geometry of the problem corresponding displacements.
under consideration. This study extends the range of avail- The elastic response of rigid circular footings subjected to
able elastic solutions for problems commonly encountered in V, H, M and T loading on the surface of a homogeneous
engineering practice. The elastic response of embedded elastic half-space has been examined extensively, and
circular footings of radius R subjected to vertical, horizontal, closed-form analytical solutions have been developed. These
moment and torsional loads in bonded contact with a non- solutions are summarised by Poulos & Davis (1974). Most
homogeneous half-space is examined. The geometry and practical geotechnical engineering situations, however,
embedment conditions considered are shown in Fig. 1. Case
1 represents a circular footing at the bottom of an open
trench, case 2 represents a fully embedded footing, case 3
represents an embedded footing (or short pile) with full d z
sidewall–soil contact, and case 4 represents a skirted foun-
dation (or caisson), also with full sidewall–soil contact. O R
The vertical (V ), horizontal (H ), moment (M ) and tor- w M
sional (T ) loads applied and the corresponding vertical (w), H
O θT
horizontal (u), moment (Ł M ) and torsional (ŁT ) deformations O θM
T
V
Manuscript received 24 September 2002, revised manuscript
accepted 5 June 2003
u
Discussion on this paper closes 1 April 2004, for further details see
z
p. ii.
 School of Civil and Resource Engineering, The University of
Western Australia, Crawley, Western Australia. Fig. 2. Loads and deformations

703
704 DOHERTY & DEEKS
involve footings embedded at a finite depth (d ) below the linear load–displacement behaviour of the footing. This
soil surface. In this situation, formulation of fundamental procedure allows realistic characterisation of footing behav-
elastic solutions is considerably more complex, and various iour under combined loading, which can be implemented
embedment conditions can be assumed, four of which are easily in a plane-frame analysis of an offshore gravity plat-
shown in Fig. 1. Several researchers have studied the vertical form (such as a jack-up).
response of rigid footings/anchors embedded in a homoge- In nature, most deep soil deposits exhibit an increase in
neous elastic half-space and have developed analytical solu- elastic stiffness with depth, so solutions taking this into
tions: Hunter & Gamble (1974) examined a footing fully account are more appropriate. Several authors have consid-
submerged (case 2) in an incompressible elastic half-space ered this problem and have presented analytical solutions for
and considered the effect of de-bonding on the tension side rigid surface footings. Gibson (1967) presented a fundamen-
of the soil–footing interface, whereas Selvadurai (1993) tal study of an incompressible elastic half-space with a
considered a range of values of Poisson’s ratio but only the linearly increasing shear modulus with depth subjected to a
fully bonded condition. The torsional response of embedded vertically loaded surface footing. Booker et al. (1985) pre-
foundations has also been considered: Poulos (1975) and sented analytical solutions for the vertical and moment
Randolph (1981) studied the torsional response of piled loading of smooth circular footings on a non-homogeneous
foundations (case 3), whereas Rajapakse & Selvadurai elastic half-space where the shear modulus exhibits a power
(1989) studied rigid circular foundations with geometry law variation with depth, with zero shear modulus at the
similar to case 2, as well as rigid cylindrical foundations. surface. Stark & Booker (1997a, b) and Stark (2001) also
Rajapakse & Selvadurai (1985) presented solutions for the considered this power law variation of shear modulus with
torsional stiffness of non-uniform and hollow rigid piers in depth, and presented solutions for rectangular and circular
bonded contact with a layered elastic half-space. However, loaded areas. Selvadurai (1996) calculated the vertical stiff-
no analytical solutions are available for the horizontal or ness of a smooth rigid footing on a non-homogeneous region
moment load cases. This has motivated other researchers to with an exponential variation of shear modulus with depth,
employ approximate techniques, such as the finite-element and presented a comprehensive review of the literature
method. Bell (1991) carried out a series of three-dimensional related to available elastic solutions for the surface loading
finite-element analyses to examine the effect of embedment of a non-homogeneous elastic half-space. Solutions for
on rough rigid circular footings with three types of embed- the vertical loading of footings embedded in a non-
ment (cases 1, 2 and 3 shown in Fig. 1) subject to vertical homogeneous-elastic half-space have been obtained by Rowe
horizontal and moment loads. Bell (1991) expressed the & Booker (1981) and Rajapakse & Selvadurai (1991), but,
stiffness of a footing subject to combined loading in matrix as for the homogeneous half-space, solutions are not avail-
form as able for the horizontal or moment load cases. The torsional
8 9 8 w 9 studies of Randolph (1981) and Rajapakse & Selvadurai
> V > (1989) referred to above also considered non-homogeneous
>
> 2>
> 2 3>> >
>
>
> GR >
> KV 0 0 > R >
> > soil profiles. Randolph (1981) considered a linearly increas-
>
< >
= >
< >
=
H 6 7 u ing modulus with depth with zero shear modulus at the
¼ 4 0 K H K MH 5 (1)
> 2> > surface. Rajapakse & Selvadurai (1989) presented solutions
> GR >
>
> >
>
>
>
> R >
>
>
>
>
> > 0 K MH K M > > for the torsional response of rigid foundations embedded in
: M > ; :
ŁM
;
a non-homogeneous elastic half-space with a weathered
GR 3
crust.
where KV , K H , K M , and K MH are dimensionless elastic This paper examines the elastic response of embedded
stiffness coefficients, which are dependent only on Poisson’s rigid circular footings in bonded contact with the surround-
ratio (), embedment ratio (d/R) and the embedment condi- ing non-homogeneous elastic soil subjected to vertical, hor-
tions or footing geometry. Here KV , K H and K M correspond izontal, moment and torsional loading for the embedment
to vertical, horizontal and moment degrees of freedom conditions shown in Fig. 1. Following the study of Booker
respectively, K MH is the cross-coupling that exists between et al. (1985), the shear modulus is taken to have a power
the horizontal and moment degree of freedom, and G is the law variation with depth and, for non-homogeneous cases, to
shear modulus of the soil. have a value of zero at the surface. This is appropriate for
Ngo-Tran (1996) extended the work of Bell by determin- normally consolidated sands and normally consolidated clays
ing elastic stiffness coefficients for rigid conical footings, (Hardin & Drnevich, 1972) that have not developed a
making his results applicable to modelling spudcan founda- weathered crust, such as marine clays. The relationship
tions used to support offshore jack-up structures. The results between shear modulus and depth can be expressed as
of Bell (1991) and Ngo-Tran (1996) are widely used in  Æ
soil–structure interaction problems for offshore structures. z
G(z) ¼ G R (2)
Recommended practice for site specific assessment of mobile R
jack-up units (SNAME, 1997) includes a set of vertical,
horizontal and moment stiffness coefficients derived from where G R is the shear modulus at a depth equal to the
Bell’s study, which are recommended for use in the analyses radius of the footing (R), and Æ is the non-homogeneous
of jack-up structures with embedded foundations. parameter that varies between zero and one. This equation
Recent studies by Martin & Houlsby (2001) and Houlsby encompasses homogeneous (Æ ¼ 0) and Gibson (Æ ¼ 1) soil
& Cassidy (2002) use the results of Bell (1991) and profiles, as shown in Fig. 3.
Ngo-Tran (1996) in sophisticated plasticity-based load– Closed-form solutions presented by Booker et al. (1985)
displacement footing (spudcan) models where the response show that the stiffness of a circular region on the surface of
of the footing is expressed in terms of resultant forces (V, H, a compressible ( , 0·5) elastic half-space with a Gibson
M ). In these models a yield surface is defined in terms of V, soil profile is zero, whereas for all other cases (that is,
H and M load space. Inside the yield surface the footing Æ , 1 and/or  ¼ 0·5) a finite stiffness results. Doherty &
behaviour is described by linear elastic stiffness coefficients. Deeks (2003a) compared numerical solutions, computed
When the loading combination reaches the yield surface, using the scaled boundary finite-element method, with the
work-hardening plasticity theory is used to form an elasto- closed-form solution by Booker et al. (1985), and showed
plastic stiffness matrix to incrementally describe the non- excellent agreement, except for cases where the stiffness of
CIRCULAR FOOTINGS IN A NON-HOMOGENEOUS HALF-SPACE 705
GR G(z) method was first presented in Wolf & Song (1996). A
0
complete formulation is presented in Doherty & Deeks
(2002), including non-homogeneous elasticity in the form
1 described by equation (2). Doherty & Deeks (2003a) exam-
α1 ined the accuracy and efficiency of the scaled boundary
Gibson soil
finite-element solutions by comparing computed results with
analytical solutions for the vertical, horizontal, moment and
torsional loading of rigid circular footing on the surface of
an elastic half-space. They showed that, using a boundary
discretised with 61 nodes, the discrepancies between the
0α1 scaled boundary finite-element solutions and the analytical
α0
solutions are negligible for all load cases, and incompressi-
Homogenous bility ( ¼ 0·5) is modelled accurately, with no special
soil treatment, using a Poisson’s ratio of 0·49999.
z/R

Summary of method for an axisymmetric domain subjected


Fig. 3. Variation of shear modulus with depth
to general loading
The scaled boundary finite-element method uses a coordi-
the footing approaches zero. In this paper, as footings are nate system that scales a defining curve S relative to a
embedded at a finite depth below the soil surface, situations scaling centre. The normalised radial coordinate, , has a
resulting in zero footing stiffness are not encountered. value of 0 at the scaling centre and a value of 1 at the
The main aim of the paper is to determine the dependence defining curve S, which is discretised in a finite-element
of the dimensionless stiffness coefficients KV , K H , K M , K MH manner. The domain boundary is defined by revolving the
and KT (the dimensionless torsional stiffness coefficient discretised curve S and two radial ‘side-faces’, in the plane
defined in equation (9)) on Æ and on the embedment Ł ¼ 0, around the axis of symmetry, as shown in Fig. 4.
conditions. As discussed above, these parameters cannot be Each value of  defines a scaled version of the discretised
evaluated analytically in general, so a numerical approach is curve S, which is represented by a set of points (rs (s), z s (s)).
adopted. Because of the unbounded nature of the problem, s is a boundary coordinate measuring the distance along the
the scaled boundary finite-element method is chosen. curve S, from s ¼ s0 at the first side-face to s ¼ s1 at the
The scaled boundary finite-element method for axisym- second side-face. (For convenience, s0 may be taken to be
metric non-homogeneous situations under general loading zero, so s1 is the actual length of the curve S.)
conditions is briefly outline in the next section. (A complete To model an axisymmetric domain, the scaling centre and
derivation is presented by Doherty & Deeks (2003a)). This first side-face are restricted to the axis of symmetry, and
section also indicates how each embedment case is modelled therefore s ¼ s0 defines a vertical line on the axis of
using the scaled boundary finite-element method. The third symmetry. To permit Young’s modulus to be scaled with
section describes how the stiffness coefficients can be used depth, the scaling centre is positioned on the surface, and
to predict the load–displacement response of the embedded the second side-face at s ¼ s1 is horizontal.
footings, and presents stiffness coefficients that have not A bounded domain is defined by s0 < s < s1 , 0 < Ł < 2
been available until now. A fourth section provides a simple and 0 <  < 1 (Fig. 4(a)), whereas an infinitely deep do-
example demonstrating the usage of these coefficients. main, referred to as an unbounded domain because of its
infinite extent in the  direction, is defined by s0 < s < s1 ,
0 < Ł < 2 and 1 <  , 1 (Fig. 4(b)). The truncated cone
THE SCALED BOUNDARY FINITE-ELEMENT METHOD shown in Fig. 4 is an example of an axisymmetric domain.
Background However, more general shapes, where S is not piecewise
The scaled boundary finite-element method is a novel linear, can also be handled.
semi-analytical approach to continuum analysis developed by The scaled boundary and cylindrical coordinate systems
Wolf & Song (1996). The method combines many of the are related by the scaling equations
advantages of the standard finite element method and the r ¼ rs ð sÞ (3a)
boundary element method. In particular, unbounded pro-
blems may be handled with ease, as solutions are obtained z ¼ z s ð sÞ (3b)
analytically to infinity in the radial direction. The technique
involves discretisation of the boundary of the solution do- Ł¼Ł (3c)
main, but, unlike the boundary element method, does not
require the use of a fundamental solution. Because of the The scaled boundary finite-element method seeks an approx-
complexity of the original formulation, a simplified virtual imate solution for displacement in the form
8 9
work derivation for plane stress, plane strain and axisym- < u r (, s, Ł) = X 1 n   
metric elasto-statics has recently been presented (Deeks & u z (, s, Ł) ¼ F su ð n, ŁÞ ½ N ð sÞ u s ð, nÞ
Wolf, 2002), but this derivation is restricted to two- : ; n¼0
uŁ (, s, Ł)
dimensional situations and axisymmetric domains with axi-
   o
symmetric loading. Although the method was originally þ F au ð n, ŁÞ ½ N ð sÞ u a ð, nÞ (4)
derived for situations of homogeneous isotropic and aniso-
tropic elasticity, Doherty & Deeks (2003b) have extended it Here, the variation of displacement in the circumferential
to allow variation of shear modulus with depth in the form direction (Ł) is represented by a Fourier series, where
specified by equation (2), which is also restricted to the 2 3
cos nŁ 0 0
axisymmetric load case.
The use of a Fourier series to model general loading of an [F su (n, Ł)] ¼ 4 0 cos nŁ 0 5 (5)
axisymmetric region in the scaled boundary finite-element 0 0 sin nŁ
706 DOHERTY & DEEKS
θ Horizontal θ Horizontal
side-face Scaling centre
Scaling centre side-face
s axis
s  s1 s axis s  s1
[ξ  0]
r r
uz
uθ Discretised curve S
uz [ξ  1]
ur

Discretised curve S
[ξ  1] ur

Vertical Vertical ξ axis


side-face
z side-face [ξ → ]
s  s0 s  s0
(a)

z
(b)

Fig. 4. (a) Bounded and (b) unbounded axisymmetric scaled boundary finite-element domains, with curve S discretised
with five nodes

describes a symmetric variation of displacements about the Modelling details of embedded footings
r-axis at Ł ¼ 0, whereas In the scaled boundary finite-element method, displace-
ments can be prescribed at nodes on the discretised bound-
2 3
sin nŁ 0 0 ary in the same way that displacements are prescribed in the
[F au (n, Ł)] ¼ 4 0 sin nŁ 0 5 (6) standard finite-element method. Constant-displacement
0 0 cos nŁ boundary conditions can also be prescribed on side-faces, as
detailed by Doherty & Deeks (2003a). In each load case,
displacements are prescribed only at degrees of freedom on
describes an anti-symmetric variation of displacements about
the boundary/boundaries representing the rigid rough circular
the r-axis at Ł ¼ 0. footing; all other degrees of freedom are unrestrained.
[N(s)] contains shape functions corresponding to the dis-
Based on a study by Doherty & Deeks (2003a), each
cretisation of S, and {u s (,n)} and {u a (,n)} contain nodal solution in the next section is obtained using a boundary
displacement functions for the symmetric and anti-symmetric
curve discretised with 61 nodes. Fig. 5 shows the discretised
Fourier terms respectively. Nodal functions {u s (,n)} and boundary curves, S, the scaling centre, and each node
{u a (,n)} are found analytically in the solution process, and
displacement function for the four footing types considered.
represent the variation of the solution in the  direction. As The parts of the boundary used to represent the footing are
the displacements along the side-faces are represented by
shown with a thickened line. For all footing types and
nodal displacement functions, no discretisation along these
loading conditions considered here, the discretised boundary
domain boundaries is required.
curve is defined (with s0 ¼ 0 and s1 ¼ R + d) by
Based on equation (4), Doherty & Deeks (2003a) obtained 
expressions for stress and strain that are used to derive an s 0<s< R
rs (s) ¼ (8a)
expression for internal virtual work. This is equated to R R< s< Rþd
expressions for external work done by forces acting at nodes 
on the discretised boundary and the radial side-faces. The d 0<s<R
z s (s) ¼ (8b)
resulting set of equations reduces to a series of decoupled dþ Rs R< s< Rþd
quadratic eigenvalue problems, one for each symmetric and The bounded domain is a cylinder, whereas the unbounded
anti-symmetric Fourier term, which are solved to form a domain is a half-space with a cylindrical cavity.
stiffness matrix for the domain relative to degrees of free- To model the vertical load case, only the symmetric
dom at nodes on the discretised curve on the boundary of component of n ¼ 0 in the Fourier series is used, resulting in
the domain. This is done separately for bounded and un- a stiffness matrix formed relative to radial and vertical
bounded domains, and the stiffness matrix for each is degrees of freedom only. A uniform displacement of the
assembled together as in the standard finite-element method vertical degree of freedom of each node on the boundary
to form a global stiffness matrix for the entire problem. A representing the footing is prescribed, while the radial degree
linear equation of the form of freedom is restrained to comply with the rough footing
condition. For the horizontal and moment load cases only
fPn g ¼ [K n ]fu n g (7) the symmetric component of n ¼ 1 in the Fourier series is
used, which results in each node having all three degrees of
results for the symmetric and anti-symmetric component of freedom. For the horizontal load case the vertical degree of
each Fourier term n, where {Pn } is the vector of nodal force freedom of each node on the boundary representing the
amplitudes, [K n ] is the stiffness matrix, and {u n } is the footing is restrained, while uniform radial and tangential
vector of nodal displacement amplitudes. Equation (7) is displacements are imposed, with u r ¼ uŁ . For the moment
solved independently for each component present in the load case the radial and tangential degrees of freedom are
Fourier series, in the standard finite-element manner. Pro- restrained, while the vertical degree of freedom of each node
blems involving the translation or rotation of a rigid circular on the boundary representing the footing is displaced with a
footing, such as those considered here, contain only one magnitude proportional to the distance of the node from the
Fourier term in the solution. For full details the reader axis of symmetry, with zero displacement on the axis of
should refer to Doherty & Deeks (2003a). symmetry. The torsional load case is modelled using only
CIRCULAR FOOTINGS IN A NON-HOMOGENEOUS HALF-SPACE 707
R
r r r r

z z z z

(a) (b) (c) (d)

Fig. 5. Scaled boundary finite-element mesh for each footing type: (a) case 1; (b) case 2; (c) case 3; (d) case 4

the anti-symmetric component of n ¼ 0, resulting in only a cases shown in Fig. 1. These stiffness coefficients can be
tangential degree of freedom at each node. A displacement used to predict the load–displacement response through the
of the tangential degree of freedom of each node on the following equation:
boundary representing the footing is imposed with a magni-
tude proportional to the distance of the node from the axis of
8 9
symmetry, with zero displacement on the axis of symmetry. > V > 8 w 9
>
> >
>
As illustrated in Fig. 5(a), case 1 is modelled is using an >
> G R R2 >
> > >
unbounded domain only, and the footing is represented by >
> >
> 2 3>> R >
> >
>
>
> >
> KV 0 0 0 > > >
>
the boundary with 0 < s < R. Case 2, as illustrated in Fig. >
> H >
> > u >
> >
< G R2 >
> = 6 7>< >
=
5(b), is modelled in the same way as case 1, except that a R 6 0 KH K MH 0 7
¼6
6 0
7 R
7 (9)
bounded domain is included. For case 3, the entire discre- >
> M >
> 4 K MH KM 0 5>> >
>
>
> >
> >
> >
>
tised curve is use to represent the footing, and the bounded >
> G R R3 >
> >
> Ł M >
>
domain is omitted. Case 4 is modelled by combining >
> >
> 0 0 0 KT >> >
>
>
> > > >
bounded and unbounded domains, as shown in Fig. 5(d). >
> T > >
>
:
ŁT
;
: ;
The caisson’s skirt is modelled using prescribed nodal G R R3
displacements on the vertical section of the discretised
boundary—that is, the section of the boundary defined by
R < s < R + d, whereas the top of the caisson is modelled This equation is similar to equation (1) but with G replaced
by prescribing the displacement of the horizontal side-face, by G R and the additional stiffness coefficient KT describing
at s ¼ s1 , of the bounded domain. the torsional response, which is independent of Poisson’s
ratio, included. In order to use equation (9) to predict the
load–displacement response of an embedded footing, para-
STIFFNESS COEFFICIENTS FOR EMBEDDED meters G R , Æ (in equation (2)) and Poisson’s ratio must be
FOOTINGS evaluated using conventional site investigation techniques.
Using the stiffness coefficients Given Æ and Poisson’s ratio, stiffness coefficients can be
In the following subsection, stiffness coefficients KV , K H , read from Figs 6–22 for the relevant footing geometry and
K M , KT and K MH are presented for the four embedment embedment ratio.

40 80
35 d/R  0·5 ν  0·2 70 d/R  0·5 ν  0·5
d/R  1 d/R  1
30 60
d/R  2 d/R  2
25 50
d/R  4 d/R  4
KV

20
KV

40
15 30
10 20
5 10
0 0
0 0·2 0·4 0·6 0·8 1 0 0·2 0·4 0·6 0·8 1
α α
(a) (b)

Fig. 6. Case 1, KV : (a) í 0·2; (b) í 0·5


708 DOHERTY & DEEKS
50 80
d/R  0·5 ν  0.2 70 d/R  0·5 ν  0.5
40 d/R  1 d/R  1
60
d/R  2 d/R  2
30 d/R  4 50 d/R  4
KV

KV
40
20 30

20
10
10
0 0
0 0·2 0·4 0·6 0·8 1 0 0·2 0·4 0·6 0·8 1
α α
(a) (b)

Fig. 7. Case 2, KV : (a) í 0·2; (b) í 0·5

70 100

60 d/R  0·5 ν  0·2 d/R  0·5 ν  0·5


d/R  1 80 d/R  1
50
d/R  2 d/R  2
40 d/R  4 60 d/R  4
KV

KV
30
40
20
20
10

0 0
0 0·2 0·4 0·6 0·8 1 0 0·2 0·4 0·6 0·8 1
α α
(a) (b)

Fig. 8. Case 3, KV : (a) í 0·2; (b) í 0·5

60 100
ν  0·2 ν  0·5
d/R  0·5 d/R  0·5
50
d/R  1 80 d/R  1
40 d/R  2 d/R  2
d/R  4 60 d/R  4
KV

KV

30
40
20

20
10

0 0
0 0·2 0·4 0·6 0·8 1 0 0·2 0·4 0·6 0·8 1
α α
(a) (b)

Fig. 9. Case 4, KV : (a) í 0·2; (b) í 0·5

35 60

d/R  0·5 ν  0.2 d/R  0·5 ν  0.5


30 50
d/R  1 d/R  1
25
d/R  2 40 d/R  2
20 d/R  4 d/R  4
KH

KH

30
15
20
10

5 10

0 0
0 0·2 0·4 0·6 0·8 1 0 0·2 0·4 0·6 0·8 1
α α
(a) (b)

Fig. 10. Case 1, K H : (a) í 0·2; (b) í 0·5


CIRCULAR FOOTINGS IN A NON-HOMOGENEOUS HALF-SPACE 709
45 70
40 d/R  0·5 ν  0·2 d/R  0·5 ν  0·5
60
35 d/R  1 d/R  1
50
30 d/R  2 d/R  2
25 d/R  4 40 d/R  4
KH

KH
20 30
15
20
10
5 10

0 0
0 0·2 0·4 0·6 0·8 1 0 0·2 0·4 0·6 0·8 1
α α
(a) (b)

Fig. 11: Case 2, K H : (a) í 0·2; (b) í 0·5

70 90
ν  0·2 80 ν  0·5
60 d/R  0·5 d/R  0·5
d/R  1 70 d/R  1
50
d/R  2 60 d/R  2
40 d/R  4 d/R  4
50
KH

KH
30 40
30
20
20
10
10
0 0
0 0·2 0·4 0·6 0·8 1 0 0·2 0·4 0·6 0·8 1
α α
(a) (b)

Fig. 12. Case 3, K H : (a) í 0·2; (b) í 0·5

90
70
80 d/R  0·5 ν  0·5
60 d/R  0·5 ν  0·2
70 d/R  1
d/R  1
50 60 d/R  2
d/R  2
d/R  4
40 d/R  4 50
KH
KH

40
30
30
20
20
10 10
0 0
0 0·2 0·4 0·6 0·8 1 0 0·2 0·4 0·6 0·8 1
α α
(a) (b)

Fig. 13. Case 4, K H : (a) í 0·2; (b) í 0·5

25 40
ν  0·2 ν  0·5
d/R  0·5 35 d/R  0·5
20 d/R  1 d/R  1
30
d/R  2 d/R  2
25
15 d/R  4 d/R  4
KM

KM

20
10 15

10
5
5
0 0
0 0·2 0·4 0·6 0·8 1 0 0·2 0·4 0·6 0·8 1
α α
(a) (b)

Fig. 14. Case 1, K M : (a) í 0·2; (b) í 0·5


710 DOHERTY & DEEKS
35 50

d/R  0·5 ν  0·2 d/R  0·5 ν  0·5


30
d/R  1 40 d/R  1
25
d/R  2 d/R  2
20 d/R  4 30 d/R  4
KM

KM
15 20
10
10
5

0 0
0 0·2 0·4 0·6 0·8 1 0 0·2 0·4 0·6 0·8 1
α α
(a) (b)

Fig. 15. Case 2, K M (a) í 0·2 (b) í 0·5

250 300

d/R  0·5 ν  0·2 d/R  0·5 ν  0·5


250
200 d/R  1 d/R  1
d/R  2 200 d/R  2
150 d/R  4 d/R  4
(d/R)

(d/R)
KM

KM
150
100
100

50
50

0 0
0 0·2 0·4 0·6 0·8 1 0 0·2 0·4 0·6 0·8 1
α α
(a) (b)

Fig. 16. Case 3, K M : (a) í 0·2; (b) í 0·5

200 300
ν  0·2 d/R  0·5 ν  0·5
d/R  0·5
250 d/R  1
150 d/R  1
d/R  2
d/R  2 200
d/R  4
d/R  4
(d/R)
(d/R)

KM
KM

100 150

100
50
50

0 0
0 0·2 0·4 0·6 0·8 1 0 0·2 0·4 0·6 0·8 1
α α
(a) (b)

Fig. 17. Case 4, K M : (a) í 0·2; (b) í 0·5

0 1
0·5
0
1
1
1·5
2 2
KMH

KMH

2·5 d/R  0·5 3


d/R  0·5
3 d/R  1
4 d/R  1
3·5 d/R  2
d/R  2
d/R  4 ν  0·2 5 ν  0·5
4 d/R  4
4·5 6
0 0·2 0·4 0·6 0·8 1 0 0·2 0·4 0·6 0·8 1
α α
(a) (b)

Fig. 18. Case 1, K MH : (a) í 0·2; (b) í 0·5


CIRCULAR FOOTINGS IN A NON-HOMOGENEOUS HALF-SPACE 711
0·4 0·8

0·3 d/R  0·5 0·6 ν  0·5


d/R  1
0·2 ν  0·2 d/R  2
0·4

0·1 d/R  4 0·2


Linear (d/R  4) 0

KMH
KMH

0
0·2 d/R  0·5
0·1 d/R  1
0·4
d/R  2
0·2
0·6 d/R  4
0·3 0·8 Linear (d/R  4)
0·4 1
0 0·2 0·4 0·6 0·8 1 0 0·2 0·4 0·6 0·8 1
α α
(a) (b)

Fig. 19. Case 2, K MH : (a) í 0·2; (b) í 0·5

0 0

10
10
20
20
30
(d/R)
KMH

(d/R)
KMH

30 d/R  0·5 40 d/R  0·5


d/R  1 d/R  1
50
d/R  2 d/R  2
40 ν  0·2 ν  0·5
d/R  4 60 d/R  4

50 70
0 0·2 0·4 0·6 0·8 1 0 0·2 0·4 0·6 0·8 1
α α
(a) (b)

Fig. 20. Case 3, K MH : (a) í 0·2; (b) í 0·5

0 0

10
10
20
20 30
(d/R)

(d/R)
KMH

KMH

40
30 d/R  0·5 d/R  0·5
d/R  1 50 d/R  1
40 d/R  2 d/R  2
ν  0·2 60 ν  0·5
d/R  4 d/R  4
50 70
0 0·2 0·4 0·6 0·8 1 0 0·2 0·4 0·6 0·8 1
α α
(a) (b)

Fig. 21. Case 4, K MH : (a) í 0·2; (b) í 0·5

Computed coefficients been divided by their respective embedment ratio in Figs 16,
Results are presented for Poisson’s ratios of 0·2, to Fig 17, Fig 20 and 21 to improve readability of the figures.
represent a wide range of sands (Lade, 1977), and Poisson’s
ratio of 0·5 to represent undrained clays. Embedment ratios
(d/R) of 0·5, 1·0, 2·0 and 4·0 are considered, which ade- Discussion
quately cover most foundation applications. For each case, It can be seen from the Figures in the preceding subsec-
non-homogeneous soil profiles with Æ ranging between 0 tion that the stiffness coefficients vary smoothly with respect
and 1 are considered, encompassing realistic variations of to the non-homogeneity parameter Æ for all load cases,
shear modulus with depth for both normally consolidated except for the cross-coupling term (K MH ) for case 2, shown
sands and clays (Hardin & Drnevich, 1972). Based on the in Fig. 19. For R/d ¼ 4, the results are scattered and a linear
study of Doherty & Deeks (2003a), all results presented here line of best fit is included. In this case, the cross-coupling
use a boundary discretised with 61 nodes, giving an accu- terms are small relative to the corresponding K H and K M ,
racy of within 1%. K M , and K MH for cases 3 and 4, have terms, so these values are of little practical significance.
712 DOHERTY & DEEKS
40 50
35 d/R  0·5 d/R  0·5
d/R  1 40 d/R  1
30
d/R  2 d/R  2
25
d/R  4 30 d/R  4

KT
KT
20
15 20

10
10
5
0 0
0 0·2 0·4 0·6 0·8 1 0 0·2 0·4 0·6 0·8 1
α α
(a) (b)

140 140
d/R  0·5
120 120 d/R  0·5
d/R  1 d/R  1
100 d/R  2 100 d/R  2
d/R  4 d/R  4
80 80
KT

KT
60 60

40 40

20 20

0 0
0 0·2 0·4 0·6 0·8 1 0 0·2 0·4 0·6 0·8 1
α α
(c) (d)

Fig. 22. KT : (a) case 1; (b) case 2; (c) case 3; (d) case 4

EXAMPLE APPLICATION OF STIFFNESS


COEFFICIENTS
This section provides an example demonstrating the appli- 8 MN
cation of the stiffness coefficients presented in the previous
section to obtain a first estimate of the footing displacements
for a given set of design loads.
As a result of a growing demand to produce clean renew-
able sources of energy, offshore wind turbines are becoming
increasingly common worldwide. Byrne & Houlsby (2002)
proposed the use of novel foundation systems to support
these offshore wind turbines, which involves the use of one
or more caisson (case 4) foundations. 50 m
The configuration of the structure used in this example is 4 MN
shown in Fig. 23: it consists of a single caisson supporting a
wind turbine. The loads acting on the structure are assumed
to be V ¼ 8 MN, H ¼ 4 MN at a height 20 m above 20 m
the caisson resulting in an over-turning moment of
M ¼ 80 MN m at reference point O and an eccentricity about O 3 MPa G
the vertical axis of 3 m, resulting in a torsional moment of
T ¼ 12 MN m. The soil properties are assumed to be d  12 m
Æ ¼ 0·8, G R ¼ 3 MPa and Poisson’s ratio  ¼ 0·2, and the 12 m
caisson is assumed to be 12 m in radius (R) and 12 m in
R  12 m
depth (d ).
For d/R ¼ 1,  ¼ 0·2 and Æ ¼ 0·8 for case 4, KV ¼ 11·5 z
from Fig. 9(a), K H ¼ 10·3 from Fig. 13(a), K M ¼ 16·9 from
Fig. 17(a), K MH ¼ 8·4 from Fig. 21(a) and KT ¼ 14·5 from
Fig. 22(d). Fig. 23. Caisson supporting offshore wind turbine
From equation (9), the following expressions can be
obtained for the displacement components:

V K MH H KH M
w¼ (10) ŁM ¼ þ
KV GR R K M K H  K 2MH G R R2 K M K H  K 2MH G R R3
(12)
KM H K MH M T
u¼  (11) ŁT ¼ (13)
K M K H  K 2MH G R R K M K H  K 2MH G R R2 K T G R R3
CIRCULAR FOOTINGS IN A NON-HOMOGENEOUS HALF-SPACE 713
Substituting in the stiffness coefficients, design loads, G R fu a (, n)g vector of ant-symmetric nodal displacement functions
and R specified above, the resultant displacements at the fu s (, n)g vector of symmetric nodal displacement functions
intersection of the centreline of the footing and ground
surface are w ¼ 0·019 m, u ¼ 0·033 m, Ł M ¼ 0·002 radians
and ŁT ¼ 0·00016 radians. This results in total horizontal
and vertical displacements at the centre of the wind turbine REFERENCES
hub, 50 m above the ground surface, of 0·133 m and Bell, R. W. (1991). The analysis of offshore foundations subjected
to combined loading. MSc thesis, University of Oxford.
0·019 m respectively. Booker, J. R., Balaam, N. P. & Davis, E. H. (1985). The behaviour
of an elastic non-homogeneous half-space. Int. J Numer. Anal.
Methods Geomech. 9, 353–367.
CONCLUSION Byrne, B. W. & Houlsby, G. T. (2002). Investigating novel founda-
The scaled boundary finite-element method is used to tions for offshore windpower generation. Proc. 21st Int. Conf.
obtain semi-analytical stiffness coefficients for rigid circular on Offshore Mechanics and Arctic Engineering, Oslo.
footings subjected to vertical, horizontal, moment and tor- Deeks, A. J. & Wolf, J. P. (2002). A virtual work formulation of the
sional loads embedded in a non-homogeneous half-space. scaled boundary finite-element method for elastostatics. Comput.
Mech. 28, No. 6, 489–504.
The non-homogeneous profiles considered are appropriate
Doherty, J. P. & Deeks, A. J. (2003a). Scaled boundary finite-
for normally consolidated sands and clays, and the founda- element analysis of a non-homogeneous axisymmetric domain
tion geometries and embedment ratios examined cover many subjected to general loading. Int. J. Numer. Anal. Meth. Geo-
of the footing types used in industry. mech. 27, 813–835.
The soil non-homogeneity parameter Æ has a significant Doherty, J. P. & Deeks, A. J. (2003b). Scaled boundary finite-
effect on all stiffness coefficients. The effect of Æ is greatest element analysis of a non-homogeneous elastic half-space. Int. J
for the vertical and moment load cases as displacement Numer. Methods Engng 57, 955–973.
fields extend deeper into the half-space encountering materi- Gibson, R. E. (1967). Some results concerning displacements and
al of increasing stiffness. The interaction between Poisson’s stresses in a non-homogeneous elastic half-space. Géotechnique
17, 58–67.
ratio and Æ is greatest for loading cases that have a vertical
Hardin, B. O. & Drnevich, V. P. (1972). Shear modulus and
component. damping in soils: design equations and curves. J. Soil Mech.
Found. Div., ASCE 98, No. SM7, 667–692.
Houlsby, G. T. & Cassidy, M. J. (2002). A plasticity model for the
ACKNOWLEDGEMENTS behaviour of footings on sand under combined loading. Géotech-
The first author wishes to acknowledge the support of an nique 52, No. 2, 117–129.
Australian Postgraduate Award funded by the Common- Hunter, S. C. & Gamble, D. (1974). The theory of a rigid circular
wealth Government of Australia and a Jean Rogerson Post- disc ground anchor in an elastic soil either with or without
graduate supplementary scholarship from the University of adhesion. J. Mech. Phys. Solids 22, 371–399.
Lade, P. V. (1977). Elasto-plastic stress–strain theory for cohesion-
Western Australia. less soil with curved yield surface. Int. J Solids Struct. 13,
1019–1035.
Martin, C. M. (1994). Physical and numerical modelling of offshore
NOTATION foundations under combined loading. PhD thesis, University of
d embedment depth of footing Oxford.
G soil shear modulus Martin, C. M. & Houlsby, G. T. (2001). Combined loading of
H horizontal load spudcan foundations on clay: numerical. Géotechnique 51, No.
M moment load 8, 687–699.
n Fourier term Ngo-Tran, C. L. (1996). The analysis of offshore foundations sub-
O load and displacement reference point on footing jected to combined loading. PhD thesis, University of Oxford.
R radius of footing Poulos, H. G. (1975). Torsional response of piles. J. Geotech.
r radial coordinate Engng, ASCE 101, No. GT10, 1014–1035.
S defining curve Poulos, H. G. & Davis, E. H. (1974). Elastic solutions for soil and
s boundary coordinate rock mechanics. New York: John Wiley & Sons.
T torsional load Rajapakse, R. K. N. D. & Selvadurai, A. P. S. (1985). Torsional
u horizontal displacement of footing stiffness of non-uniform and hollow rigid piers embedded in
V vertical load isotropic elastic media. Int. J. Numer. Anal. Methods Geomech.
w vertical displacement of footing 9, 525–539.
z vertical coordinate Rajapakse, R. K. N. D. & Selvadurai, A. P. S. (1989). Torsion of
GR soil shear modulus at d ¼ R foundations embedded in a non-homogeneous soil with a weath-
KV vertical stiffness coefficient ered crust. Géotechnique 39, No. 3, 485–496.
KH horizontal stiffness coefficient Rajapakse, R. K. N. D. & Selvadurai, A. P. S. (1991). Response of
KM moment stiffness coefficient circular footings and anchor plates in non-homogeneous elastic
K MH cross-coupling stiffness coefficient soils. Int. J. Numer. Anal. Methods Geomech. 15, 457–470.
ur radial nodal displacement component Randolph, M. F. (1981). Piles subjected to torsion. J. Geotech.
uz vertical nodal displacement component Engng, ASCE 107, 1095–1111.
uŁ tangential nodal displacement component Rowe, R. K. & Booker, J. R. (1981). The elastic displacement of
Æ non-homogeneous parameter single and multiple underream anchors in a Gibson soil. Géo-
 Poisson’s ratio of soil technique 31, No. 1, 125–141.
Ł circumferential coordinate Schofield, A. & Wroth, C. P. (1968). Critical state soil mechanics.
 normalised radial coordinate Maidenhead: McGraw-Hill.
ŁM moment rotation of footing Selvadurai, A. P. S. (1979). Elastic analysis of soil-foundation
ŁT torsional rotation of footing interaction, Developments in Geotechnical Engineering, No. 17.
[K n ] stiffness matrix for the nth Fourier term Amsterdam: Elsevier.
[N (s)] shape function matrix Selvadurai, A. P. S. (1993). The axial loading of a rigid circular
fPn g vector of nodal force amplitudes for the nth Fourier anchor plate embedded in an elastic half-space. Int. J. Numer.
term Anal. Methods Geomech. 17, 343–353.
fu n g vector of nodal displacement amplitudes for the nth Selvadurai, A. P. S. (1996). The settlement of a rigid circular
Fourier term foundation resting on a half-space exhibiting a near surface
714 DOHERTY & DEEKS
elastic non-homogeneity. Int. J. Numer. Anal. Methods Geomech. non-homogeneous elastic half-space subjected to uniform sur-
20, 351–364. face tractions. Part I: Loading on arbitrarily shaped areas. Int. J.
SNAME (1997). Recommend practice for site specific assessment of Numer. Anal. Methods Geomech. 21, 361–378.
mobile jack-up units, Technical and Research Bulletin 5–5A. Stark, R. F. & Booker, J. R. (1997b). Surface displacement of a
Society of Naval Architects and Marine Engineers. non-homogeneous elastic half-space subjected to uniform sur-
Stark, R. F. (2001). Integration of singularities in FE/BE analyses face tractions. Part II: Loading on arbitrarily shaped areas. Int.
of soil-foundation interaction with non-homogeneous elastic J. Numer. Anal. Methods Geomech. 21, 379–395.
soils. Meccanica 36, 329–350. Wolf, J. P. & Song, Ch. (1996). Finite-element modelling of
Stark, R. F. & Booker, J. R. (1997a). Surface displacement of a unbounded media. Chichester: John Wiley & Sons.

You might also like