Академический Документы
Профессиональный Документы
Культура Документы
1 1,
INTERNATIONALJOURNAL FOR NUMERICAL A N D ANALYTICAL METHODS I N GEOMECHANICS, 1-1 5 (1987)
Y . K.CHOW
Department of Civil Engineering, National University of Singapore. Singapore
SUMMARY
A numerical procedure is described for the analysis of vertical deformation of smooth, rigid foundations of
arbitrary shape on homogeneous and layered soil media. The contact area at the interface of the foundation
and soil medium is approximated by square subdivisions.The response of the system is then obtained from the
superpositionof the influence of the individual subdivisions.The flexibility influence coefficients are based on
equivalent smooth, rigid circular areas with the same contact area as the square subdivisions.For foundations
on a homogeneous, isotropic elastic half-space, the flexibility coelfcients are given analytically by the
integrated forms of the Boussinesq’s solution. For a layered soil medium, the flexibility coefficients are
determined from an axisymmetric finite element analysis which is essentially two dimensional. Thus, there is
no necessity for a full three-dimensionalfinite element analysis.Comparison with solutions obtained using the
integral transform technique for smooth, rigid rectangular foundations on a homogeneous, isotropic elastic
half-space shows good agreement. Parametric solutions are presented for the response of rectangular
foundations on some ‘typical’ soil profiles. The use of a simplified method to estimate the settlement of
rectangular foundations on a layered soil medium by superposing solutions for homogeneous,elastic strata is
discussed.
INTRODUCTION
The finite element method is a very powerful tool available to an engineer for the analysis of
geotechnical engineering problems. However, for the analysis of vertical deformation of rigid
foundations of arbitrary shape, the three-dimensional nature of the problem requires excessive
computer resources, as well as data preparation time, making the method uneconomical except for
important projects. More economical methods have been developed which enable the response of
arbitrary shaped foundations to be determined at a reasonable cost. Cheung and Zienkiewicz’
proposed a method for the analysis of surface foundations of arbitrary flexibility. Butterfield and
Banerjee’ extended the approach to include both surface and embedded rigid foundations.
However, in both cases, the soil model is restricted to a homogeneous, isotropic elastic half-space.
Brown and Gibson3 considered a soil with stiffness increasing linearly with depth, but only results
for flexible foundations were reported. Wardle and Fraser4 and Fraser and Wardle’ generalized
the foregoing methods to enable the response of rectangular foundations on layered soil media to
be determined. The method involves the use of integral transforms6 which are difficult to evaluate.
Moreover, computer programs based on this approach are not widely available. However, the
approach is a significant advancement, enabling foundations resting on layered soil media to be
analysed in a rational manner.
This paper describes an alternative numerical procedure which enables the vertical deformation
of smooth, rigid foundations of arbitrary shape resting on layered soil media to be determined at a
reasonable cost. The stresses experienced by the soil are assumed to be sufficiently small such that
the media behave essentially in a linear elastic manner. The soil properties are assumed to be
isotropic. Parametric solutions are presented for smooth, rigid rectangular foundations on a
homogeneous half-space, a homogeneous elastic stratum, a layered soil overlying a rigid base and a
soil layer with stiffness increasing linearly with depth. Comparisons are made with published
solutions where these are available.
NUMERICAL PROCEDURE
The irregular geometry of the contact area at the interface of the arbitrary shaped rigid foundation
and the soil medium is approximated by square subdivisions (see Figure 1). A ‘relaxed’ boundary
condition is assumed at the interface of the foundation and the soil medium such that the tangential
shear stresses are zero. Physically, this implies a perfectly smooth base for the foundation. The
influence of the roughness of the foundation base on its settlement is studied in some detail by
Carrier and Christian.’ For a rigid circular foundation resting on a homogeneous half-space, the
maximum difference between the rough foundation and smooth foundation settlements occurs at
a soil Poisson’s ratio v = 0 and is approximately 10 per cent. For v 2 0.3, the rough foundation and
smooth foundation solutions are nearly equal. For a rigid circular foundation resting on a non-
homogeneous half-space where the soil stiffness is proportional to depth, the solutions for a rough
foundation and smooth foundation are the same when v = 0.5, and the maximum difference occurs
at v = 0 and is less than 3 per cent.
If the vertical stresses (or loads) over each subdivisions are known, then the deformation of a
subdivision i (see Figure 1) may be obtained by superposition from the following equation:
j=n
wi= 1 F..P.
j= 1
1J J
in which F i j is the flexibility coefficient denoting the deformation at subdivision i due to a unit load
acting on subdivisionj (see Figure 1); P, is the value of load acting on subdivision j (see Figure 1);
and n is the number of square subdivisions. Equation (1) may be written for each of the subdivisions
making up the rigid foundation resulting in the following matrix equation:
Iw> = CF1 {PI (2)
Thus, the flexibility coefficients for a square subdivision of side a may be determined from
equations (3) and (6)by substituting ro = a / , / n .The flexibility coefficient of the square subdivision
due to load acting on itself (i.e. at i ) is given by
and the influence on subdivision i due to loads on the remaining subdivisions may be obtained
from
1. Discretize the contact area of foundation into square subdivisions (see Figure 1).
2. Evaluate the flexibility coefficients in matrix [ F ] in equation (2) using equation (7) for the
diagonal elements and equation (8) for the ofl-diagonal elements.
3. Solve equation (2) with { w ) = ( 1 ) (i.e. uniform deformation of the foundation) to give the
contact forces { P } .
4. Sum the forces in { P } to give the total force acting on the foundation to cause a unit
deformation.
Since the soil medium is assumed to behave in a linear elastic manner, the load-deformation
behaviour of the foundation has now been established.
Homogeneous soil conditions are rarely encountered in practice. This soil model is useful for
preliminary analysis due to its relative simplicity. For a more realistic estimate of the deformation
of the foundation, the inhomogeneity of the soil may have to be considered. Analytical solution for
the load-deformation behaviour of a smooth, rigid circular foundation resting on the surface of a
generally layered soil medium is not available. Thus one needs to resort to some form of numerical
methods of analysis. A numerical procedure for the determination of the flexibility coefficients for a
layered soil medium is described in the following subsection.
1 2 3
Figure 2. Local co-ordinates in isoparametric element
relatively simple matter. Thus results from a two-dimensional analysis may be used to determine
the response of a three-dimensional problem. The flexibility coefficients are determined directly
from a smooth, rigid circular footing analysis. Thus, there is no further need to perform numerical
integrations to obtain the influence of the distributed loads. In the axisymmetric finite element
analysis, the smooth, rigid footing is modelled through a prescribed displacement procedure in
which a uniform settlement is prescribed to the element nodes supposedly in contact with the
footing. The loads required to cause this uniform settlement is recovered though the stresses at the
Gaussian integration points via a numerical integration procedure. Details of the numerical
procedure for a prescribed displacement analysis may be seen in Reference 11.
Although the deformations in the finite element analysis are only explicitly given at the nodal
points, the deformation at any location on the ground surface can be determined using the element
shape functions of the isoparametric elements. For the quadratic isoparametric elements used
exclusively in this paper, the vertical deformation at the ground surface may be evaluated from (see
Figure 2):
+
w = N ~ w ~N ~ w , + N 3 ~ 3 (9)
in which wl,w 2 , w 3 are vertical deformations at nodes 1, 2, and 3, respectively;
N
r
--(t- 1);
'-2
N2=1- r2;
N3 =:(l
L
+ t);
and t is the local co-ordinate in the isoparametric element. The local co-ordinate may be
determined from
(1 1)
in which T is the radial distance from the centre of rigid circular footing; rl,rz,r3 are radial co-
ordinates of nodes 1,2 and 3, respectively (see Figure 2).
The flexibility coefficients Fij which correspond to the deformation of subdivision i due to a unit
load applied at subdivision j may be obtained from equation (9) for a unit load acting on the
smooth, rigid circular foundation. Thus, the response of the arbitrary shaped foundation on a
layered soil medium may be obtained following the procedure summarized in the foregoing
subsection with the flexibility coefficients calculated using equation (9) at stage 2.
In this procedure, only a single axisymmetric finite element analysis is required should
foundation bases with different geometries need to be investigated. The influence of other
subdivisions may be readily obtained by a physical shift in the reference axis. The computing costs
are probably higher when compared to the integral transform technique in evaluating the
flexibility coefficients, but the wide availability of such finite element computer programs should
6 Y.K.CHOW
encourage the use of this approach. Isotropic and anisotropic soil properties may be dealt with, but
the method is restricted to horizontally layered soil of 'inlinite' lateral extent. In this paper, only
isotropic soil properties are considered.
It may be noted here that the axisymmetric finite element procedure described herein may be
used to obtained the approximate flexibility coefficients for the homogeneous, elastic half-space
described in the foregoing subsection,but a suficiently large model of the domain must be used and
thereby increasing the cost of analysis. Thus, for a thick homogeneous soil layer, it would be more
cost effective to use the analytical expressions in equations (7) and (8) to obtain the approximate
flexibility coefficients. However, in many practical situations, the stiffness of the soil generally
varies with depth and this has to be reflected in the finite element model through the appropriate
modelling of the soil layers.
In the finite element analyses, the stiffness matrices of the quadratic isoparametric elements are
numerically obtained with 'exact' integration (i.e. 3 x 3 Gaussian integration points). The
programs are run on an IBM 3081 computer with double precision arithmetic which gives an
approximate accuracy of 17 significant digits. Poisson's ratio of the soil of up to v = 0.4999 has been
used without numerical problems. The linearly increasing stiffness of the soil within the element is
easily handled through the numerical integration process.
NUMERICAL RESULTS
L '
Figure 3. Discretization of contact area of rectangular foundation with LIB = 2 into 16 x 8 subdivisions
LAYERED SOIL MEDIA 7
0 2 L 6 8 10 12
LIB
Figure 4. Displacement influence values for smooth, rigid rectangular foundations on a homogeneous, isotropic elastic
half-space
t+-i
Rigid
finite element analyses. The response of rectangular foundations is computed as in the previous
example. The accuracy of the finite element method in determining the vertical deformation of a
smooth, rigid circular foundation on a homogeneous soil layer is assessed by comparing with
solutions by Brown" in Figure6. In general, the agreement is reasonable.
Table I1 compares the displacement influence values I, (see equation (12)) for square foundations
resting on layers of different thickness and their equivalent circular foundations (based on equal
contact areas). The maximum difference in the solutions is only about 2 per cent. This further
supports the approximation of the square subdivisions by equivalent circular areas used in the
numerical procedure. The response of rectangular foundations on layers of various thickness and
Poisson's ratio of the soil v = 0, 0.3 and 0.4999 are also computed and the results are shown in
LAYERED SOIL MEDIA 9
b
B
- Finite elements
0 Brownlo
I I
0 2 L 6 0 10 12
H /re
Figure 6. Comparison of solutions for the response of smooth, rigid circular foundations on a homogeneous layer
1 2 4 6
I 2 4 6
Figure 7. For this problem, Fraser and Wardle' presented a limited range of results and they are
also included in Figure 7 for comparison. The agreement between the results is reasonably good. A
useful design chart given by Janbu et aL'* and also published in a book by Simons and Menzies13
provides estimates of the average immediate settlement of uniformly loaded, flexible rectangular
foundations on an incompressible elastic soil stratum. Comparison of solutions by Janbu et af.12
Fraser and Wardle' and the present approach indicates that the average settlement of uniformly
loaded, flexible rectangular foundations generally overestimates the deformation of equivalent
smooth, rigid rectangular foundations. For example, Table 111 shows the displacement influence
10 Y.K.CHOW
1.6
8
6
c
1.2
2
I,
0.8 1
0.4
B Fmrw and Wardlc’
0
/ I I I I
9 =o.o
I
1.6 c-
LB
‘I
e-
1.2
1,
0 .a
0.4
0
1.6
LIB
1.2
1,
0.8
04
C
1 $ = 0.4999
0 1 2 3 4 5 6
H/B
Figure 7. Displacement influence values for smooth, rigid rectangular foundations on a homogeneous elastic stratum.
values I , for rectangular foundations with LIB = 2 resting on soil layers with varying thickness
obtained using the three methods.
For this particular problem, the Poisson’s ratio of the soil is shown to have a significant
influence on the deformation of the foundation. For the incompressible soil, the solutions for
LAYERED SOIL MEDIA 11
0.58 1 2E E 2E
B m fRigid
Case 1 Case 2 Case 3
I , (v = 0 3 ) I , (v = 04999)
Equivalent Equivalent
Case Square circle Square circle
1 0-454 0.464 0376 0377
2 0370 0375 0.322 0322
3 0.297 0.300 0.269 0.270
rectangular foundations with increasing lengthbreadth ratios converge rapidly to the plane strain
case. The plane strain solutions are obtained using the finite element method. Comparing Figures 4
and 7 indicates that even at HIE = 6, the rigid base has an important influence on the deformation
of the foundation. This influence is more pronounced for foundations with higher lengthbreadth
ratios and foundations where the soil tends towards a state of incompressibility.
Table V. Displacement influence values I I for rectangular foundations on a layered soil medium
(a) v = 0-3
(b) v =0.4999
in the two solutions is only about 2 per cent. Thus, the assumptions made in the procedure appear
to be reasonable.
Settlement of rectangular foundations on layered soil may be estimated by superposing the
separate compression of each layer obtained using single layer solutions. The reader is referred to
the book by Simons and Menzies13 for a detailed description of the method. The single layer
solutions commonly used are based on the works of Janbu et a/.” However, as discussed in the
foregoing subsection, Janbu et a!.” solutions are for determining the average settlement of
uniformly loaded, flexible rectangular foundations which generally overestimates the settlement of
equivalent smooth, rigid rectangular foundations. Thus, the more appropriate single layer
solutions for smooth, rigid rectangular foundations may be obtained from the present work, and
they are given in Figure7. The expression for the displacement influence value If is given in
equation (12).
Table V compares the solutions for rectangular foundations resting on the layered soil profile
shown in Figure8 obtained using the present approach (i.e. the numerical procedure) and those
-
obtained by superposing the single layer solutions. For soil with a constant Poisson’s ratio v = 0.3,
the solutions by the two methods agree reasonably well; the maximum differences for rectangular
foundations with various LIB ratios for cases 1,2 and 3 are about 4,5 and 13 per cent, respectively.
E(01
1 I \
Rigid
Figure 9. Rectangular foundation on a layer with stiNness increasing linearly with depth
LAYERED SOIL MEDIA 13
4 1 1 I I L y 7
I I I I
(C) 3 :0.3
I2
E (0 I/E(B)
Figure 10. Displacement influence values for smooth, rigid rectangular foundations on a layer with stiffness increasing
linearly with depth (a) LIB = I, (b) 4 8 = 2; (c) 4 8 = 4; (d) 4 B = 8
Whereas for soil with a constant Poisson’s ratio, v = 0.4999,the maximum differences for cases 1,2
and 3 are about 8, 11 and 32 per cent respectively. The solutions are less satisfactory in cases where
the soil stiffness decreases with depth, in particular when v=O4999. However, in view of the
uncertainties involved, for example in the estimated elastic soil properties and their distribution,
the solutions may well be acceptable in practice. Thus, it appears that for most practical problems,
the single layer solutions in Figure7 can be superposed to give a reasonable estimate of the
settlement of smooth, rigid rectangular foundations resting on the surface of a layered soil medium;
in particular in soil where the stiffness increases with depth.
Rectangular foundation on soil layer with stiflness increasing linearly with depth
There are wide practical applications in considering a soil layer which has a stiffness that
increases linearly with depth. This problem is depicted schematically in Figure 9.The Young’s
14 Y. K.CHOW
in which q is the applied vertical pressure; B the breadth of foundation; E(B) the value of Young’s
modulus at depth B; and I , the displacement influence value. The flexibility coefficients for this
problem are obtained from finite element analyses. The linearly increasing Young’s modulus
within the finite element layer is easily incorporated in the numerical integration process during the
formulation of the element stiffness matrices. Parametric solutions are presented for a constant soil
Poisson’s ratio 1’ = 0.3 and varying degrees of soil inhomogeneity for rectangular foundations with
length/breadth ratios, LIB = 1, 2, 4 and 8 in Figure 10. For H / B = 0.2 (corresponding to a thin
layer), increasing the length/breadth ratio does not affect the deformation of the foundation
significantly for a given applied pressure q. However, as the soil layer thickness increases, this effect
is more pronounced. As the soil becomes more non-homogeneous (i.e. as the value of E(O)/E(B)
ratio decreases), the influence of the layer thickness appears to become less significant
for H / B > I .
CONCLUSIONS
A numerical procedure has been described which enables the vertical deformation of smooth, rigid
foundations of arbitrary shape resting on homogeneous or layered soil media to be determined.
Comparison of this present method with the integral transform technique’ for determining the
response of a square foundation on a homogeneous, isotropic elastic half-space shows that the
present method converges more rapidly and thus is more economical. Similar solutions for
foundations on layered soil media are not available for comparison. Parametric solutions have
been presented for smooth, rigid rectangular foundations resting on a homogeneous half-space, a
homogeneous elastic stratum and a soil layer with stiffness increasing linearly with depth. These
solutions enable a quick estimate of the foundation response under the conditions considered. The
settlement of smooth, rigid rectangular foundations on a layered soil medium has also been
computed for a limited number of cases. It appears that in most practical problems, the solutions
for homogeneous, elastic strata may be superposed to give a reasonable estimate for the layered soil
solutions; in particular when the stiffness of the soil increases with depth.
Although solutions have only been reported on rectangular foundations, the approach is general
enough to consider foundation bases with unusual geometries, e.g. some offshore gravity platforms
have tripod-like base configurations. The interaction of foundations placed in close proximity to
one another may also be assessed.
APPENDIX I: NOTATION
a = Dimension of square subdivision.
B = Breadth ofrectangular foundation.
E = Young’s modulus of soil.
E(O) = Young’s modulus of soil at ground surface.
E(B) = Young’s modulus of soil at depth B.
LAYERED SOIL MEDIA 15
REFERENCES
1. Y.K. Cheung and 0.C. Zienkiewicz, ‘Plates and tanks on elastic foundation-an application of finite element
method’, Int. J . Solids Struct., 1 (4), 451-461 (1965).
2. R. Butterfield and P. K. Banerjee, ‘A rigid disc embedded in an elastic half-space’, Geotech. Eng., 2 (I), 35-52 (1971).
3. P. T. Brown and R. E. Gibson, ‘Rectangular loads on inhomogeneous elastic soil’. J . Soil Mech. Found. Div., Proc.
ASCE, 99 (SMlO), 917-920 (1973).
4. L. J. Wardle and R. A. Fraser, ‘Finite element analysis of a plate on a layered cross-anisotropic foundation’, Proc. First
Int. Conf. Finite Element Methods Engng, University of New South Wales, Australia, pp. 565-578 (1974).
5. R. A. Fraser and L. J. Wardle, ‘Numerical analysis of rectangular rafts on layered foundations’, Geotech., 26 (4),613-
630 (1976).
6. C. M. Gerrard and W. J. Harrison, ‘The analysis of a loaded half-space comprised of anisotropic layers’, Technical
Paper N o . 10, CSlRO Division of Applied Geomechanics, Melbourne (1971).
7. W. D. Carrier and J. T. Christian, ‘Rigid circular plate resting on a non-homogeneous elastic half-space’, Geotech., 23
(1). 67-84 (1973).
8. S . P. Timoshenko and J. N. Goodier, Theory o/Elasticity, 3nd edn, McGraw-Hill, New York, 1970.
9. A. P. S. Selvadurai, Elastic Analysis of Soil-Foundation Interaction, Elsevier, Amsterdam, (1979).
10. P. T. Brown, ‘Numerical analyses of uniformly loaded circular rafts on elastic layers of finite depth’, Geotech., 19 (2)
301-306 (1969).
11. I. M. Smith, Programming the Finite Element Method with Application to Geomechanics, Wiley, New York, 1982.
12. N. Janbu, L. Bjerrum and B. Kjaernsli, Veiledning ved Losning av Fundamenteringsoppgaver, N.G.I. Publication N o .
16 (1956).
13. N. E. Simons and Menzies, A Short Course in Foundation Engineering, Butterworth, London, 1977.