Finite Element Analysis in Geotechnical Engineering Volume Two - Application 6. Shallow Foundations

6.
Shallow foundations
6.1 Synopsis
This chapter considers the analysis of shallow foundations. It begins by defining
shallow foundations and categorising them into two groups, those that are founded
at the ground surface and those that are founded at shallow depth. Each category
is then considered separately and the implications for numerical analysis discussed.
Issues such as the bearing capacity of pre-loaded foundations, the effect of
anisotropic strength on bearing capacity and instability due to insufficient soil
stiffness are discussed. Examples are given, where ever possible, to help clarify
some of the issues raised. The chapter ends by describing some of the analyses
performed for the stabilisation of the leaning Tower of Pisa.
6.2 Introduction
In terms of complexity, shallow foundations are probably the simplest of
geotechnical structures. The loads applied to them are often well defined and their
purpose is to transfer these loads to the soil. This is in contrast to slopes and /or
retaining wall problems, where the soil provides both the activating and resisting
forces and there is therefore a complex interaction between the two.
Shallow foundations come in many shapes and sizes and this chapter begins by
categorising them as either surface or shallow foundations. Each category is then
considered in turn and the implications for finite element analysis are discussed.
Examples are given, where possible, to help clarify some of the issues raised.
These include the analysis of different shaped footings, the bearing capacity of pre-
loaded foundations, the effect of anisotropic strength and instability due to
insufficient soil stiffness. Many of these examples are yet to be published and have
been selected to demonstrate the ability of the finite element method to provide
accurate solutions to conventional problems and also insights into novel problems.
For all analyses an accelerated modified Newton-Raphson scheme, with a sub-
stepping stress point algorithm, was employed to solve the nonlinear finite element
equations, see Chapter 9 of Volume 1. For the plane strain, axi-symmetric and
Fourier Series Aided finite element analysis, 8 noded isoparametric elements were
used, with reduced (2x2) integration. For full three dimensional analysis 20 noded
isoparametric elements were used, with reduced (2x2x2) integration.
Downloaded by [ University of Liverpool] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
Shallow foundations / 21 5
6.3 Foundation types

Shallow foundations may be broadly categorised into one of the following two
types.
6.3.1 Surface foundations

These are foundations which are placed
on, or very near to, the soil surface. They
can come in many shapes and sizes, but
are often categorized as either strip,
circular or rectangular footings, see
Figure 6.1.
Strip Rectangle
6.3.2 Shallow foundations

Shallow foundations are similar to Figure 6.1: Types of surface
surface foundations, except that they are foundations
embedded below the soil surface, see
Figure 6.2. For economic reasons their
depth below the surface is usually not
great, hence the term shallow. Again,
they are often categorized as either strip, Strip
Circle
circular or rectangular. Embedment Rectangle
6.4 Choice of soil model Figure 6.2: Shallow foundations

Conventional foundation design
considers bearing capacity and
deformations separately. Using these
approaches it is not possible to establish
Ultimate
the complete load-displacements curve. It load
is only possible to provide estimates of
the ultimate foundation load and of the
initial gradient of the load-displacement
curve, see Figure 6.3. Displacement
The calculation of bearing capacity is
based on stress field solutions and/or Figure 6.3: Typical load-
limit analysis, combined with some
displacement curve
empirical correlations. The soil is
essentially assumed to behave as an elastic Tresca material, if undrained bearing
capacity is being considered, and as an elastic Mohr-Coulomb material, if drained
bearing capacity is under investigation. Such simple constitutive models can also
be used in numerical analysis and it can be shown, see below, that numerical
analysis can recover the analytical bearing capacity solutions where they are
available. However, while the conventional bearing capacity solutions are only
216 / Finite element analysis in geotechnical engineering: Application
available for simple foundations resting on one soil type, with either a constant
strength or a strength that varies linearly with depth, numerical analysis can
consider complex foundation geometries resting on inhomogeneous foundations.
Numerical analysis can also cope with complex loading combinations (e.g. vertical,
horizontal and moment loading).
Conventional bearing capacity solutions are restricted to either undrained soil
or fully drained soil. They do not therefore provide information on the variation of
bearing capacity with time. This may be important for foundations constructed on
clay soils. Using a coupled numerical approach, see Chapter 10 of Volume 1, it is
possible to simulate such time dependent behaviour, as long as a realistic
constitutive model is used to represent the soil. In this respect the use of a Tresca
model is not appropriate, as the strength of the soil remains constant. Although, in
principle, a Mohr-Coulomb model could be used, this is not advisable as unrealistic
predictions of bearing capacity can be obtained under undrained and partially
drained conditions, see Chapter 9. To obtain realistic predictions it is necessary to
use a strain hardening/softening model of at least the complexity of a simple
critical state model (e.g. modified Cam Clay).
Conventionally, foundation displacements are predicted using elastic theory.
Several approaches are available and they differ in the manner in which the
stiffness parameters are derived. Some assume the soil is linear elastic, others
obtain the stiffness from an oedometer test. While these approaches appear to give
reasonable predictions of average foundation settlement, they do not provide
accurate predictions of differential settlements, deformations under combined
loading, or movements in the soil adjacent to the foundation. This is perhaps not
surprising as the stiffness of the foundation is usually ignored, with the foundation
assumed to apply a uniform pressure to the soil. Numerical analyses have the
ability to deal with a range of constitutive models and can therefore produce more
realistic predictions.
In principle, a soil model that accurately simulates the behaviour of the soils
supporting the foundation under investigation should be used. If the foundation
consists of layered soils, then several soil models might be used. If bearing
capacity is of interest, then clearly the soil model should accurately simulate the
soil strength. If deformations are of concern, and in particular those adjacent to the
foundation, then it is advisable to use a constitutive model that can accurately
represent the nonlinear behaviour of the soil under small strains.
6.5 Finite element analysis of surface foundations

6.5.1 Introduction
If the foundations are subjected to vertical loading only, then a strip footing can be
idealised as plane strain, a circular footing as axi-symmetric, but a rectangular
footing remains a three dimensional problem. Consequently, while it is relatively
straight forward to analyse strip and circular footings under vertical loading, as
they can be reduced to two dimensions, considerably more computer resources are
Shallow foundations / 21 7
required to analyse rectangular footings

which involve a full three dimensional
analysis, although account can often be taken
of the two vertical planes of symmetry to
reduce the complexities of the analysis, see
Figure 6.4 (i.e. in this case only a quarter of
the full geometry needs to be considered).
If the loading is not vertical but inclined,
and the inclination is in the plane Plane of
perpendicular to the length of the footing, symmetry
plane strain analysis can still be performed
for a strip footing. Otherwise, and for both
the circular and rectangular footings, a three
Figure 6.4: Vertical planes of
dimensional analysis is required. For the
symmetry in 3D analysis of a
circular footing this analysis can be
rectangular footing
performed using either a full three
dimensional, or a Fourier series aided approach. A conventional three dimensional
analysis will be required for the rectangular footing.
In general, it is necessary to discretise both the soil and the foundation into
finite elements, see Figure 6.5. This enables the correct stiffness of the foundation
to be included in the analyses. Interface elements can also be positioned between
the underside of the foundation and the soil, so that the correct interface behaviour
can be modelled. However, as the foundation is often very stiff compared to the
soil, numerical problems can arise due to the large difference in stiffness between
adjacent elements on each side of the interface. Although these numerical
instabilities are not so much of a problem with modern computers, where it is
possible to use increased numerical precision (e.g. double or quadruple precision),
surface foundations are often analysed using one of two extreme assumptions (i.e.
a perfectly flexible or rigid foundation). Such assumptions are also used in
conventional soil mechanics practice. If one of these assumptions is made, then
there is no need to include the foundation itself in the analysis.
Foundation Interface elements
\
flfl \
/
\ / \ / \/ \
Figure 6.5: Finite element discretisation of
both soil and foundation
6.5.2 Flexible foundations

If the footing is assumed to be flexible, then it is possible to simulate its behaviour
in a finite element analysis as shown in Figure 6.6. As the footing is flexible, it is
assumed that any loading is uniform and can be represented by a surface surcharge
pressure applied to the surface of the soil immediately below the position of the
footing. As noted in Chapter 3 of Volume 1, such a uniform surcharge load can be
converted to give equivalent nodal forces, AFy. In fact, most computer programs
do this automatically for the user.
If the footing is smooth, Flexible foundation:
then the other boundary Load control Displacement control
rough
condition that must be applied smooth
to the nodes under the footing 1111
is that the horizontal nodal aj plk d
NOT POSSIBLE
forces are zero (i.e. AFX=O). >() \u
Most computer programs -TCu)
assume that AFX=O if no
boundary condition is _. _ _ n
• o. , . Ai ,• ,. r Figure 6.6: Boundary condition options for
specified in the x direction for * * , - , , * .
.. , , flexible footing
a particular node.
Alternatively, if the footing is rough, then the horizontal displacement (Aw) of
the nodes on the soil surface below the position of the footing must be restricted
and therefore set to zero (Aw =0), see Figure 6.6.
It is not possible to apply displacements to the footing and perform the analysis
in displacement control. Consequently, care must be taken when approaching the
collapse load of the foundation, as the application of too much load will cause
unlimited displacements. This will probably manifest itself as a convergence
problem in the nonlinear solution algorithm. Some finite element programs have
facilities for automatically controlling the size of the incremental loads. This is
particularly useful for load control problems and can result in accurate estimates
of limit loads.
6.5.3 Rigid foundations

If the footing is rigid, analyses can be performed under either load ox displacement
control. The various alternatives for the boundary conditions to be applied to the
soil surface below the position of the footing are given in Figure 6.7. As noted in
this figure, for load control it is necessary to tie the vertical displacements of the
nodes below the position of the footing, see Section 3.7.4 of Volume 1. This will
ensure that all the nodes move vertically by the same amount. As the vertical
displacements are tied, the load can be applied as a uniform pressure over the width
of the footing, or as a single point load at the node on the centre line of the footing.
In addition, as with flexible foundations, care must be taken when approaching the
collapse load of the foundation if the analysis is being performed under load
control.
Shallow foundations / 219
Rigid footing:
Load control Displacement control
smooth rough smooth rough
U
1T T 1T r
iF v aj pli< d ap plie d Av a >pli
y(v) ^d =0 AH =t 0
"*(«]
Figure 6.7: Boundary condition options for rigid

footing
If the analysis is performed under displacement control, with vertical
displacements applied to the soil surface below the position of the footing, then the
footing load is obtained from summing the vertical reactions of the nodes which
have been subjected to this displacement. In a nonlinear finite element analysis,
these nodal reactions are best calculated from the total stresses in the elements
connected to the displaced nodes, using:
(6.1)
where {R} is the vector of reactions in the coordinate directions, [B] is the strain
matrix (see Section 2.6 of Volume 1), {&} is the vector of current total stresses,
{&}} is the vector of total stresses existing prior to displacing the footing, and the
volume integration is performed for all soil elements connected to the displaced
nodes. Most finite element software have facilities for obtaining such values.
6.5.4 Examples of vertical loading

6.5.4.1 In troduc tion
As a simple example of the application of numerical analysis, the problem of a
vertically loaded surface footing resting on either an undrained clay or a drained
soil (clay or sand) is considered. To be consistent with conventional foundation
design, the undrained clay and drained soil are modelled using a linear elastic
Tresca and a linear elastic Mohr-Coulomb model respectively. The results of these
analyses are compared with the conventional bearing capacity solutions.
6.5.4.2 Strip footings on undrained clay

Analyses have been performed to obtain the load-displacement curves for rigid
strip surface footings. The soil was assumed to be elasto-plastic, with a Tresca
yield surface and properties: £=100MPa, // = 0.49 and Su = lOOkPa. The mesh
shown in Figure 6.8 was used. The two vertical sides of the mesh have been
restrained in the horizontal direction, while the base of the mesh was not allowed
to move in either the vertical or horizontal direction. Loading was simulated by
applying increments of vertical displacement to the soil surface below the position
of the footing, as described above.
Load-displacement curves for the

strip footing are shown in Figure 6.9.
II ft-r
In conventional design the bearing 1
y1 1
y
\
10m
capacity of a vertically loaded strip
footing on undrained clay is
expressed as:
\/ \ \ 20 m
\
Qam=ANeSy (6.2)
where gmax is the maximum vertical
load applied to the footing, A its area Figure 6.8: Finite element mesh for
and Nc the bearing capacity factor. In strip and circular footing analyses
Figure 6.9 the load is therefore
expressed in terms of the mobilised
1 ^
bearing capacity factor iVcmob (= Q/(A
Si,)), where Q is the load on the
footing. The displacement is
normalised by B, where B is the half
— ^ = 5.19, smooth
width of the footing. Results from --•N = 5.29, rough with interface
c
N = 5.39, rough
c
three analyses are given. One of these
modelled a smooth footing and the
other two a rough footing. The
0.02 0.03 0.04
difference between the two analyses
Normalised settlement,
modelling a rough footing is that in
one analysis a row of interface
elements was added between the soil Figure 6.9: Load-displacement
surface and the underside of the curves for strip footing
footing, see Figure 6.10. In the other
analysis and in the analysis modelling Foundation width
Interface elements
a smooth footing, these interface
elements were not present. The
interface elements were given a shear
strength of lOOkPa and shear and
normal stiffness values of
K=K =\05kWm3 (note that these
stiffness values do not have a major
Figure 6.10: Detail of mesh with
influence on the results). It can be
interface elements
seen that the three analyses give
slightly different load-displacement curves and slightly different Nc values at
failure. These values are quoted on Figure 6.9.
Conventional bearing capacity theory indicates that, for a strip footing resting
on undrained clay with a constant strength, the bearing capacity factor Nc should
be 5.1416 (= 2+7i), for both smooth and rough footings. This solution can be
obtained from limit analysis (i.e. both upper (unsafe) and lower (safe) bound
solutions) and from a closed form plasticity solution (i.e. combination of stress and
velocity field solutions). It is therefore

theoretically exact. The finite element
analyses give values of 5.19,5.39 and t\ \ \ y
5.29 for the analyses with smooth,
rough and rough with interface 4.5 m
elements footings respectively. smooth
Compared to the analytical solution,
the numerical analyses are in error by
0.94%, 4.6% and 2.8% respectively.
Clearly, the error is larger for the
rough footing analyses, and in
particular is larger for the analysis
with no interface elements. The rough
reason for this is partly explained in
Figure 6.11, which shows the vectors
of incremental displacement for the
last increment of each analysis (i.e. at
failure). The orientation of these
vectors indicates the direction of
movement and their length the rough with interface elem.
magnitude of movement. As it is the
orientation of the vectors and their Figure 6.11: Vectors of incremental
relative magnitude that indicates the displacements at failure for three
failure mechanism, the absolute different soil-footing interfaces
magnitude of the incremental
displacements is irrelevant and consequently no magnitude scale is given on the
figure.
Inspection of these failure mechanisms indicates that some horizontal
movement occurs at the soil surface immediately below the smooth footing. For
the rough footing this is not possible due to the boundary conditions.
Consequently, the failure mechanism for the rough footing is slightly deeper and
wider than that for the smooth footing. This in rum implies a larger area to the slip
surface which, when combined with a constant Su, leads to a greater ultimate
footing load and hence Nc.
The error for all three analyses arises because of the nature of the finite element
mesh immediately below and adjacent to the footing. As shown in the insets on
Figure 6.11, the vectors of displacement adjacent to the edge of the rough footing
change direction very abruptly, from vertically downwards under the comer of the
footing to an angle of 45° to the horizontal in an upward direction, on the soil
surface adjacent to the footing. This change occurs over the single element
positioned at the comer. The accuracy of the solution will therefore depend on the
ability of this single element to accommodate this rapid change in displacement.
The rapid change in displacement at the edge of the footing is less pronounced in
the smooth footing analyses and in the analysis where interface elements have been
used. It can be seen from the inset on

Figure 6.11 that for the latter analysis
some horizontal movement of the soil
surface immediately under the edge of - - • ^ = 0.5
the footing occurs. This is possible

because the interface elements in this
vicinity are plastic and although the
tops of these elements are constrained
to move vertically downward, the
i\• 0.01 0.02
: 0.03 0.04 0.05
Normalised vertical displacement, 8/fl
bottoms can have a horizontal
component of movement (note the
displacements of the tops of the Figure 6.12: Effect of Ko on
interface elements are not shown on load-displacement behaviour
Figure 6.11).
Su = 100 kPa, E= 100 MPa, B = lm
a)
b)
c)
Figure 6.13: Effect of Ko on failure mechanism and

plastic zone beneath the footing
As shown in Chapter 9 and by Day and Potts (2000), more accurate solutions
can be obtained by using smaller elements around the edge of the footing.
However, these solutions involve larger computational times because of the
singularity at the edge of the footing. Hence, for the work described in this chapter
the mesh shown in Figure 6.8 has been retained. The qualitative comparisons that
are made are therefore likely to be correct, but it should be remarked that more
accurate quantitative results could have been obtained by using a more refined
mesh.
The effects of different initial stress conditions in the ground are shown in
Figures 6.12 and 6.13. Results from analyses with Ko values of 0.5,1.0 and 2.0 are
shown. All analyses had the same soil properties (i.e. same stiffness and strength),
and had a saturated bulk unit weight of 20kN/m3 and a ground water table at the
soil surface. The load-displacement curves shown in Figure 6.12 indicate that Ko
has only a minor influence on the behaviour of the foundation.
Figure 6.13 shows vectors of incremental displacement at three stages, SIB =
0.006, 0.02 and 0.05, for each analysis, where S is vertical displacement and B is
the half width of the footing. Also shown on these plots are the zones of soil which
have become plastic. All three analyses indicate that at (5/5=0.02 the vectors of
incremental displacement show significant relative movements at depth below the
footing. This is accompanied by an extensive zone of plastic soil. However, at
failure the pattern of displacements have changed and the final failure mechanism
involves quite shallow movements. There is some indication that the zones of
yielded soil extended further laterally and less vertically as Ko increases.
6.5.4.3 Effect of footing shape on the bearing capacity of

undrained clay
Further analyses have been „
performed to obtain the load-
displacement curves for rigid m\-v\ >
circular and square surface
\\
50 m
foundations. The same soil

properties, as used for the strip 50 m
foundations, were adopted. S>
For the circular footing, the
mesh of 8 noded elements
shown in Figure 6.8 was used,
while for the square footing
Figure 6.14a: Horizontal cross-section of
the mesh of 20 noded elements
a 3D mesh for square footing analysis
shown in Figure 6.14 was
used. As can be seen from Figure 6.14b, the aspect ratios of the elements at depth
below the edge of the footing are not ideal. However, as they are a considerable
distance from the footing, they did not appear to adversely affect the results.
Horizontal movements on the vertical boundaries of the mesh were restrained,
while the base of the mesh was not allowed to move in either horizontal or vertical
directions. Loading of the footing was 5m

again simulated by imposing vertical
displacements to the soil surface <a
below the position of the footing.
A comparison of the load-
displacement curves for strip, circular
and square footings is shown in 6
©
Figures 6.15 and 6.16 for smooth and 50 m
rough footings respectively. Again,
the load is expressed in terms of the a>
mobilised bearing capacity factor
Ncmoh(=Q/(ASu)\ and the displacement
is normalised by B, where B is the
half width of the footing, for the strip
and square foundations, and the Figure 6.14b: Vertical cross-section
radius for the circular foundation. It of a 3D mesh for square footing
should be noted that, while there is an analysis
analytical solution for the ultimate
bearing capacity factor for the strip
foundation (i.e. Nc = 5.14) and a semi-
analytical/numerical solution for the — Strip N = 5.19
c
Circle^ = 5.87
circular foundation (i.e. Nc = 5.69 and - - - Square N = 5.72 c
6.2 for smooth and rough footing

respectively), there is no such solution
for the square foundation. However, it
is commonly assumed that circular 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
and square footings have the same Normalised displacement, hIB
bearing capacity, although there is no Figure 6.15: Load-displacement

theoretical justification for this curves for smooth strip, circular and
assumption. The finite element results square footings
indicate that the bearing capacity of a
square foundation is, in fact, less than I * 7
that of a circular foundation. o 6
Analyses have also been
performed for rectangular footings Rough:
— Strip tf = 5.39
with length to breadth ratios, LIB, of 2 e
Circle N. = 6.52
% 3 --Square JV = 6.37
and 4. These were performed with
f
e
meshes similar to that used for the

square footing, but with additional
elements added to account for the 0.02 0.03 0.04 0.05
increase in length of the footing. The Normalised settlement, 8/B
resulting load-displacement curves
Figure 6.16: Load-displacement
are presented in the form of mobilised
moh curves for rough strip, circular and
bearing capacity factor, Nc , versus
square footings
vertical normalised displacement of

the footing, SIB, in Figures 6.17 and
6.18, for smooth and rough footings
Smooth:
respectively. LIB = 1/1 (5.72)
Conventionally, the undrained 1/5 = 2/1(5.63)
1/5 = 4/1(5.45)
bearing capacity, Qmax , of a
rectangular surface footing resting on
a clay soil is expressed in terms of the
ultimate bearing capacity factor of a 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
strip footing, 7Vcstrip, and a shape Normalised settlement, d/B
factor, sc, by the following equation:

Qmax = AscNfnpSu (6.3) curves for smooth rectangular
The shape factor sc depends on the footings
LIB ratio of the footing and is usually
determined from the following
empirical relationship (Skempton
1951):
Rough:
LIB = 1/1 (6.37)
sc = l + 02 B/L (6.4) LIB = 2/\ (6.02)
LIB = 4/1 (5.72)
This relationship does not distinguish
between rough and smooth footings.
The results presented in Figures 6.17
and 6.18 can be used to calculate 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
values of sc for rectangular footings, Normalised settlement, 6/B
with LIB values of 1, 2 and 4. These

are compared with Equation (6.4) in Figure 6.18: Load-displacement
Figure 6.19. curves for rough rectangular
footings
6.5.4.4 Strip footings on
weight/ess drained 1.3
soil 1.25 Skempton's empirical expression

/ % Rough footing
In conventional design the bearing ., \ f • Smooth footing
capacity of a vertically loaded strip
1.15 \
footing on a drained soil (i.e. sand) is
expressed as: 1.1
• — - ^ ^ ^
1.05
^ ^
1
(6.5) 0 1 2 3 4 5
Ratio of footing dimensions, LIB
where B is the half width of the

footing, q' is the magnitude of any Figure 6.19: Variation of shape
surcharge pressure existing on the factor with L/B
ground surface adjacent to the
footing, see Figure 6.20, and Nc, Nq and Ny

are bearing capacity factors. The factors Nc
and Nq are derived theoretically assuming the II IIII . j , I I III I
soil is weightless, while Ny is found from an //Al} • • • l ^ f e ^ ^ 1111 \y/
approximate calculation assuming the soil ^ H
has weight, but no cohesion or surcharge.
Superposition is then assumed to give Figure 6.20: Strip footing
Equation (6.5). with surface surcharge
Analytical stressfieldsolutions have been
used to obtain the following expressions for N and Nc (Prandtl (1920)):
(6.6)
(6.7)
For the situation where the angle of dilation equals the angle of shearing resistance
(i.e. v = cp'\ it is possible to show that a compatible displacement mechanism is
associated with these stress fields. It is also possible to extend the stress fields
throughout the soil mass, satisfying equilibrium and without violating the yield
condition. The solutions expressed by Equations (6.6) and (6.7) are therefore
theoretically exact, see Chapter 1 of Volume 1. The solutions can also be shown
to be applicable to both smooth and rough footings. If the angle of dilation, v, does
not equal the angle of shearing resistance, q>\ then the above solutions are only
approximate.
To assess the ability of numerical analysis to predict these results, finite element
analyses have been performed using the geometry and finite element mesh shown
in Figure 6.8. The Mohr-Coulomb model was used to model soil behaviour, see
Section 7.5 of Volume 1. The soil was assumed to be weightless (y = 0) and
cohesionless (c'= 0), with an angle of shearing resistance, cp', of 25°, a Young's
modulus, E', of lOOMPa and a Poisson's ratio, ju, of 0.3. Four analyses were
performed with a surcharge, q', of lOOkPa and one analysis with q' = lOkPa. The
four analyses with q' = lOOkPa differed in that different combinations of the angle
of dilation and roughness of the footing were assumed. For the analysis with q' =
lOkPa the footing was smooth and the angle of dilation was zero. For the analyses
with rough footings interface elements were positioned between the footing and the
soil, see Figure 6.10. These were given the same values ofq>' and v as the soil, and
stiffness values K=K =\05 kN/m3.
The predicted load-displacement curves are shown in Figure 6.21. The load is
expressed as the mobilised bearing capacity coefficient, A^mob (=Q/(Aq')). This
figure shows that there is little effect of the footing roughness, but that the angle
of dilation affects both the shape of the load-displacement curve and the ultimate
value of N(j. The magnitude of the surcharge load, q\ affects the amount of
displacement required to reach failure. However, it does not affect the ultimate
value of NCj.
For <p'=25° Equation (6.6) gives a smooth, v = 0°, q' =10 kPa
value of Nq = 10.7. The ultimate

(failure) values ofNq for each analysis
are listed in Table 6.1. As noted
rough, v = 25°
above, Equation (6.6) is theoretically smooth, v = 25°
rough, v = 0°
exact for a soil with an angle of Q
smooth, v = 0°
dilation equal to the angle of shearing

resistance and is valid for both
smooth and rough footings. The 0.05 0.1 0.15 0.2
Normalised settlement, bIB
analyses in which v = 25° can
therefore be directly compared with
the analytical value and the errors are Figure 6.21: Load-displacement
0.9% and 3%, for the smooth and curves for strip footings on
rough footing respectively. As with weightless soil (effect of roughness,
the analyses for undrained clay, the dilation and surcharge)
numerical predictions are improved if
a finer mesh is used in the vicinity of the edge of the footing. The three analyses
with a dilation angle of zero give a value of Nq of approximately 10.0. This is
some 7% lower than value given by Equation (6.6).
Vectors of incremental displacement at failure for the four analyses with
#'=100kPa are shown in Figure 6.22. These show that for all four analyses the
failure mechanism is deep seated and has a wide lateral extent.
10 m
smooth, v = 0°
, f = 25°
rough, v = 0 rough, v = 25°
Figure 6.22: Effect of roughness and dilation on

the failure mechanism for a strip footing on
weightless soil
6.5.4.5 Strip footings on a drained soil

If the soil is cohesionless, has no surcharge but has weight, then only the third
term in Equation (6.5) is non-zero. As noted in Section 1.9.2 of Volume 1, it is
rarely possible to obtain analytical stress field solutions for a frictional soil (i.e.
cpf>Q) which has weight, due to the complexity of the governing hyperbolic partial
differential equations. Consequently, there is no theoretically exact expression for
Ny. In fact, several alternative expressions can be found in the literature. One of the
most popular is that by Hansen (1970):
Ny=l5(Nq-l)ton<p' (6.8)
This equation, as with many of the alternatives, is often used for both rough and
smooth footings. However recent work by Bolton and Lau (1993) has shown that
footing roughness affects the magnitude of Ny. Their results are based on numerical
integration of the stress field equations and are therefore, again, approximate.
Results from four finite element |
analyses, using the mesh shown in
s" rough, v = 0
Figure 6.8 and with the properties rough, v = 25°
given above for the A^ analyses, are
shown in Figure 6.23. Dry conditions
smooth, v = 0°
were assumed and the soil had a bulk smooth, v = 25°
3
unit weight of 18kN/m and K =0.511
(i.e. 1 -sin#?'). A different combination
of footing roughness and angle of soil
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02
dilation was used in each analysis. Normalised settlement, 8/B
The results are expressed as graphs of
mobilised Nymob against normalised
footing displacement, SIB. In contrast
curves for a strip footing on
to the analyses presented above for
cohesion/ess soil
Nq, the final value of Ny is dependent
(effect of roughness and dilation)
on footing roughness, but not on the
magnitude of the angle of dilation.
The final values of NY are tabulated in ,
Table 6.2, where they are also
compared with values from Bolton
and Lau (1993) and from Equation
4.5 m
(6.8). It can be noted that for smooth
footings the finite element results are smooth strip, v = 0° smooth strip, v = 25"
in agreement with those of Bolton and Ko = 0.577, f=25°
Lau (1993). However, for the rough
footings the numerical analyses are in
much better agreement with Equation
(6.8).
The vectors of incremental
displacement at failure are shown in rough strip, v = 0"
Figure 6.24. The smooth footings

have much shallower failure Figure 6.24: Failure mechanism for
mechanisms and the difference a strip footing on cohesionless soil
between these and those for the rough (effect of roughness and dilation)
footings explains the difference in the Ny values.

Vectors of incremental displacement at three different stages of loading for the
analysis with a smooth footing and with a zero dilation angle are shown in Figure
6.25. Also shown on these figures are the zones of plasticity. As for the undrained
analyses described above, the mode of deformation changes as loading proceeds.
The zone of plasticity is much more extensive than the final failure mechanism.
Drained
smooth strip:
K. = 0.577
Figure 6.25: Development of failure beneath a smooth

strip footing on cohesion/ess soil
In all of the Ny analyses presented above Ko = 0.577. To investigate the effect

of Ko, the analysis with a smooth footing and zero dilation was repeated with Ko
values of 1.0 and 2.0. The resulting | 4
load-displacement curves are shown ^
in Figure 6.26. The value of Ko has an
effect on the load-displacement curve
prior to failure, but does not affect the
ultimate value of Ny. The vectors of
incremental displacement at failure
for each analysis are compared in
Figure 6.27. Also shown on this 0.004 0.006 0.008 0.01 0.012
figure are the plastic zones at failure. Normalised settlement, b/B
While the failure mechanism is
independent of Ko , the size of the Figure 6.26: Effect of Ko on load-
plastic zone is not. displacement behaviour
Figure 6.27: Effect of Ko on failure mechanism and

development of plastic zone
If a surcharge loading exists on the soil surface adjacent to the foundation and
the soil has weight, then both the Nq and Ny terms of Equation (6.5) are non-zero.
As noted above, this equation assumes that the effects of soil weight, y, and
surcharge, q\ can be superimposed. To investigate this hypothesis the analysis
presented above for a rough footing on sand with an angle of dilation v = 25° was
repeated with a surcharge q' = lOkPa. This analysis gave an ultimate load on the
footing Qmax = 556 kN/m. Based on Nq=U.03 (see Table 6.1) and Ny = 6.72 (see
Table 6.2), Equation (6.5) predicts an ultimate load of 463 kN/m. Comparing these
values indicates that the superposition assumption implied in Equation (6.5) is
conservative. In this particular case by 17%.
One reason why the superposition assumption is not valid is that the failure
mechanisms associated with Nq and Ny differ. This can be seen by comparing the
failure mechanisms shown in Figures 6.22 and 6.24. The failure mechanism for Nq
is deeper and wider than that for Ny.
It should also be noted that the magnitudes of the footing settlements required
to mobilise full Nq and Ny values differ (see Figures 6.21 and 6.23). In addition, the
settlement required to mobilise the ultimate Nq value depends on the magnitude of
the surcharge, q'. Conventional design procedures do not explicitly account for
this. However, in practice this could have significant implications, especially in
situations where the soil strength degrades (i.e. <pf reduces) with straining. For
example, at ultimate load the average strength associated with the first mechanism
to form (i.e. Ny) is likely to have decreased from its peak value, while that
associated with the second mechanism (i.e. A^) is unlikely to have reached its peak
value. In such a case Equation (6.5) may not be conservative (see Chapter 4 for a
detailed discussion on progressive failure).
6.5.4.6 Circular footings on a weightless drained soil

To evaluate the bearing capacity for a vertically loaded circular footing on a
drained soil Equation (6.5) is again used in conventional design. However, the
bearing capacity coefficients Nq, Nc and Ny are modified to account for the circular
shape of the footing. This can be done by using different values for N(J,NC and Np
or by maintaining the values associated with a strip footing, but introducing shape
factors sq, sc and sy in a similar manner as shown in Equation (6.3) for undrained
clay.
In contrast to the strip footing, where the expressions given by Equations (6.6)
and (6.7) for Nq and Nc are theoretically exact, for a circular footing all of the
bearing capacity factors used in conventional design are approximate. To quantify
the effect of footing shape on Nq the finite element analyses, presented above in
Section 6.5.4.4 for weightless sand with <7'=100kPa, were repeated with a circular
footing.
The predicted load-displacement curves are shown in Figure 6.28 and these can
be compared with those for the strip footing given in Figure 6.21. In contrast to the
strip footing analyses, which indicate that the footing roughness does not affect the
ultimate value ofNq, the circular footing analyses indicate that the rougher the
footing the larger the ultimate Nq

value. In addition, and again in
contrast to the results for the strip
footing, the angle of dilation has only
a minor influence on the ultimate
value ofN(J.
One of the reasons for the
differences outlined above can be
seen in Figure 6.29, which shows 0.2 0.3 0.4 0.5
Normalised settlement, d/B
vectors of incremental displacement
at failure. These vectors indicate that
the failure mechanisms for the smooth Figure 6.28: Load-displacement
circular footings are shallower and of curves for a circular footing on
a smaller lateral extent than those for weightless soil
the rough circular footings. (effect of roughness and dilation)
Comparison of these failure
mechanisms with those for the strip footings shown in Figure 6.22 indicates that
even for the rough circular footings the failure mechanisms are smaller both
laterally and vertically than for the strip footings.
10m
smooth, v = 0 smooth, v = 25
rough, v = 0° rough, v = 25°
Figure 6.29: Failure mechanism for a circular

footing on weightless soil
The ultimate values of Nq are listed in Table 6.1, where they are compared with
the equivalent values from the strip footing analyses. The values for the circular
footings are substantially larger than those for the equivalent strip footings. This
is in agreement with conventional design practice which, depending on the actual
design code or advice manual, suggests a shape factor of between 1.2 and 1.5 to
be applied to the Nq value for a strip footing.
Table 6.1: Bearing capacity factor Nq
A^ at failure
Analysis
Strip footing Circular
footing
Smooth footing, q' = lOOkPa, v = 0° 10.03 15.57
Smooth footing, q' = lOOkPa, v = 25° 10.8 16.01
Rough footing, q' = lOOkPa, v = 0° 9.92 18
Rough footing, q' = lOOkPa, v = 25° 11.03 19.42
Smooth footing, q' = lOkPa, v = 0° 10
6.5.4.7 Circular footings on a drained soil

To investigate the effect of foundation
shape on the Ny term in Equation (6.5)
the analyses presented in Section
6.5.4.5 were repeated using an axi-
symmetric, as opposed to a plane
strain, idealisation. The resulting
load-displacement curves are shown
in Figure 6.30 and can be compared
to those from the equivalent strip
footing analyses presented in Figure 0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02
Normalised settlement, 8/B
6.23. The two sets of analyses show
similar behaviour, with the exception
that the ultimate Ny values are smaller Figure 6.30: Load-displacement
for the circular footings. This is in curves for a circular footing on
contrast to the Nq analyses, which cohesionless soil
showed that higher values were (effect of roughness and dilation)
obtained for circular footings.
The ultimate values of Ny are listed in Table 6.2, where they are compared to
those from the strip footing analyses and also to those obtained by Bolton and Lau
(1993). Comparing the values obtained from the finite element analyses, the ratios
of circular to strip footing Ny values are 0.88 and 0.98, for the smooth and rough
footings respectively. Compared to the values given by Bolton and Lau (1993), the
smooth footing results are in reasonable agreement, whereas the rough footing
results are not. In fact, Bolton and Lau give values of Ny for rough circular footings
which are substantially larger than for the equivalent strip footing. This is not
predicted by the finite element analyses and is also not in agreement with
conventional design practice, where shape factors of between 0.6 to 0.9 are
recommended (Sieffert and Bay-Gress (2000)).
Table 6.2: Bearing capacity factor NY

Bolton & Lau Equation (6.8) Finite elements
Analysis
Strip Circle Strip Circle Strip Circle
Smooth, v = 0° 3.51 3 6.76 - 3.59 3.14
Smooth, v = 25° 3.51 3 6.76 - 3.74 3.31
Rough, v = 0° 11.6 13.5 6.76 - 6.74 6.62
Rough, v = 25° 11.6 13.5 6.76 - 6.72 6.64
The vectors of incremental displacements at failure are shown in Figure 6.31.

Comparing these with those for the equivalent strip footings given in Figure 6.24,
indicates that for circular footings the failure mechanisms are much shallower and
of smaller lateral extent.
4.5 m
smooth circle, v = 0°
Ko = 0.577, <j>'=25°
rough circle, v = 0° rough circle, v = 25°
Figure 6.31: Failure mechanism for a

circular footing on cohesion/ess soil
6.5.5. Undrained bearing capacity of non-homogeneous clay

In Sections 6.5.4.2 and 6.5.4.3 discussing the undrained bearing capacity of clay,
the strength was assumed to remain constant with depth. In many field situations
this is not a realistic assumption. For some simple linear and bi-linear strength
variations with depth, approximate analytical solutions are available (Davis and
Booker (1973)), however for more general situations such solutions do not exist.
6.5.5.2 Constitutive model

Accounting for a variation in undrained strength with depth is relatively easy in a
finite element analysis. A Tresca model could be employed, in which the undrained
strength is allowed to vary spatially across the finite element mesh. The analysis
could then be carried out as described above.
However, in anticipation of the content of Section 6.5.6 of this chapter, and to
illustrate the use of a more complex constitutive model, a form of the modified
Cam Clay model will be used here. In this model the shapes of the yield and plastic
potential surfaces are given by a Mohr-Coulomb hexagon and a circle respectively
(see Sections 7.9.2 and 7.12 of Volume 1). Both stiff and soft clay soils are
considered and the parameters, listed in Table 6.3, have been selected to represent
real soil types. It may be noted that these parameters are essentially effective stress
parameters. It is not possible to input directly the undrained strength or its change
with effective stress. However, as shown in Appendix VII.4 of Volume 1, the
undrained strength can be derived from the basic parameters and is given by the
following equation:
OCy
(6.9)
NC
sirup
where:
smOsirup'
cos<9 +
B=-
6 is the Lode's angle; OCR is the overconsolidation ratio; oj is the

initial vertical effective stress.
In the present analyses, KONC and Ko0C have been assumed to be given by:
KOC (6.10)
= \-sirup' =
Table 6.3: Soil properties for pre-loaded strip footings

Soil K X 2nd elastic y Vi k
type parameter (kN/m3) (m/sec)
Stiff 23° 0.03 0.16 //=0.2 18 2.84 1 x 10 1 0

clay
Soft 32° 0.02 0.22 G=1700kPa 17 3 5 x 10 1 0

clay
The stiff clay parameters are based on the

behaviour of London Clay. An OCR of 6 has
been used and the water table has been
OCR = 6
assumed to be 2.5m below the ground
surface. Above this level the clay remains • Data from triaxial
compression tests
saturated and can sustain tensile pore water
pressures. The undrained strength profile is
shown in Figure 6.32. Also shown in the
figure is field data for London Clay from
Carswell^a/. (1993).
The soft clay parameters are based on site
investigation and laboratory data from a site
in Grimsby, Yorkshire (Mair et ah (1992)).
At this site the clay is normally consolidated
-10
below a depth of approximately 2m, with a 50 100 150 200
stronger crust at the surface. The water table Undrained strength, Su (kPa)
has been assumed to be at a depth of 2m and,
as with the stiff clay, the soil above this Figure 6.32: Undrained
surface is assumed saturated and able to strength profile fitted through
sustain tensile pore water pressures. To London Clay data
obtain a realistic distribution of undrained
strength with depth, the OCR has been varied as shown in Figure 6.33. Also shown
in this figure is the variation ofKooc, according to Equation (6.10). The resulting
distribution of undrained strength Su, together with the field data, is show in Figure
6.34. The values of Su plotted are those appropriate to triaxial compression.
0
-2
7- < 0
-2
-<
H i Fitted
Q -6 Q -6 profile
OCR
-8
-10 i i -10
0 " 2 4 0 10 20 30 40
Undrained strength, Su (kPa)
Figure 6.33: OCR and Ko Figure 6.34: Undrained strength

profiles for soft clay profile fitted through soft clay
data
6.5.5.3 Geometry and boundary conditions

Analyses have been performed of 2m and 10m wide rigid strip footings. As
undrained conditions are required, two alternative types of analysis are available.
A fully coupled analysis, combining mechanical deformation and pore water flow
could be performed, see Chapter 10 of Volume 1. Alternatively, only the
mechanical behaviour could be considered, using a large value for the bulk
compressibility of the pore fluid, see Section 3.4 of Volume 1. Although the latter
approach is the simplest and probably the best option, the first approach is used
here. This is because the analyses are to be further extended to drained conditions
in Section 6.5.6 of this chapter.
The finite element mesh for the 2m wide footing is that shown previously in
Figure 6.8. Eight noded plane strain isoparametric elements have been used, with
the four corner nodes having pore water pressure, as well as displacement, degrees
of freedom. The two vertical sides of the mesh have been restrained in the
horizontal direction, while the base of the mesh was not allowed to move either in
the vertical or horizontal direction. Loading of the footing was modelled by
applying increments of equal vertical displacement to nodes located at the soil
surface underneath the position of the footing. Both smooth and rough footings
have been simulated. For the rough footings the horizontal movement of the nodes
immediately below the footing has been restrained.
Throughout the analyses no flow of water has been allowed through the base
of the mesh, immediately beneath the footing, across the ground surface adjacent
to the footing, nor through the left hand boundary which forms the vertical plane
of symmetry through the footing. On the right hand boundary the pore water
pressures have been maintained equal to their original values determined by the
position of the water table. As the footing was loaded rapidly to failure in 21 days
and the permeability of the soil was low (see Table 6.3), undrained conditions
prevailed in the soil and therefore the pore pressure boundary conditions on the
right hand side of the mesh had little influence on the results. Similar results would
have been obtained if it had been assumed to be a no flow boundary, however such
a boundary condition was not appropriate for the intended analysis discussed in
Section 6.5.6 of this chapter.
For the analyses performed with a 10m wide strip footing, the mesh dimensions
have been increased by a factor of 5. The soil properties and initial conditions
have, however, been left unchanged.
6.5.5.4 Failure mechanisms

For a smooth strip footing on a clay soil, having a constant undrained shear
strength, two classes of failure mechanisms are theoretically possible: the Hill (Hill
(1950)) and the Pmndtl (Prandtl (1920)) mechanisms. These are shown graphically
in Figure 6.35. For a rough strip footing the Hill mechanism is not appropriate as
it implies horizontal soil movements at the soil-footing interface. Consequently,
only the Prandtl mechanism is valid for a rough strip footing. Interestingly, for a
soil with a constant undrained shear strength, both the Hill and Prandtl solutions
give the same value of the bearing capacity

factor Nc (i.e. N=2+n). As the Prandtl
solution is valid for both smooth and rough
footings, this also implies that in this case the
bearing capacity is independent of footing a) Hill mechanism of failure
roughness. Q
2B IB ^ IB
While the above is true for a soil with a
constant undrained shear strength, Stn it does
not hold if Su varies spatially (Davis and
Booker (1973)). The effects of a spatial
variation of S,f on the failure mechanism and
bearing capacity can be seen from the results b) Prandtl mechanism of failure
of the present analyses. Vectors of
incremental displacement at failure, for a g 6 3 5 ; pQssjble
selection ofthe analyses, are shown in Figure mechanjsms of footing failure
6.36. The vectors clearly indicate the nature
and extent ofthe failure mechanism for each analysis.
Results from 2m wide smooth and rough footings on stiff clay are shown in
Figures 6.36a and 6.36b respectively. The failure mechanism for the smooth
footing is essentially of the Hill type, with soil immediately under the footing
having both vertical and horizontal components of displacement. In contrast, the
failure mechanism for the rough footing is more like the Prandtl type, with the soil
immediately below the footing moving predominantly vertically. Clearly, the
failure mechanism for the rough footing penetrates deeper below the footing and
involves a considerably larger volume of soil than that for the smooth footing. This
occurs because the horizontal restraint imposed at the soil-footing interface causes
the failure mechanism to penetrate deeper into the soil for the rough footing case.
As the undrained shear strength increases with depth, see Figure 6.32, then it is
perhaps not surprising that the bearing capacity for the rough footing is higher than
for the smooth footing (note: the bearing capacities are noted on Figure 6.36 as
QmJA).
Because the initial distribution of 5fl with depth is linear for the stiff clay, it is
possible to obtain an estimate ofthe bearing capacity using results from Davis and
Booker (1973). This gives a value for the smooth footing of Qmax/A=\45 kPa. To
obtain this value interpolation and scaling from the figures supplied in the above
paper are necessary. Davis and Booker's figures are based on calculations which
involve some finite difference approximations and have been drawn to provide
conservative estimates for design purposes. The above result is therefore subject
to error. Nevertheless, this value of ultimate bearing capacity is within 5% ofthe
finite element result.
Similar results are also shown in Figures 6.36c and 6.36d, which are from
analyses of a 10m wide footing on soft clay. These results imply that the surface
crust, see Figure 6.34, does not have a major influence on the failure mechanism.
This is not so for the 2m wide footing on soft clay, as can be seen in Figures 6.36e
and 6.36f. In this case a deep Prandtl type mechanism is predicted for both the
smooth and rough footing. Here the failure mechanism, the extent of which
depends partly on the footing width, is controlled by the 2m deep surface crust
which has a strength reducing with depth. The failure mechanisms for both the
smooth and rough footings are forced into the weaker soil, as this provides a failure
mechanism involving the least resistance. As the failure mechanism is the same for
both smooth and rough footings, then the bearing capacity is also the same, which
is evident from the results shown on Figures 6.36e and 6.36f.
4.5m •
a) 2m smooth footing on stiff clay b) 2m rough footing on stiff clay
, 7 0 » « ^ = 58kPa
• 1
22.5m •
c) 10m smooth footing on soft clay d) 10m rough footing on soft clay
l = 61kPa
e) 2m smooth footing on soft clay f) 2m rough footing on soft clay
Figure 6.36: Failure mechanisms for smooth and

rough footings on stiff and soft clay
6.5.6 Undrained bearing capacity of pre-loaded strip

foundations on clay
The bearing capacity of a soil is essential for the design of near surface
foundations. It controls the load which can be supported by the ground and will
influence the size and type of foundation to be used. The problem is well
understood for both undrained and drained loading situations, where the soil
strength parameters are either constant or vary in a simple manner with depth, and
there are several different ways, theoretical and empirical, of determining the
bearing capacity (see above). The estimation of the bearing capacity of new
foundations is therefore relatively straight forward and in this respect it is the short
term (undrained) bearing capacity that usually governs the design of foundations
on clay soils.
The situation is not so clear, however, if extra load is to be added to an existing
foundation on clay soil, some time after initial construction was completed. This
could occur, for example, if new machinery is to be installed, or if additional floors
are to be added. It can also occur if old foundations are to be reused for a new
building, a situation that is becoming more common as the civil engineering
profession is becoming more sensitive to issues related to sustainability. Due to
consolidation of the soil after initial loading there is a change in undrained shear
strength. The undrained bearing capacity at the time the additional load is applied
will therefore differ from that during initial construction. At present, there are no
guidelines for determining the magnitude of this change. This is perhaps not
surprising as there will be a complex distribution of undrained strength below the
foundation, which will render simple analytical solutions almost impossible.
This problem is therefore an ideal candidate for finite element analysis and has
been investigated first by Jackson et al. (1997) and then by Zdravkovic et al.
(2001), who extended the range of soil conditions considered. They performed a
series of coupled finite element analyses of strip foundations. A strip footing was
first placed on a clay soil, with known initial conditions, and rapidly loaded to
failure. This provided the initial short term (undrained) bearing capacity. A series
of further analyses were then performed in which the footing was first rapidly
loaded (undrained) to a percentage of the original short term bearing capacity. Pre-
load values of 20, 40, 60, 80 and 100% were used. The load was then held at this
value while all excess pore water pressures in the clay dissipated (i.e. full
consolidation). The footing was then subjected to further rapid loading (undrained)
to failure, and the new undrained bearing capacity determined.
6.5.6.2 Cons titutive model

An important facet of this problem is the change of undrained strength which
occurs during consolidation, as this governs the undrained bearing capacity under
any subsequent rapid increase in loading. Consequently, it is necessary to employ
a constitutive model that can accurately predict such behaviour and therefore the
soil was modelled using a form of modified Cam clay. In fact, the same model and
soil parameters as described in the previous section were used. The undrained
analyses for the stiff clay (OCR=6) and the soft clay described previously, were
used to determine the initial undrained bearing capacities. Additional analyses have
been performed with the stiff clay properties with OCRs of 1, 2, 4, 9 and 25.
6.5.6.3 Geometry and boundary conditions

Analyses have been performed on 2m and 10m wide rigid strip footings and the
finite element mesh for the 2m wide footing is shown in Figure 6.8.
The boundary conditions for the loading stages of the analyses were the same
as those described in Section 6.5.5. However, during the consolidation stages of the
analyses the nodes located at the soil surface underneath the position of the footing
(i.e. those displaced during the undrained loading) were free to displace vertically,
but by the same amount, with the total load on the foundation remaining constant.
This was achieved by using the tied degrees of freedom concept, see Chapter 3 of
Volume 1. The magnitude of the movement that occurred during these
consolidation stages of the analysis was therefore not prescribed, but was a result
of the analysis.
The same hydraulic boundary conditions (i.e. prescribed flows and pore water
pressures), as described in Section 6.5.5, were also maintained throughout the
consolidation stages, except that during these stages of the analyses the pore water
pressures along the ground surface adjacent to the footing have been set and
maintained at the initial values existing prior to footing construction.
Small time steps were used to maintain undrained conditions during the loading
stages. For example, the initial loading to failure was achieved in 21 days.
Sufficient time was allowed during the consolidation stages for all excess pore
water pressures to fully dissipate. Typically this was achieved in 80 years and 30
years for the soft and stiff clay analyses respectively.
6.5.6.4 Results of the analyses

Detailed results from the investigation are given in Zdravkovic et al. (2001).
Consequently only the main conclusion concerning the effect of the pre-load will
be discussed here.
As an example of the results, the load-displacement curves for the 2m wide
smooth footing are shown in Figure 6.37a and 6.37b, for the soft and the stiff
(OCR=6) soils respectively. The curve labelled' initial' corresponds to the situation
where the footing is loaded rapidly from the initial conditions and therefore
produces the initial short term bearing capacity. These results are from the analysis
discussed in Section 6.5.5. The other curves are the results from the analyses in
which the footing has been pre-loaded to a percentage of this initial bearing
capacity and then allowed to consolidate with no change in load before being
subjected to rapid loading to failure. It can be seen that the higher the pre-load, the
higher the bearing capacity.
When the footing is initially loaded rapidly, excess pore water pressures are
established in the soil. The excess pore water pressures can be considered to consist
of two components. The first (positive) being due to an increase in mean total
stress, and the second (positive or negative) due to an increase in the deviatoric
(shear) stress. The first component will be compressive, with its magnitude
dependent on the footing load. The magnitude of the second component will also
depend on the footing load, but in addition it will depend on the type of clay and
in particular on its overconsolidation

ratio. For a clay with a low OCR the
second component will also be
compressive. However, as the OCR
increases, the magnitude of this
compressive pore water pressure
component will decrease, becoming
tensile at high OCRs. Consequently,
the magnitude of the excess pore 0.4 0.6 0.8 1
water pressure that is generated Displacement (m)
during the initial undrained loading

stages of each analysis will depend on Figure 6.37a: Load-displacement
the amount of load applied and on the curves for pre-loading on soft clay
initial OCR of the clay.
The changes in the undrained
180
shear strength of the clay during the 160
Stiff clay (QCR=6)
consolidation stages of the analyses 140

D D D
depend directly on the magnitude of 120
the excess pore water pressures and 100
the soil properties. For a given soil, 80 Initial

100% pre-load
the greater the compressive 60 80% pre-load
60% pre-load
40 40% pre-load
magnitude of the excess pore 20% pre-load
20
pressure, the greater the gain in
strength. If the excess pore water 0.2 0.3
Displacement (m)
pressures are negative, then the
undrained strength will reduce as
water will be sucked into the soil. Figure 6.37b: Load-displacement
Results from all the rough footing curves for pre-loading on stiff clay
analyses on stiff clay (in which the
undrained strength increased linearly
with depth) are plotted in Figure 6.38,
as undrained bearing capacity after
pre-loading normalised by the initial
undrained bearing capacity, against
amount of pre-load for different OCR
values. The analyses with OCRs less J*S 120
than 6 show an increasing bearing
capacity with the amount of pre-load. 40 50 60 70 80 90 100
For the OCR = 9 analyses the bearing Percentage of pre-loading (%)
capacity actually decreases when the

pre-load increases above 80%. A Figure 6.38: Undrained bearing
similar trend of reducing bearing capacity after pre-loading vs.
capacity is also predicted for the percentage of pre-load
OCR=25 analyses when the pre-load
exceeds 60%. In fact for this soil, when an analysis with a 100% pre-load was
attempted, the footing failed during the consolidation stage before it could be
loaded any further.
This behaviour can be explained by considering the excess pore water pressures
developed during pre-loading. As noted above, this consists of two components.
For soil with a high OCR the relative magnitude of these two components changes
with the amount of pre-load. For the OCR=9 and OCR=25 analyses the component
due to a change in total mean stress is compressive, whereas that due to deviatoric
shearing is tensile. As the amount of pre-load increases above 80% for the OCR=9
and 60% for the OCR=25 analyses, the tensile component increases in magnitude
relative to the compressive component, and consequently the total magnitude of
the excess pore water pressure drops and results in a smaller increase in undrained
strength during the consolidation stage. For the OCR=25 soil with a 100% pre-
load, the tensile deviatoric component of the excess pore water pressure exceeds
that component due to the increase in mean stress. This results in tensile excess
pore water pressures which on consolidation cause swelling, a drop in undrained
strength and therefore a reduction in bearing capacity. Further investigations
indicated that for pre-loads in excess of 93%, footings on a stiff OCR=25 soil failed
during the consolidation stages, before any further load could be added.
To show more clearly the effect of
OCR, the results presented in Figure
6.38 have been re-plotted in Figure
6.39, as normalised undrained bearing
capacity against OCR (plotted on a
logarithmic scale) for different
amounts of pre-load. For pre-load
values less than 50% the gain in
undrained bearing capacity first
reduces with OCR, but later recovers OCR
slowly as OCR is increased still

further. This may at first seem to Figure 6.39: Undrained bearing
contradict the arguments put forward capacity after pre-loading vs. OCR
above, which suggest that as the OCR
increases then the relative magnitude of the excess pore water pressures which are
available to dissipate is likely to reduce and therefore lead to smaller gains in
undrained bearing capacity. However, as the OCR increases, so does the SJav'
ratio, see Equation (6.9). Consequently, at the higher OCRs there will be a greater
increase in Su for a given change in ov'. This effect therefore acts in the opposite
sense to that associated with the magnitude of the excess pore water pressure, and
explains the behaviour observed in Figure 6.39. For a 100% pre-load the excess
pore pressure effect dominates and the gain in bearing capacity falls continuously
with increase in OCR, see Figure 6.39. In contrast, for the 80% pre-load the gain
in bearing capacity first reduces (excess pore water pressure effect dominating),
then increases slightly (SJov' effect dominating) before reducing again (excess
pore water pressure effect dominating) as the OCR increases.
It is also of interest to note that for the stiff clay analyses, in which the initial
undrained shear strength increases linearly with depth, the gain in bearing capacity
due to pre-load is independent of footing width. Figures 6.38 and 6.39 are therefore
applicable to a footing of any width and may be useful for design purposes.
6.5.6.5 Concluding remarks

The results show that increases of bearing capacity of up to 74% are possible for
pre-loaded foundations on normally consolidated clays, whilst for moderately
overconsolidated clays the maximum increase is only 15%. In the former case the
bearing capacity increases with the amount of pre-load applied. It will also depend
on the width of the footing in relation to the depth of any surface crust in the clay.
However, for the moderately overconsolidated clays the maximum bearing
capacity may occur at pre-loads less than 100%. For heavily overconsolidated
clays pre-loading may cause a reduction in undrained bearing capacity.
In practice most foundations will be designed with a factor of safety on load of
at least 2. This means that the maximum pre-load is therefore likely to be only
50%. The results of the present investigation indicate that such a pre-load will
cause at most an increase in undrained bearing capacity for subsequent loading of
35% for a normally consolidated soil, but only 12% for a soil with an OCR greater
than 4. Consequently, for the majority of real situations it is unlikely that pre-
loading will give rise to a substantial improvement in undrained bearing capacity.
6.5.7 Effect of anisotropic strength on bearing capacity

6.5.7.1 Introduction
In reality most soils behave in an anisotropic manner. This anisotropic behaviour
is often assumed to consist of two components, inherent and induced anisotropy.
Inherent anisotropy is that component which is controlled by the fabric of the soil
before it is loaded. Induced anisotropy is that component which arises due to
further loading.
In principle, if anisotropic behaviour is to be modelled, the constitutive model
should be able to simulate both components. However, if the inherent component
dominates behaviour, then a constitutive model that accounts for this component,
but not the induced component, might be acceptable. In such a situation a model
of the form described in Section 7.10.2 of Volume 1 might be appropriate. Because
such models do not represent induced anisotropy, they will not predict any effect
of rotating the direction of the principal stresses.
If both components of anisotropy are to be accounted for, then a model which
is formulated in terms of all six components of stress and strain, and not in terms
of invariants, is required. The MIT models described in Chapter 8 of Volume 1 are
such models. MIT-E3 has been used to investigate the influence of anisotropic soil
behaviour on the bearing capacity of both strip and circular rough, rigid surface
footings, subjected to inclined loading.
6.5.7.2 Soil behaviour

Zdravkovic (1996) performed an
investigation into the anisotropic
behaviour of a silt, using the Imperial
College large Hollow Cylinder
Apparatus (HCA). Results from one
of the series of tests (the M-series) are
shown in Figure 6.40. In these tests
all samples were initially Ko
consolidated to point A, with a Lode's
100 120 140 160
angle #=-30°, (i.e. triaxial p' (kPa)
compression) and the major principal
stress vertical, i.e. a=0°, where a is
Figure 6.40: Effective stress paths
the angle of inclination of the
from hollow cylinder tests on silt
direction of the major principal stress
to the vertical. At Point A the Lode's angle was changed to #=0°, maintaining the
major principal stress in the vertical direction, i.e. a=0°. The test labelled M0 was
then sheared undrained to failure, keeping #=0°, by increasing the major principal
stress, again at a=0°. The test labelled M90 was sheared undrained to failure,
keeping 0=0°, by reducing the vertical stress so that when failure occurred this was
the minor principal stress and hence the major principal stress was now horizontal,
i.e. a=90°. Test Ml5 was initially unloaded to point B15 with 6=0° and a=0°. The
major principal stress was then rotated to be inclined to the vertical by 15° (i.e
a=15°). This was done maintaining a constant J, p and 6 and with the soil
undrained. The stress path moved to point C15. The sample was then sheared
undrained keeping a=\5° and #=0°. A similar procedure was followed for tests
M30, M45 and M70, except that the direction of the major principal stress was
rotated to be at 30°, 45° and 70° to the vertical respectively. The details of testing
are presented in Zdravkovic (1996) and Zdravkovic and Jardine (2000).
The results given in Figure 6.40 therefore indicate the effect of changing the
direction of the major principal stress on the behaviour of the soil. Clearly this
behaviour is anisotropic, with both the drained angle of shearing resistance and
equivalent undrained strength reducing with an increase in the angle of rotation of
the direction of the major principal stress.
Parameters were obtained for the MIT-E3 constitutive model, see Table 6.4.
Because the model was to be used to simulate both isotropic and anisotropic soils,
the effect of the bounding surface plasticity below the state boundary surface had
to be switched off. If this was not done, it was impossible to make the model
behave in an isotropic manner. Consequently, no values for the parameters y and
h are given in Table 6.4. In any case, as the soils were normally consolidated, the
effect of the bounding surface plasticity would not be significant. The model was
then used to reproduce the test series given in Figure 6.40. The results are shown
in Figure 6.41. Overall, the analyses followed the same steps as imposed in the
experiments, except that all samples were unloaded by the same amount, i.e. A to
B. Comparison of Figures 6.40 and 6.41 indicates that the model is able to
reproduce the key aspects of the experimental results.
Table 6.4: MIT-E3 material properties

Parameter description MIT-E3 value
VlOO Specific volume for a Ko normally consolidated 1.688

sample at p'=100kPa
Ko"c Normally consolidated value of Ko 0.5
Vic Critical state angle of shearing resistance in 35°

triaxial compression
Viz Critical state angle of shearing resistance in 30°

triaxial extension
c Ratio of semi-axes of the bounding surface 1.0

ellipsoid
Parameter affecting rotation of bounding surface 100
X Slope of the VCL in v-ln// space 0.014
K Initial slope of the swelling line in v-\np' space 0.0018
Poisson's ratio 0.3
s, Parameter affecting strain softening 1.0
c Parameter affecting the hysteretic elasticity 1.0
n Parameter affecting the hysteretic elasticity 1.5
CO Parameter affecting the hysteretic elasticity 0.0
However, it is deficient in one important aspect: while the silt soil showed
dilatant behaviour after reaching phase transformation, MIT-E3 indicates ductile
failure. This is not a serious deficiency, as the analyses to be described here were
intended to model foundations on a soft clay, which shows similar behaviour to the
silt up to phase transformation, without dilating at large strains (Porovic (1995),
Leroueil(1977)).
The undrained triaxial compression strength profile for this soft clay is shown
in Figure 6.42. It varies linearly with depth, starting from a finite value at the soil
surface, giving Su/av'=0.36. This would be the profile normally adopted in any
analysis using an isotropic soil model. However, MIT-E3 allows the undrained
strength profile to change, depending on the direction of the major principal stress,
as shown in Figure 6.42.
NC
=0.5
\\\\\ e =
.W CT;V = 3o + ioz
= -30°
\\ \
-10 .
\\ \ Triaxial compression
150 \\ \\ V y
125
M 0 _ ^ ^ \\ \* \
\\ \ \\
\\ \ \ \
2 ioo M30 ^ ^ O \ -20 -
3
^ 75 M7
°m/ V \ \-' 11
3 50
25
M9O
%k\ \ \i -25 -
^\ \ \
x
0
0..<^\
20 40 60 80 100 120
p> (kPa)
140 \y
160 180 200 220
-30
0 20 40 60 80
Undrained strength, Su (kPa)
100 120
Figure 6.41: Hollow cylinder Figure 6.42: Anisotropic

effective stress paths simulated undrained strength profiles
with MIT model from MIT model
6.5.7.3 Behaviour of strip footings

Two series of strip footing analyses have been performed, in the first the properties
given in Table 6.4 were adopted, whereas in the second the values of cplc' and (pTE'
were reassigned to be <pTC/=35° and <pTE'=63°. This had the effect of making the
model isotropic and consequently, when the experimental results given in Figure
6.40 were simulated, each test was predicted to give the same result which was
very similar to the M0 test in Figure 6.41. Comparison of the analyses with these
different soil properties therefore allows the effects of anisotropy to be quantified.
For each set of material properties five analyses were performed, each having
a different inclination of the applied
load. This inclination was kept
constant in each analysis, and the y mm* y y
footing was loaded to failure under
load control. To achieve this, / Y E
/ T
r \ \
components of vertical (V) and 40 m
horizontal (//) loads were applied as

point loads, see Section 3.7.7 of
Volume 1, to the node at the centre of
the top of the footing. The magnitudes
of the load increments were reduced Figure 6.43: Finite element mesh
as failure was approached. The finite for analysis of strip footing
element mesh is shown in Figure

6.43. As the analyses were performed
in load control, the footing itself had
to be included in the analysis.
Load-displacement curves from a
typical analyses (H/V=0.3) are shown
in Figure 6.44. It can be seen that
failure is well defined, with both the
horizontal and vertical loads reaching > 0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
limiting values. Vertical (v) and horizontal (M) displacement (m)
The results of both sets of analysis Figure 6.44: Typical load-

are shown in Figure 6.45 in the form displacement curves for combined
of an interaction diagram showing loading analyses
values of the vertical (Fult) and
horizontal (Huh) loads at failure. The
results clearly show the adverse effect
of the anisotropic behaviour for all
IT
Strip footing
inclinations of loading. . III**

i t Tin
It should be noted that for the
HIV "03
analyses with horizontal loading only, IsotropicMIT model
Huh=2BSu, where Su is the value of the
undrained strength mobilised at the
soil surface below the footing. In the
present analysis this will be the
strength appropriate to plane strain Figure 6.45: Envelopes of ultimate
conditions and will therefore be load for strip footing
different to the values plotted in
Figure 6.42, which are for triaxial Jm
compression conditions.
m \/ \ <@
6.5.7.4 Beha viour of circular
footings
The above exercise was repeated for
V a
a circular footing. The Fourier series \ \
aided finite element approach was
adopted using the parallel symmetry 20 m
option, see Section 12.3.2 of Volume
1. By performing a small parametric <@
study it was found that only 5
harmonic coefficients were required
to obtain accurate results. The finite
element mesh is shown in Figure ///////
6.46. Vertical and horizontal loading Figure 6.46: Finite element mesh
were applied as line loads (in the for analysis of circular footing
circumferential direction) at the node

representing the top right hand corner JfL
Circular footing
of the footing, point A, in Figure

6.46. For the vertical load (V) only the
zero harmonic coefficient was Isotropic MIT model
specified as non zero (i.e.

A K=AF°+0...). To obtain a horizontal
load, both radial (Hr) and
circumferential (//#) line loads were
specified. For the radial load only the
first cosine harmonic was specified as figure 6.47: Envelopes of ultimate
—i load for circular footing
non zero (i.e. A//r=0+A Hr cos#+0...),
whereas for the circumferential load the first sine harmonic coefficient was set (i.e.
= i
A//#=0+A He sin#+0...). The latter was of opposite sign but equal magnitude to that
set for the radial load, see Figure 12.2 and Appendix XII.2 of Volume 1.
Again predictions using both sets of soil parameters were obtained. The results
are shown as an interaction diagram in Figure 6.47. The adverse effect of the
anisotropic behaviour is evident again. Comparison with Figure 6.45 indicates that
this effect is larger for the circular footing than for the strip footing.
6.6 Finite element analysis of shallow foundations

6.6.1 Introduction
Most of the comments given above for surface foundations also apply to shallow
foundations. However, in this case it is necessary to account for the interface
between the soil and the sides of the foundation, see Figure 6.48. It is not usual to
model the construction of the foundation itself, but to simply assume it to be
'wished in place'.
6.6.2 Effect of foundation depth on undrained bearing

capacity
To account for the depth of a foundation below the ground surface, depth factors
are usually introduced into the general bearing capacity equation in a similar way
to shape factors. For undrained bearing capacity the general equation becomes:
(6.11)
where dc is the depth factor and/?o is the total overburden stress at foundation level,
see Figure 6.49. Exact theoretical solutions are not available to account for
foundation depth and therefore the magnitude of the depth factor is often based on
semi-empirical correlations. One of the most popular correlations currently in use
is that proposed by Skempton (1951) and shown in Figure 6.50. This correlation
is largely based on the results of ""r "" 1 "" Shallowfoundation

model footing experiments. It shows
that dc increases with foundation
depth until reaching a normalised
depth, DIB, of 10, after which it
1 1
a i • B
remains constant. It is worth noting
that the maximum value of dc was
partly determined by the results from
cavity expansion analysis. The curve
shown in Figure 6.50 does not Figure 6.48: General finite element
distinguish between different footing modelling of shallow foundations
shapes or different footing roughness.
It is therefore commonly applied to
all footing types and in this respect B
refers to the half width of strip and
ground level
rectangular footings and the radius of
circular footings.
To investigate the effect of footing
depth a series of finite element
analyses have been performed with
foundation level -
DIB = 0, 2, 4, 6, 8 and 10. Both
smooth and rough strip and circular
footings have been analysed. A Figure 6.49: General loading
typical finite element mesh, for scheme for shallow foundations
D/B=S,- is shown in Figure 6.51. Due
to symmetry only half the geometry is
modelled. As with the undrained
analyses presented in Section 6.5.4.2,
the clay was modelled as an elastic
Tresca material with £=100MPa,
ju=0A9, Sf = 100kPa, K=\ and
ysat=20KN/m3. 2 4 6 8
The foundation itself was not Ratio of foundation depth to width, DIB
modelled and therefore displacement
boundary conditions were applied to Figure 6.50: Skempton's
that part of the mesh boundary relationship for depth factor
adjacent to the footing, see Figure
6.51. For a smooth footing zero horizontal displacements were applied along 'ab',
while increments of downward vertical displacement were applied along 'be'. For
a rough footing zero horizontal displacements and increments of vertical
displacement were applied along 'abc'.
If the mesh shown in Figure 6.51 is used to analyse a rough circular footing,
then the ultimate footing load is 16500kN. Assuming that the full undrained shear
strength of the soil is mobilised in shear along the side of the footing, and noting
that the diameter of the footing is 2m

and its depth below the ground surface T
is 8m, gives a shear force on the side
of the footing of 2x8x7rxl00=5026kN. \ /
-Js
Taking this value away from the total
footing load at failure gives the \ s/ / \
contribution from the base of the
footing as 11474kN. Comparing this a
with the ultimate load from a similar
analysis performed with the footing at \
the surface implies a depth factor, dc of 50 m

5.6. This value is considerably larger
than that given in Figure 6.50.
Closer inspection of the analysis
indicates the shear stress mobilised ///////
along the side of the footing is too
large. Summing the vertical nodal Figure 6.51: Finite element mesh
reactions along the side of the footing used for analysis of depth factor
gives a force of 12686kN. This is
considerably larger than the maximum
value of 5026kN calculated above,
assuming that the full undrained soil
strength is mobilised in shear along the
side of the footing.
At first, such a result might Interface elements
(zero thickness)
indicate some problem with the finite
element program. However closer
inspection of the analysis shows that at
all integration points in the finite
element mesh the stresses nowhere
exceed the Tresca failure condition.
The program is therefore functioning
correctly. The problem arises because
in the finite element method the
stresses are only sampled at integration
points. While they may obey the
failure condition at these . Detail of finite element
Figure 6 52:
points, the shape functions associated mesn adjacent t0 footing
with the elements used might imply a
variation of stresses over the element, which is not representative of those
occurring in the real problem. In the present case eight noded elements have been
used with 2x2 Gauss integration. This implies a linear variation of stresses,
whereas in the axi-symmetric circular footing problem the shear stresses are likely
to vary exponentially away from the side of the footing.
To overcome this problem thinner elements could be used to represent the soil
immediately adjacent to the side of the footing. Alternatively, interface elements
could be used, as shown in Figure 6.52. These are given normal and shear
stiffnesses compatible with that of the soil (i.e. A^=A^=105kN/m3) and a shear
strength of lOOkPa (i.e. the same as the soil). (Note that the vertical and horizontal
scales are different in Figure 6.52 and that, for clarity, the zero thickness interface
elements have been drawn with a finite thickness). Repeating the analysis now
gives an ultimate footing load of 9307kN. The contribution from the shear stresses
on the sides of the footing is 4945kN, which is now less than the maximum
possible value of 5026kN and therefore acceptable. (Note that for the maximum
possible value to be mobilised, the major principal stress would have to make
everywhere an angle of 45° to the vertical sides of the footing). Comparing the
contribution from the base of the footing with that from a similar footing located
at the ground surface gives a depth factor dc of 22.
The above problem is less severe
in the analyses of a rough strip g /
footing and clearly does not occur in 1
the analyses of both smooth strip and | 3 -
circular footings, as the shear stress £ 4 —
\
f//AN
i
along the side of the footing is zero. J 5 -

While the analyses of the rough ^
circular footing with interface Q 7 -
elements is an improvement over the 8
0 200 400
original analysis, it is still flawed as Horizontal stress on side of footing, kPa
tensile horizontal total stresses are
predicted between the sides of the
Figure 6.53: Horizontal total stress
footing and the adjacent soil. This is
on side of footing with tension
shown in Figure 6.53, which presents
allowed in interface elements
the distribution of total horizontal
stress down the side of the footing at 0
failure. These stresses are taken from
»p
1
the interface elements. If the soil is ". 2 - i

1 1 i I 11
unable to sustain such tensile stresses

then the analysis is unrealistic. In
reality the soil is likely to separate
5 —
from the footing, leaving a vertical
crack.
To overcome this problem, the
1 i >>
interface elements positioned down -200 0 200 400
the side of the footing can be given a Horizontal stress on side of footing, kPa
zero tensile strength. This means that
they can sustain compressive normal Figure 6.54: Horizontal total stress
stresses, but will open when the stress on side of footing with no tension
becomes zero, essentially modelling a allowed in interface elements
vertical crack. The distribution of rough circle

horizontal total stress down the side A
;mooth circle
of the footing at failure, from an
analysis performed with such
interface elements, is shown in Figure
6.54. Zero tensile stresses are
indicated to a depth of 2.5m below
the ground surface and consequently
separation occurs over this depth. The 2 4 6 8
Ratio of foundation depth to width, DIB
ultimate footing load from this
analysis was 8613kN and the depth
factor dc = 2.14. Figure 6.55: Depth factor from FE
Potential separation of soil from study compared to Skempton's
the side of the footing also occurs for relationship
smooth footings. Analyses for such
footings should therefore also include interface elements as shown in Figure 6.52.
To model the smooth nature of the interface between the soil and footing, these
elements are given very low shear stiffness and strength values (e.g. ideally zero
values should be given). To model separation of the soil from the footing these
elements are again given a bulk stiffness consistent with that of the soil and a zero
tensile stress capacity.
Interface elements were used in all the analyses for the parametric study on the
influence of foundation depth. The results of this study in terms of the variation of
depth factor, dc, against normalised foundation depth, DIB, are shown in Figure
6.55. Also shown on this figure for comparison is Skempton's curve. The results
for the strip footing are in reasonable agreement with Skempton's curve, while
those for the circular footing give higher values of dc. This agrees with results from
pile tests and the recent work of Randolph et al (2000), which indicate higher
values than implied by Skempton's curve.
6.6.3 Example: The leaning Tower of Pisa

In 1989 the civic tower of Pavia, in Italy, collapsed without warning, killing four
people. This prompted the Italian Minister of Public Buildings and Works to
appoint a Commission to advise on the stability of the Pisa Tower, which was
perceived to have a high risk of collapse. The Commission recommended closure
of the Tower to the general public and this was instituted at the beginning of 1990.
There was an immediate outcry by the Major and citizens of Pisa who, correctly,
foresaw the damage that the closure would inflict on the economy of Pisa, heavily
dependent on tourism as it is. In March 1990 the Prime Minister of Italy set up a
new Commission to develop and implement measures for stabilising the Tower.
One of the early decisions of the Commission was to develop a numerical model
of the Tower and the underlying ground that could be used to assess the
effectiveness of various possible remedial measures. The purpose of this section
is to describe this model, the work that went into its calibration and how it was
used to investigate both the temporary lead counterweight solution and some of the
various permanent solution options.
The only practical means of calibrating the model was to attempt to get it to
match the history of inclination of the Tower during and subsequent to its
construction. Therefore, the first part of this section is devoted to a brief
description of the history of the Tower. The differences between bearing capacity
failure and leaning instability are then discussed before the numerical model is
described and some of the results presented. Further details of the analyses can be
found in Burland and Potts (1994) and Potts and Burland (2000).
6.6.3.2 Details of the Tower and ground profile

Figure 6.56 shows a cross-section
through the Tower. It is nearly 60m high
and the foundations are 19.6m in th
7 c
diameter. The weight of the Tower is
14500t. At present the foundations are
inclined due south at 5.5° to the
horizontal. The average inclination of the
axis of the Tower is somewhat less, due
its slight curvature. The seventh cornice
overhangs the first cornice by about
4.1m.
Construction is in the form of a 1" cornice
hollow cylinder. The inner and outer Floor of
surfaces are faced with marble, and the instrument room
annulus between these facings is filled
with rubble and mortar within which
extensive voids have been found. A spiral
staircase winds up within the annulus.
Figure 6.56 clearly shows that this Figure 6.56: Schematic cross-
staircase forms a large opening on the section through Pisa Tower
south side, just above the level of the first
cornice where the cross section of the masonry reduces. The high stresses within
this region are a major cause of concern and could give rise to an instantaneous
buckling failure of the masonry without warning. In the summer of 1992 this
masonry was stabilised by applying lightly prestressed steel strands around the
Tower in the vicinity of the first cornice. At present, the masonry is being
consolidated by grouting and the temporary steel strands will soon be reduced in
number.
Figure 6.57 shows the ground profile underlying the Tower. It consists of three
distinct horizons, the properties of which are described in detail later in this chapter
and in AGI (1991). Horizon A is about 10m thick and primarily consists of
estuarine deposits laid down under tidal conditions. As a consequence, the soil
types consist of rather variable sandy

and clayey silts. At the bottom of
Horizon A is a 2m thick medium •a1
11 WT
Layer i
=1 -= S?a
Sandy and clayey silts ^^^ level
dense fine sand layer (the upper "~"^^ ——-
sand). Based on sample descriptions Upper clay (Pancone)
CQ
and piezocone tests the material to the ft Intermediate clay
Intermediate sand
south of the Tower appears to be •3
Lower clay
more silty and clayey than to the
north and the sand layer is locally
Layer C
Lower sand
thinner.
Horizon B consists of marine clay
which extends to a depth of about
40m. It is subdivided into four distinct Figure 6.57: Soil profile beneath
layers. The upper layer is a soft Pisa Tower
sensitive clay known as the Pancone. It is underlain by a layer of stiffer clay (the
intermediate clay), which in turn overlies a sand layer (the intermediate sand). The
bottom of Horizon B is a normally consolidated clay known as the lower clay.
Horizon B is laterally very uniform in the vicinity of the Tower.
Horizon C is a dense sand which extends to a considerable depth (the lower
sand). The water table in Horizon A is between lm and 2m below the ground
surface. Pumping from the lower sand has resulted in downward seepage from
Horizon A, with a vertical pore water pressure distribution through Horizon B
slightly below hydrostatic.
The many borings beneath and around the Tower show that the surface of the
Pancone clay is dished beneath the Tower, from which it can be deduced that the
average settlement is approximately 3m.
6.6.3.3 History of construction

The Tower is a campanile for the Cathedral, construction of which began in the
latter half of the 11th century. Work on the Tower began on 9th August 1173 by the
modern calendar. By about 1178 construction had progressed to about one quarter
of the way up the fourth storey when work stopped. The reason for the stoppage
is not known, but had it continued much further the foundations would have
experienced an undrained bearing capacity failure. The work recommenced in
about 1272, after a pause of nearly 100 years, by which time the strength of the
ground had increased due to consolidation under the weight of the Tower. By
about 1278 construction had reached the 7th cornice, when work again stopped due
to military action. Once again there can be no doubt that, had work continued, the
Tower would have fallen over. In about 1360 work on the bell chamber was
commenced and was completed in about 1370 - nearly 200 years after
commencement of the work.
It is known that the Tower must have been tilting to the south when work on
the bell chamber began, as it is noticeably more vertical then the remainder of the
Tower. Indeed, on the north side there are four steps from the seventh cornice up
to the floor of the bell chamber, while on the south side there are six steps. Another
important detail of the history of the Tower is that in 1838 a walk-way was
excavated around the foundations. This is known as the catino and its purpose was
to expose the column plinths and foundation steps for all to see, as was originally
intended. This activity resulted in an inrush of water on the south side, since here
the excavation is below the water table, and there is evidence to suggest that the
inclination of the Tower increased by as much as a half of a degree as a result.
6.6.3.4 History of tilting

As mentioned in the Introduction, the only possible means of calibrating a model
of the Tower is to attempt to simulate the history of tilting of the Tower during and
subsequent to its construction. Hence it was necessary to learn as much as possible
about the history of the tilt of the Tower. The only reliable clues on the history of
tilt lie in the adjustments made to the masonry layers during construction and in the
shape of the axis of the Tower.
Based on the measured thickness
of each masonry layer and a 8
hypothesis on the manner in which 15 r r
the masons corrected for the
progressive lean of the Tower,
Burland (see Burland and Potts
(1994)) deduced the history of
inclination of the foundations of the 1 1178 to 1272
Tower shown in Figure 6.58. In this
figure the weight of the Tower is
plotted against the deduced •a Loading
inclination. During the first phase of "o 5 Consolidation
construction to just above the third Plane strain
cornice (1173 to 1178), the Tower • 3D analysis
o Historical
inclined slightly to the north. The
northward inclination increased
slightly during the rest period of 0 1 2 3 4 5 6
nearly 100 years to about 0.2°. When Inclination of foundations (deg)
construction recommenced in about
1272, the Tower began to move
Figure 6.58: Predicted and historical
towards the south and accelerated
shortly before construction reached inclination of the Tower during and
the seventh cornice in about 1278 after construction
when work again ceased, at which stage the inclination was about 0.6° towards the
south. During the next 90 years the inclination increased to about 1.6°. After the
completion of the bell tower in about 1370, the inclination of the Tower increased
significantly. In 1817, when Cressy and Taylor made the first recorded
measurement with a plumb line, the inclination of the Tower was about 4.9°. The
excavation of the catino in 1834 appears to have caused an increase in inclination
of approximately 0.5° and the present day inclination of the foundations is about
5.5°. It can be seen from Figure 6.58 that significant inclination of the Tower only
began once the height exceeded the sixth cornice. If the inclination had been due
to much more compressible ground beneath one side than the other, it would have
developed much earlier. Therefore, another explanation for the rapid onset of
inclination is required and will be discussed later. It is the history of inclination
depicted in Figure 6.58 which was used to calibrate the numerical models
described later in this section.
For most of this century the inclination of the Tower has been increasing. These
changes in inclination are extremely small compared with those that occurred
during and immediately following construction. The rate of inclination of the
Tower in 1990 was about 6 seconds per annum. The cause of the continuing
movement is believed to be due to fluctuations of the water table in Horizon A. No
attempt has been made to model these small movements.
6.6.3.5 The motion of the Tower foundations

Previously, studies have concentrated
on the changes of inclination of the
Tower. Little attention has been
devoted to the complete motion of the
foundations relative to the Centre of
surrounding ground. The theodolite rotation
and precision levelling measurements
made in the last century help to
clarify this. These observations can be
used to define the rigid-body motion
of the Tower during steady-state
rotation, as shown in Figure 6.59 (see
Burland and Potts (1994)).
It can be seen that the Tower is
rotating about a point approximately
Figure 6.59: Motion of the Tower
located level with point V, and
vertically above the centre of the foundation. The direction of motion of points FN
and F s are shown by vectors and it is clear that the foundations are moving
northwards with FN rising and F s sinking. It can therefore be concluded that the
seat of the continuing long term tilting of the Tower lies in Horizon A and not
within the underlying Pancone clay, as has widely been assumed in the past.
6.6.3.6 Stability of tall towers

Before considering the numerical analysis of the Pisa Tower, it is illuminating to
consider the possible mechanisms of failure associated with the stability of the
foundations of a tall tower. There are two possible mechanisms that could account
for failure of such a tower: (i) bearing capacity failure due to insufficient soil
strength, and (ii) leaning instability due to insufficient soil stiffness. Bearing
capacity failure is the more common type of instability and the one covered in most
text books and codes of practice. All the analyses presented so far in this chapter
showed this type of failure. Leaning instability is not so common and is only
relevant to tall structures. It occurs at a critical inclination when the overturning
moment, generated by a small increase in inclination, is equal to or greater than,
the resisting moment of the foundations generated by the same rotation. In all but
the simplest of cases it is difficult, probably impossible, to analyse without using
numerical analysis.
These two alternative failure
mechanisms are best demonstrated by
a simple example. Figure 6.60 shows
a simple tower resting on a uniform
deposit of undrained clay. The clay is Initial tilt
modelled as a linear elastic Tresca oftower = 0.5°
material, with an undrained strength
SM=80kPa. The dimensions of the
tower are similar to those of the Pisa
Tower. To trigger a rotation failure
some initial defect (imperfection) 20 m
must be present. In this example the Undraiaed clay
tower was given an initial tilt of 0.5°
(Elasto-plastic)
(i.e. the initial geometry of the tower
had a tilt). The self weight of the Tiesca model - &,« 80 JcPa
tower was then increased gradually in
a plane strain large displacement
finite element analysis. Figure 6.60: Geometry of simple
tower
Three analyses were performed,
each with a different value of shear
4"
stiffness, G, of the soil, and the results GISU = 10 GISU = 100
are presented in Figure 6.61. Here the 3-
JJ
increase in rotation of the tower
above the initial 0.5° imperfection is 1 2"
plotted against weight of the tower, GIS = 1000
1-
for analyses with G/Su values of 10,
100, 1000. Real soils are likely to o-
___ J
have properties such that they lie 50 100 150
between the two extreme values. It Weight of Tower (MN)
should be noted that in all cases the

strength of the soil Su was 80kPa. The Figure 6.61: Rotation of simple
results show that failure occurs very tower for varying soil stiffness
abruptly, with little warning and that
the weight of the tower at failure is dependent on the shear stiffness of the soil. The
weight at failure for the analysis with the softer soil, G/Su=\0, is about half of that
for the analysis with the stiffest soil, G/Su=l000.
It is of interest to examine the

analyses with the two extreme values
of GISU in more detail. In particular, it
is instructive to consider what is
happening in the soil at failure. Figure
6.62 shows vectors of incremental
displacements for the soft soil. The
vectors represent the magnitude and
direction of the displacements from
the last increment of the analysis.
This figure shows that the movements
are located in a zone below the Figure 6.62: Pattern of movement
foundation and indicate a rotational at failure for soil with a low
type of failure. At first sight this looks stiffness
like a plastic type collapse
mechanism. However, examination of
the zone in which the soil has gone
plastic (also shown on Figure 6.62),
indicates that it is very small and not
1.5 MN
M
consistent with a plastic failure
mechanism. Consequently, this figure
indicates a mechanism of failure
consistent with a leaning instability.
• • • - • ' . ' / / / * '
m
In view of the temporary G/Su = 10
counterweight scheme, which
involved adding lead weights to the
north side of the Pisa Tower and Figure 6.63: Effect of
which will be discussed in more detail counterweight for soil with a low
later in this chapter, it is of interest to stiffness
examine the response of the simple
tower in the above example if, at the
point of collapse, weight is added to
the higher side of the foundation. The
effect of a 1.5MN/m load is shown in
Figure 6.63. Again vectors of
incremental displacement are shown.
These indicate the nature of the
movements due only to this additional
load. It is noted that under this load
the sense of movement is reversed
and the tower rotates back and
collapse is arrested. Figure 6.64: Pattern of movement
Considering the results from the at failure for soil with a high
analysis performed with the stiffer stiffness
soil, vectors of incremental 1 5MN

displacement just before collapse are
shown in Figure 6.64. The mechanism
of failure indicated by these vectors is . . . :: *» x* xx xxxxfc
[ / / / / „
TO
very different to the one shown in
Figure 6.62 for the softer soil. Instead
» » * x
ft:
of the soil rotating as a block with the
foundation, the vectors indicate a
more traditional bearing capacity type G/Su = 100C
mechanism, with the soil being
pushed outwards on both sides. The
plastic zone, also indicated on Figure Figure 6.65: Effect of
6.64, is very large and therefore the counterweight for soil with a high
results clearly indicate a plastic stiffness
bearing capacity type mechanism of
failure.
If, as before, load is applied to the higher side of the foundation at the point of
collapse, the vectors of incremental displacement given in Figure 6.65 are
obtained. These show that, in contrast to the softer soil analysis, the tower
continues to increase its inclination. In fact, it was not possible to obtain a
converged solution when the weight was added. The addition of the load initiates
collapse even though the load acts to reduce the overturning moment.
As well as demonstrating the difference between the two types of instability,
these analysis also indicate that a counterweight type scheme will only be
beneficial to the Pisa Tower if it is suffering a predominately leaning instability.
To complicate matters further, real soils are likely to have stiffness values between
the two extremes considered above and therefore both mechanisms of behaviour
are likely to be active to some degree. In this respect the motion of the Pisa Tower
discussed in Section 6.6.3.5 and shown in Figure 6.59 is more consistent with the
movements shown in Figure 6.63 than those shown in Figure 6.65, indicating that
the Tower is probably suffering predominately from a leaning instability.
6.6.3.7 Soil properties

The constitutive model chosen for the clay strata was a form of modified Cam clay
(Roscoe and Burland (1968)), in which the shapes of the yield and plastic potential
surfaces in the deviatoric plane are given by a Mohr-Coulomb hexagon and a circle
respectively (see Section 7.9 of Volumel). For the sand layers a Mohr-Coulomb
model was used. Fully coupled consolidation for all the soil layers was
incorporated into the analyses. In order to implement these models the following
soil parameters are required:
saturated unit weight,
critical state angle of shearing resistance,
C compression index (note Cc = 2.30251),
Cs - swelling index - taken as 0.1 Cc (note C, = 2.3025/c),

G/po' - elastic shear modulus,
e{ - void ratio on the VCL when/?'=lkPa (note e, = v , - 1 . 0 ) ,
k - permeability,
K() - coefficient of earth pressures at rest,
OCR - overconsolidation ratio,
v - angle of dilation (sands).
Given the above parameters, the value of the undrained strength Su can be
determined from Equation (6.9). In Horizon A (see below) Su turned out to be a
more reliable parameter than OCR and was therefore specified in its place.
The soil profile at Pisa was characterised in detail by the Polvani Commission
(Ministero dei Lavori Pubblici 1971). Horizons A and B were divided into a
number of sub-layers, the descriptions of which are as follows:
Horizon A:
MG Top soil and made ground,
Al Loose to very loose yellow sandy silt to clayey silt without
stratification,
A2 Uniform grey sand with interbedded clay layers, broken fossils - Upper
sand,
Horizon B:
B1 Highly plastic grey clay with fossils,
B2 Medium plastic grey clay with fossils,
B3 Highly plastic grey clay with fossils,
B4 Dark grey organic clay,
B5 Blue grey to yellow silty clay with calcareous nodules,
B6 Grey, sometimes yellow, sand and silty sand - Intermediate sand,
B7 Medium to highly plastic clay, with fossils and thin sand layers in the
upper part,
B8 Grey clay with frequent thin sand lenses,
B9 Blue grey silty clay with yellow zones; calcareous nodules; some dark
organic clay at centre,
BIO Grey clay with yellow zones; fossils in the lower part.
Laterally Horizon B is very uniform. However, there is much evidence to show that
in Horizon A, layer A l changes from predominantly silty sands and sandy silts
north of the Tower to clayey silts south of the Tower. There is also some evidence
from piezocones that the Upper sand layer A2 thins just south of the Tower. A
careful study of the detailed sample descriptions given in the Polvani Report
suggests that beneath the Tower, in Horizon A, there exists a lense of clayey silt
which thins from south to north.
It will become evident that it is the compressibility of the underlying soils
which has played the dominant role in the historical behaviour of the Tower (i.e.
leaning instability), so that emphasis is placed in this section on these properties.
Two major programmes of thin wall sampling and testing have been carried out,
one in 1971 (at the instigation of the Polvani Commission) and another in 1986.
The results of these studies have been summarised by Calabresi et al. (1993) and
Lancellotta and Pepe (1990) respectively. Figures 6.66a, b and c show the
experimental values of C c \ Cc and OCR respectively, where C* is the
compressibility of the reconstituted material - defined by Burland (1990) as the
intrinsic compressibility. The values of C* were derived from the correlation with
the water content at the liquid limit, wL, established by Burland. Also shown on
Figure 6.66 are the sub-layers established by the Polvani Commission and
described above. It can be seen from Figure 6.66a that the values of C* are
reasonably well defined for each sub-layer and the average values are shown by the
vertical lines. It is of interest to note that sub-layer B7 is made up of two distinct
soil types and should perhaps be considered as two separate sub-layers.
cc OCR
0.5 1.0 0 0.5 1.0 0 1 2
I A2 1
o^ Bl
^ \ B3 ~-K_. °-222 o. l_
> -20 "vr(f"
B5"
"vr
B6 1
B7
* ° ^£° •"'""••4
-30 £
•
°|
J o B9
BIO
•T T - 0 O o
a)
Tb) c)
-40 -
Figure 6.66: Experimental values of Cc\ Cc and OCR;

values used in the analysis are shown as full lines
It can be seen from Figure 6.66b that the experimental values of Cc, determined
from high quality samples of the natural clay, show considerable scatter and it is
not easy to decide on suitable representative values for analysis. The use of the
intrinsic compressibility C* has been particularly useful in this respect (Burland
and Potts (1994)). It is well known that measured values of Cc are particularly
sensitive to sample disturbance and the scatter in Figure 6.66b is a reflection of
this. Nash et al. (1992) presented the results of oedometer tests on a sensitive
marine clay from the Bothkennar test bed site and showed that the values of Cc
obtained from high quality block samples were significantly higher than for other
sampling methods including thin wall sampling. Nash's results showed that the
measured values of Cc for the best samples were between 1.9 and 2.3 times larger
than C*. In Figure 6.66b the vertical lines were obtained by multiplying the
average values of C* by a factor depending on the plasticity of the material. For
the Pancone clay (Bl to B3) the factor was 2, for the Lower clay (B7 to B10) the
factor was 1.5 and for Horizon A and the Intermediate clay (B4 and B5) the factor
was unity. It is these values that have been used in the finite element analyses and
it can be seen that they tend to lie at, or a little beyond, the upper limits of the
experimental values, as is to be expected.
Figure 6.66c shows experimental values of OCR. The vertical lines are the
values used in the finite element analyses. Values were chosen near to the upper
limit of the experimental values, as the effect of sample disturbance is to reduce the
value somewhat.
Table 6.5 gives the values of all the soil parameters that were used in the
analyses. The mean values of cpj for each sub-layer were obtained from undrained
triaxial tests. KIX was taken as 0.1 for all the soil layers. The values of permeability
were derived from the oedometer tests, taking values in the upper quartile of the
range of results for each layer. The values of Glpo' were chosen on the basis of
experience with small strain testing of a wide variety of materials. The choice of
values was not important for modelling the history of inclination of the Tower, but
proved to be more crucial when predicting its response to small perturbations of
load in its present condition. The water table was assumed to be lm below ground
level and in hydrostatic equilibrium - a condition that must have existed at the time
of construction.
Table ( 5.5: Soil parameters used in the ianalysis

Layer Elev. (Pcs' v Ko OCR X v, Glpo k
(m) (%) kN/m3 (m/s)
MG +3.0 _ 18.0 34° 17° 0.44 1.0 _ _ 500 io- 5

Al" +0.0 38 19.1 34° - 1.0 vary 0.06 2.23 400 io- 5
Al' vary 45 19.1 34° - 1.0 vary 0.11 2.23 400 10" 9
A2 -5.2 - 18.2 34° 17° 0.44 1.0 - - 1000 10 s
Bl -7.4 75 17.3 26° 0.56 1.8 0.43 4.53 300 5xlO-1()

B2 -10.9 55 17.8 26° - 0.56 1.5 0.31 3.93 300 5xlO 1 ( )
B3 -12.9 85 16.7 26° - 0.56 1.4 0.47 5.10 300 5xl0- 10
B4 -17.8 68 20.0 28° _ 0.56 2.0 0.13 2.48 500 2xlO- 10

B5 -19.0 45 20.0 28° - 0.74 2.0 0.13 2.48 500 2xlO- 10
B6 -22 - 19.1 34° 17° 0.44 1.0 - - 1000 37168
B7(a) -24.4 35 19.6 27° 0.55 1.4 0.15 2.67 300 5xl0' 1()
B7(b) -26.0 80 17.8 27° - 0.55 1.4 0.32 4.17 300 5xl0" 1()
B8 -29.0 53 19.1 25° - 0.58 1.4 0.22 3.11 300 3xl0- 1()
B9 -30.4 55 19.1 25° - 0.58 1.4 0.22 3.11 300 3xlO- 10
BIO -34.4 50 19.1 25° - 0.58 1.4 0.22 3.11 300 3xlO 1 ( )
Cl -37.0 - - - - - - - - - perm
6.6.3.8 Finite element analysis

Geometry
It must be emphasised that a prime objective of the analysis was to develop an
understanding of the mechanisms controlling the behaviour of the Tower. It was
felt that until these had been clarified, it would be unhelpful to attempt highly
sophisticated and time consuming three dimensional analysis. Accordingly, a plane
strain approach was used initially, recognising that the interpretation of the results
would require some care. Some of the earlier analyses were carried out using a
large strain formulation, but it was found that this introduced time consuming
complications in relation to the excavation of the catino and also relative ground
water level changes (i.e. settlements exceeded the depth of the original ground
water table). The results did not differ significantly from traditional infinitesimal
strain analysis, which was therefore adopted for the work described here. Later a
limited number of three dimensional analyses were carried out using the Fourier
series aided finite element method (see Chapter 12 of Volume 1). The results of
both the plane strain and three dimensional analyses are reported here.
The layers of the finite element mesh matched the soil sub-layering discussed
above. The lower sand was assumed to be rigid but permeable. The full mesh, for
the plane strain analyses, is shown in Figure 6.67 and extends laterally 100m either
side of the axis of the Tower. In Horizon B the soil was assumed to be laterally
homogeneous. However, a tapered layer of slightly more compressible material
was incorporated into the mesh for layer A1 as shown in Figure 6.68, which shows
a detail of the mesh in the vicinity of the Tower. The shaded elements beneath and
to the south have the properties listed in Table 6.5 for layer AT, with Cc.=0.25 and
fc=10"9m/s. The remaining elements in layer A1 have the properties listed for layer
Al", with Cc=0.15 and A=10"5m/s. In applied mechanics terms the insertion of this
slightly more compressible tapered layer beneath the south side may be considered
to be an 'imperfection'.
m
4- I 1m i
=4=
Figure 6.67: Finite element mesh

For the three dimensional Fourier series aided analysis the geometry is assumed
to be axi-symmetric. However, the spatial distribution of soil properties and
loading can be three dimensional. A mesh similar to that shown in Figure 6.67, but
only considering the geometry to the right of the centre line of the Tower, was
used. However, no tapered layer was incorporated to provide an 'imperfection'.
Instead, the soil properties were assumed to vary linearly beneath the Tower in
sub-layer A1. To the north of the Tower the properties were as listed for layer A 1"
in Table 6.5, to the south as listed for w M=Wh cgsin0/c

layer AT, and beneath the Tower
they varied linearly from north to North j/;;,,,'-
South
south from those listed for layer A l " —1

Al
to those listed for layer AT. r::: = a* A2
Bl
Note that, in modelling the B2
foundation of the Tower, the central
hole was neglected for the plane strain
J ±_
j—^ ^ ^J— B3
B4
analysis, since it only represents a

small proportion of the total area of Figure 6.68: Finite element mesh
the foundation. close to Tower foundation
Foundation overturning moment

If any inclination 8 of the Tower takes place, the centre of gravity moves
horizontally and its weight JF generates an overturning moment M. If the height of
the centre of gravity above foundation level is hcg, then:
M=Whcg(sin0)Ic (6.12)
Ic is a correction factor which takes account of the ratio between the second
moments of area of a rectangular and a circular foundation. For a rectangular
foundation of width 19.6m and the same area as the foundation of the Pisa Tower,
the value of Ic for rotation about the centre is 1.266. For the three dimensional
analyses in which the circular shape of the foundation is modelled Ic should
theoretically be unity. Equation (6.12) was incorporated into the analysis such that
any inclination of the foundations during or subsequent to construction
automatically resulted in the application of an appropriate overturning moment to
the foundations.
Analyses
All analyses involved coupled consolidation. The calibration analyses were carried
out in a series of time increments in which loads were applied to the foundation to
simulate the construction history of the Tower together with the rest periods, as
summarised Section 6.6.3.4. During a construction period it was assumed that the
load was applied at a uniform rate. The excavation of the catino was also simulated
in the plane strain analysis.
The only factor that was adjusted to calibrate the model was the factor Ic in
Equation (6.12). For the first run, the value of Ic was set equal to unity. At the end
of the run the final inclination of the Tower was found to be less than the present
value of 5.5°. A number of runs were carried out with successive adjustments being
made to the value of Ic until good agreement was obtained between the actual and
predicted value of the final inclination. It was found that, for the plane strain
analysis with a value of Ic = 1.27, the final calculated inclination of the Tower was
5.44°. Any further increase in Ic resulted in instability of the Tower. It is therefore
clear from this analysis that the Tower must have been very close to falling over.
The final value of Ic is very close to the theoretical value for rotation about the
centroid, but this is probably coincidental.
For the three dimensional analysis a final inclination of 5.48° was obtained with
a calibration factor Ic of 1.7. As noted above, theoretically Ic should be unity for
these analyses. However, in these three dimensional analyses the excavation of the
catino was not simulated.
6.6.3.9 Simulation of the history of inclination

Figure 6.69 shows a plot, from the ^o nn im 183g
plane strain analysis, of the predicted 1 | | J / - — • —544<
changes in inclination and settlement
of the Tower with time since the start
of construction in 1173. It is
important to appreciate that the only
1170 1370 1570 1770 1970
point that has been pre-determined on
this plot is the final inclination of
2 -
about 5.5°. All other displacements Predictions
were generated by the analysis. It can

4 -1
be seen that completion of the Tower
up to the seventh cornice (-1278) _. ^ ^ „ , . ,. ,
u . v 11- r 4.- T,.- Figure. 6.69: Relationship between
results in quite small inclinations. It is . ,. . ,
time
only when the bell chamber is added ' inclination and settlement for
p/a e
in 1360 that the inclinations increase " strai" s'mu'atlon_ °f the
, .• ii AI r J ui history of the Pisa Tower
7
dramatically. Also of considerable
interest is the fact that simulation of the excavation of the catino results in a
significant increase in inclination of about 3A°. The final settlement of the
foundations is seen to be about 3.8m, which is larger than the value deduced from
the depression in the surface of the Pancone clay (Figure 6.57). The reason for this
is thought to be due to the plane strain analysis. The depth of influence of a circular
foundation would be less and would result in a smaller settlement.
The results of the analysis are also plotted on Figure 6.58 as a graph of load
against inclination and can be compared with the deduced history. Again it must
be emphasised that it is only the present inclination that has been fitted. The
agreement between the simulated and historical behaviour is remarkable and gives
considerable confidence in the reliability of the computer model. A striking
difference is that the model does not predict the initial northerly inclination of the
Tower. This is not felt to be of importance for the intended application of the
model. It was found that, during the early stages of loading, the model did show
a small inclination to the north. This was due to the fact that consolidation of the
thin northern end of the tapered layer of compressible soil took place more rapidly
than the thicker southern end. It should be possible to devise a soil profile in
Horizon A that more accurately simulates the early history of inclination of the
Tower, but this was outside the scope of the project.
Also shown on Figure 6.58 are the results from the three dimensional analysis.
It can be seen that the agreement

between the plane strain and three
dimensional analysis is very
satisfactory. The final settlement for
the three dimensional model is 3.1m,
which is much closer to the deduced
value than the plane strain model.
Figure 6.70 shows zones of fully
Fully mobilised
mobilised strength at the end of the strength
plane strain analysis following
complete pore water pressure
dissipation after excavation of the Figure 6.70: Zones of fully
catino. As mentioned previously, at mobilised strength at the end of the
this stage the Tower is in a state of analysis
unstable equilibrium. Surprisingly,
there are no zones of contained failure within the Pancone clay, but there are
extensive zones within Horizon A. The lowest zones are in the upper sand layer
and result from the lateral extension of this layer. There is a large zone beneath and
outside the southern edge of the foundation and a smaller zone underneath the
northern side. It is evident from this figure that the impending instability of the
Tower foundations is not due to the onset of a bearing capacity failure within the
Pancone clay. The cause of the instability can be attributed to the high
compressibility of the clay which results in leaning instability.
Figure 6.71 shows the distribution
of vertical effective contact pressure
acting on the foundation,
corresponding to complete pore water
pressure dissipation after excavation
of the catino in the plane strain
analysis. The distribution is far from
linear. Beneath the southern edge of C.L.
the foundation the effective contact Before installation of counterweight
pressure is about 850kPa. At the After installation of counterweight
northern edge the effective contact

pressure is zero over approximately Figure 6.71: Distribution of
the first metre, but further south it effective foundation contact stress
increases rapidly.
In summary, the finite element models described above give remarkable
agreement with the deduced historical behaviour of the Tower. It is important to
emphasise that the history of foundation inclinations and overturning moments
were self generated and were not imposed externally in a pre-determined way. The
only parameter that was used to calibrate the model was the present inclination of
the Tower. The analyses have demonstrated that the lean of the Tower results from
the phenomenon of leaning instability due to the high compressibility of the
Pancone clay. The role of the tapered layer of slightly increased compressibility
beneath the south side of the foundations in the plane strain analysis and the linear
variation of properties assigned to sub-layer A1 in the three dimensional analysis,
is to act as an 'imperfection'. Its principal effect is to determine the direction of
lean rather than its magnitude. The model provides important insights into the basic
mechanisms of behaviour and has proved valuable in assessing the effectiveness
of various proposed stabilisation measures. Its role in evaluating the effectiveness
of the temporary counterweight solution will now be described.
6.6.3.10 Temporary counterweight

Fully aware that the selection, design and implementation of permanent
stabilisation measures for both the foundations and the masonry of the Tower
would take a long time, the Commission took an early resolution to implement
short-term temporary and fully reversible measures to increase slightly the stability
of these elements (Burland et al. (1993)). The use of lightly prestressed steel
tendons to stabilise the masonry was described earlier.
The similarity between the motion of the
Tower depicted in Figure 6.59 and the
movements associated with leaning instability a)
shown in Figure 6.62 suggested a possible
temporary means of increasing the stability of
the foundation. The observation that the northern
side of the foundation had been steadily rising
led to the suggestion that application of load to
the foundation masonry on the north side could
be beneficial in reducing the overturning J
moment. Clearly, such a solution would not have b)
been considered if it had not been recognised yield surface
that leaning instability, rather than bearing after ageing
capacity failure, was controlling the behaviour of
the Tower. Before implementing such a solution initial
it was obviously essential that a detailed analysis yield surface
should be carried out. The purpose of such an Po' Poy' P'
analysis was two-fold: firstly to ensure that the
proposal was safe and did not lead to any Fjgure 6 72: a) increase in
undesirable effects and secondly to provide a y/e/d stress due t0 agjng;
class A prediction which could be used to assess ^j genera/ shift in the yield
the observed response of the Tower as the load surface due to aqinq
was being applied.
Both the plane strain and three dimensional analyses were therefore extended
to simulate the addition of a counterweight to the north side of the Tower. As
explained by Potts and Burland (1994), before simulating the addition of the
counterweight it was necessary to account for ageing of the clay layers over the
past 100 years or so, since the end of construction. There is much evidence to show
that ageing significantly increases the stiffness of clays. For one dimensional
compression Leonards and Ramiah (1959) showed that the process of ageing
results in an increase in the yield stress ovy\ as illustrated in Figure 6.72a. More
generally a shift in the yield surface takes place as illustrated in Figure 6.72b. The
parameters po' and/?o>/ are the intersections with the/?7 axis of the yield surfaces
of a 'young' and the 'aged' clay respectively. The ratio poy'lpo' is a measure of the
degree of ageing and is defined as the yield stress ratio. The effects of ageing can
be introduced into the finite element analysis by increasing the current value of/?,/
for each integration point to give a prescribed value of the yield stress ratio.
In the laboratory, significant ageing effects have been observed over periods
of a few days. Leonards and Ramiah found that the yield stress ratio for a
reconstituted clay that was allowed to age for about 90 days was as high as 1.3. In
the present analysis the effect of introducing various values of yield stress ratio was
studied. The results presented here are for a value of 1.05, which was felt to be
conservative. It should be noted that ageing of the Pancone clay increases the
stability of the Tower foundations significantly.
The full line in Figure 6.73 shows
the predicted response of the Tower A0 (deg)
North
due the application of a counterweight 0.08 -
to the foundation masonry, at an
eccentricity of 6.4m to the north, after
allowing ageing of the Pancone and
Lower clays to give a yield stress "008
Load (tonnes)
ratio of 1.05. No ageing of Horizon A
was assumed. The soil parameters are
those given in Table 6.5. At the 5 0 -
design load of about 690t the
inclination of the Tower is predicted 100 J
to reduce by about 27.5 seconds of Settlement (mm)
arc with a settlement of 2.4mm. More
importantly, the overturning moment
Figure 6.73: Predicted response of
is reduced by about 14%. In Figure
the Tower due to application of
6.71 the broken line represents the
counterweight
predicted effective contact pressure
beneath the foundation after application of the counterweight.
Also shown in Figure 6.73 are the effects of increasing the load above the
planned level and the results are of considerable interest. It can be seen from the
full lines that, as the load is increased, the rate of increase of inclination to the
north reduces, becoming zero at about 1400t. With further increase in load the
movement reverses and the Tower begins to move towards the south. The
settlement rate also begins to increase once the load exceeds 1400t. Figure 6.73
also shows the results of increasing the eccentricity of the counterweight. At an
eccentricity of 9.4m (dotted lines) the Tower continues to rotate northwards as the
counterweight is increased. At an intermediate eccentricity of 7.8m (broken lines)
a curious response is obtained in which the Tower first moves northwards, then it
reverses and then reverses again. It can be seen that the settlement response is
similar for each eccentricity. It would be wise to ensure that (irrespective of the
eccentricity) a significant rate of increase of settlement is avoided as this implies
the onset of yield.
6.6.3.11 Observed behaviour during application of the

counterweight
A description of the counterweight
scheme is given by Burland et al. 800
(1993). It consists of a temporary : 600

prestressed concrete ring cast around
400
the base of the Tower at plinth level.
This ring acts as a base for supporting ' 200
specially cast lead ingots which were 0
placed one at a time at suitable time
spans. The movements experienced
by the Tower are measured with a Figure 6.74: History of
highly redundant monitoring system. counterweight loading
Burland et al. (1994) describe the
response of the Tower to the
application of the counterweight.
Construction of the concrete ring 30
commenced on 3rd May 1993 and the
first lead ingot was placed on 14th July
1993. Figure 6.74 shows the sequence
of load application. It can be seen 20
that, after construction of the concrete
ring, the load was applied in four
phases, with a pause between each
phase to give time to observe the 10 Observations
response of the Tower. The final Prediction
phase was split in two either side of
the Christmas break. The last ingot 400 600
was placed on 20th January 1994.
Figure 6.75 shows a comparison Applied load (t)
w 1
of the class A predictions from the
plane strain analysis and 1 2
measurements of (a) the changes in
inclination, and (b) the average I 3
settlements of the Tower relative to
the surrounding ground during the Figure 6.75: Plane strain prediction
application of the lead ingots. The and observed response due to
points in Figure 6.75 represent the application of counterweight
measurements at the end of each (class A prediction)
phase of loading and the vertical lines

extending from them show the
amount of creep movement between
30
each phase. For the final phase the
creep after one month is shown. It can
be seen that the class A predictions of
the computer model give changes in
inclination which are about 80% of 20
the measured values. However, the

settlements are in excellent agreement 3 Observations
with the measurements. O
10 3D prediction
It is perhaps worth emphasising Plane strain
that the purpose of the model was to prediction
clarify some of the basic mechanisms
of behaviour and it was calibrated 200 400 600
1
against inclinations measured in Applied load (t)
degrees. The use of the model in
studying the effects of the
counterweight was to check that 1 2
undesirable and unexpected responses
of the Tower did not occur. In this
I 3
respect the model has proved to be
very useful. It has led to a Figure 6.76: Comparison of the 3D
consideration of the effects of ageing and plane strain model for the
and it has drawn attention to the application of the counterweight
importance of limiting the magnitude (class C prediction)
of the load so as to avoid yield in the underlying Pancone clay. Also the beneficial
effects of increasing the eccentricity have been highlighted.
It is perhaps expecting too much of the model for it to make accurate
quantitative predictions of movements which are two to three orders of magnitude
less than those against which it was calibrated, and the fact that it has done as well
as it has is remarkable. However, the observed movements due to the
counterweight may be used to further refine the model. It has been found that the
difference between the predicted and measured inclinations are due largely to the
values of G/po' in Horizon A used in the analysis and given in Table 6.5.
Considerably improved agreement can be obtained if the values of Glp(' in
Horizon A are reduced. By trial and error it was found that by reducing these
values by a factor of 0.65, the northward rotation due to a counterweight of 6.9MN
was increased to 37.5 arc seconds, which corresponds to the observed value one
month after the application of the final lead weight. The results are shown in Figure
6.76 and the agreement with the observed rotations is much improved compared
with Figure 6.75. It is of interest to note that the reduction in shear modulus in
Horizon A only slightly increases the predicted settlement. The re-calibrated model
is used in the remainder of this section.
The three dimensional analyses were carried out after the application of the
counter-weight and the revised values of GIpJ were incorporated in it. The
predictions of the response of the three dimensional model to the application of the
counterweight and comparison with the plane strain model is shown on Figure
6.76. Excellent agreement is achieved for the changes in inclination. The three
dimensional model gives slightly smaller average settlements than the plane strain
model.
6.6.3.12 Permanent stabilisation of the Tower

Several alternative options for permanent stabilisation of the Tower were
considered and some of these were analysed using the numerical models described
above. One of these options involved the installation often ground anchors on the
north side of the Tower. Initially this was suggested as a replacement for the
counterweight solution, however, its use as a permanent solution was also
considered. Analysis of this option indicated that as the anchor loads increased, the
Tower initially rotated towards the north, but that after a relatively small rotation
it reverted to rapidly rotating towards the south. The associated settlements also
increased rapidly at this point. In light of the results for the counterweight solution
presented in Figure 6.73 such a result is perhaps not surprising.
Another permanent solution that was considered was the provision of a north
pressing slab. This involved casting a concrete slab on the ground surface to the
north of the Tower and loading it with ground anchors extending into the Lower
sand. Numerical analysis of this option showed that large loads were required to
cause modest rotations of the Tower and that under some loading scenarios the
Tower could revert to rotating towards the south.
The numerical analysis therefore showed that neither of the above options were
likely to be viable and consequently they were not pursued further. The permanent
solution that was finally adopted involved soil extraction from under the north side
of the Tower. This was extensively analysed and some of the results are presented
below.
6.6.3.13 Soil extraction

Introduction
Under-excavation is the process whereby small quantities of soil are excavated
locally by means of a specially designed drill, so as to leave a cavity which closes
under the overburden pressure. Closure of the cavity results in localised subsidence
of the overlying ground surface. The technique can be used for inducing controlled
subsidence of a building.
The Commission considered the use of this method as a means of controlled
reduction of the inclination of the Tower and the arrangement is shown
schematically in Figure 6.77. There was considerable uncertainty about the
response of the Tower to the process of soil extraction from beneath the north side
of the foundation, as the Tower is very close to leaning instability. The purpose of
the numerical analyses presented here

was to explore the response of the
computer model of the Tower to
under-excavation and to examine the
stress changes that the process
induces beneath the foundations and
within the soil beneath the south side.
It should be emphasised that the
finite element meshes had not been
developed with a view to modelling
under excavation. The individual
elements are rather large for
Figure 6.77: Schematic presentation
representing regions of extraction. To
of the soil extraction process
have repeated the whole analysis
using a new finer mesh would have been both expensive and very time consuming.
In any case, the purpose of the modelling was to throw light on the mechanisms of
behaviour rather than attempt a somewhat illusory 'precise' analysis.
The objective of the simulation of the soil extraction process was to reduce the
volume of any chosen element of ground incrementally, so as to achieve a pre-
determined reduction in volume of that element. Localised soil extraction was
modelled by selecting a chosen element in the finite element mesh, eliminating its
stiffness and then progressively reducing its volume by the application of equal and
opposite vertical nodal forces to its upper and lower faces so as to compress the
element. The nodal forces were increased progressively until the desired volume
reduction had taken place. The stiffness of the element was then restored. It is
worth commenting on why only the vertical stresses were reduced. The actual
operation of soil extraction is from a near horizontal element about lm long and
0.2m diameter. With such a long thin element the boundary displacements will be
primarily vertical. As mentioned above, the finite element mesh was not designed
to simulate the process of soil extraction and the length to depth ratios of the
elements are much smaller. So as to encourage vertical displacements of the upper
and lower faces, only the vertical stresses were reduced. To date only plane strain
analyses have been performed to simulate soil extraction.
Critical line
Simple studies carried out on one-g models on sand at Imperial College pointed to
the existence of a Critical line. Soil extraction from any location north of this
Critical line gave rise to a reduction in inclination, whereas extraction from south
of the line gave rise to an increase in inclination. The first objective of the
numerical analysis was to check whether the concept of a Critical line was valid.
Figure 6.78 shows the finite element mesh in the vicinity of the Tower.
Elements numbered 1,2,3,4 and 5 are shown extending southwards from beneath
the north edge of the foundations. Five analyses were carried out in which each of
the elements was individually excavated to give full cavity closure, and the
response of the Tower computed. For

excavation of elements 1, 2 and 3 the
inclination of the Tower reduced so
that the response was positive. For
element 4 the response was
approximately neutral with an initial
slight reduction in inclination which,
with further excavation, was reversed.
For element 5 the inclination of the
Tower increased as a result of Figure 6.78: Excavated elements
excavation. used to investigate the existence of
The above analyses confirm the the Critical line
concept of a Critical line separating a
positive response from a negative one. For the plane strain computer model the
location of the Critical line is towards the south end of element 4, which is at a
distance of 4.8m beneath the foundation of the Tower, i.e. about one half the radius
of the foundations.
It was noted that, as the location of excavation moved further and further south
beneath the foundation, the settlement of the south side steadily increased as a
proportion of the settlement of the north side. Excavation of elements 1 and 2 gave
a proportion of less than one quarter.
It is concluded from this study that, provided soil extraction takes place north
of a Critical line the response of the Tower is positive, even though the foundations
are close to leaning instability. For the plane strain model the Critical line is
located about half the radius of the foundations south of the northern edge of the
foundations. It is further concluded that it should be possible to keep the
settlements of the south edge of the foundations to less than one quarter of the
settlements at the north, provided soil extraction does not extend southwards by
more than about 2m beneath the foundations.
Simulation of under-excavation process

Having demonstrated that localised soil extraction gives rise to a positive response,
the next stage was to model a complete under-excavation intervention aimed at
safely reducing the inclination of the Tower by at least 0.25°. A preliminary study
of extraction was carried out, using a shallow inclined drill hole beneath the
foundations. Although the response of the Tower was favourable, the stress
changes beneath the foundations were large. Consequently, a deeper inclined
extraction hole was investigated.
The inset in Figure 6.79 shows the finite element mesh in the vicinity of the
foundations on the north side. The elements numbered 6 to 12 were used for
carrying out the intervention and are intended to model an inclined drill hole. It
should be noted that element 12 lies south of the Critical line established by
localised soil extraction as described above. The procedure for simulating the
under-excavation intervention was as follows:
a) The stiffness of element 6 was reduced to zero.

b) Equal and opposite vertical nodal forces were applied progressively to the
upper and lower faces until the volume of the element had reduced by
about 5%. The stiffness of the element was then restored.
c) The same procedure was then applied successively to elements 7,8,9, 10
and 11, thereby modelling the progressive insertion of the extraction drill
probe. For each step the inclination of the Tower reduced.
d) When element 12 was excavated the inclination of the Tower increased,
confirming that excavation south of the Critical line gave a negative
response. The analysis was therefore re-started after excavating element
11.
e) The retraction of the drill probe was then modelled by excavating
elements 10,9, 8,7 and 6 successively. For each step the response of the
Tower was positive.
f) The whole process of insertion and extraction of the drill probe was then
repeated. Once again excavation of element 12 gave a negative response.
The computed displacements of
the Tower are plotted in Figure 6.79.
The sequence of excavation of the
elements is given on the horizontal
axis. The upper diagram shows the
change of inclination of the Tower
during under-excavation and the
lower diagram shows the settlements
of the north and south sides of the
foundation.
It can be seen that as under-
6 7 8910111098 7 6 7 8 910111098 7 6 7 8 91011
excavation progresses from elements
Element number excavated
6 through to 11, the rate of change of 11 i i i i i i i i i i i i i i i i i i i i i i i i
northward inclination increases, as do

the settlements. As the drill is
retracted the rate decreases. At the
end of the first cycle of insertion and
extraction the inclination of the
Tower is decreased by 0.1°. The
settlement of the south side is rather
more than one half of the north side.
For the second cycle of insertion and
extraction of the drill a similar Figure 6.79: Response of Tower to
response is obtained, but the change soil extraction
of inclination is somewhat larger. After the third insertion of the drill the resultant
northward rotation was 0.36°. The corresponding settlements of the north and south
sides of the foundation were 260mm and 140mm respectively.
Figure 6.80 shows the contact stress distributions at various stages of the under-
excavation intervention. The process results in a small reduction of stress beneath
the south side. Beneath the north side fluctuations in contact stress take place as is
to be expected, but the stress changes are small.
Distance (m)
8 12 16 20
s 400
/Initial
800
a) After second insertion of the extraction drill

1200
0
Initial
400
800
o
U b) After second retraction of the extraction drill
1200
0
Initial
itial
400 — X. — N
800
c) After third insertion of the extraction drill

1200
Figure 6.80: Effective contact

stress distribution beneath the
Tower foundation
6.6.3.14 The response of the Tower to soil extraction

The results of the modelling work were sufficiently encouraging to undertake a
large-scale development trial of the drilling equipment. For this purpose a 7m
diameter eccentrically loaded instrumented footing was constructed in the Piazza
north of the Baptistry. Drilling was carried out using a hollow-stemmed continuous
flight auger inside a contra-rotating casing. When the drill was withdrawn to form
the cavity, an instrumented probe located in the hollow stem was left in place to
monitor its closure. The trials showed that cavities formed in the Horizon A
material closed gently and that continued extraction from the same location could
be achieved. The trial footing was successfully rotated by about 0.25° and
directional control was maintained even though the ground conditions were
somewhat non-uniform. The measured contact stress changes were very small.
In view of the encouraging results from the numerical analysis and the large-
scale trial, the decision was taken by the Commission to carry out preliminary soil
extraction beneath the north side of the Tower itself, with the objective of
observing the response of the Tower to a limited and localised intervention.
Preliminary soil extraction was carried out over a width of 6m, using twelve
200mm diameter boreholes. A target of a minimum of 20 arc seconds reduction in
inclination was set as being large enough to demonstrate unequivocally the
effectiveness of the method.
On 9th February 1999, in an atmosphere of great tension, the first soil extraction
took place. For the first week the Tower showed no discernable response, but
during the following tense days it began very gradually to rotate northwards. As
confidence grew, the rate of soil extraction was increased. At the beginning of June
1999, when the operation ceased, the northward rotation was 90 arc seconds and
by mid-September it had increased to 130 arc seconds. At that time three of the 97
lead ingots (weighing about lOt each) were removed and movement ceased.
During preliminary under-excavation soil extraction mainly took place outside
the footprint of the foundation and locally only extended beneath the north edge
of the foundation by about 1.5m. It is of interest that the southern edge of the
foundation was observed to rise by about one tenth of the settlement at the north.
This may be contrasted with the numerical model which predicted small
settlements at the south. The reason for this difference may be due to the fact that
a plane strain model was used, whereas the soil extraction process is highly three
dimensional. It is hoped to study this in more detail. Whatever the reason, the uplift
at the south is highly beneficial, as the volume of soil to be extracted is reduced
and it seems likely that reduction of stress is taking place in this critical region.
Having demonstrated that soil extraction produced a positive response, the
Commission formally approved the application of the method for permanent
stabilisation. Using 41 extraction tubes, work on the full intervention commenced
on 21 st February 2000. It is estimated that it will take about eighteen months of
careful soil extraction to reduce the inclination of the Tower by about half a
degree, which will be barely visible. At the time of writing (July 2000) a reduction
of inclination of 800 arc seconds has been achieved and the pattern of uplift at the
southern edge has been maintained. There is still a long, tense journey ahead but
without the positive results of the numerical analysis it is doubtful that this very
sensitive operation on a Tower that is on the point of leaning instability would have
been undertaken.
6.6.3.15 Co mm en ts
This section describes the development and calibration of two finite element
computer models of the Pisa Tower and underlying ground using the modified
Cam clay model with coupled consolidation. One of the models is plane strain and
the other is three dimensional. The choice of compressibility parameters turned out
to be crucial and the use of the intrinsic compressibility C* has proved useful in
interpreting the results of oedometer tests. The main purpose of the model was to
aid the understanding of the basic mechanisms of behaviour of the ground-
structure interaction. Unlike most other attempts to simulate the behaviour of the
Pisa Tower, the model is self consistent in that it generates its own overturning
moment in response to any change of inclination.
Calibration of the model was achieved by ensuring that it gave the correct final
inclination of the actual Tower. It is shown that the model is in remarkable
agreement with the deduced historical inclinations of the Tower during and
subsequent to construction. The results of the analysis confirm beyond reasonable
doubt that the lean of the Tower is caused by a mechanism known as 'leaning
instability', which results from the high compressibility of the underlying clays and
is not related to strength. At an inclination of 5.44° the Tower is predicted to be in
a state of unstable equilibrium, confirming that the actual Tower is very close to
falling over.
The computer model has proved valuable in assessing the effectiveness of the
north counterweight solution for temporally increasing the stability of the Tower
foundations. Although the model was not developed with a view to making precise
predictions of very small movements, nevertheless it predicted the response to the
application of the counterweight with remarkable accuracy. The predicted
reductions in inclination were about 80% of the observed ones and the settlements
were almost exactly as predicted. The observed response of the Tower has been
used to slightly refine the model for future use in assessing the effectiveness of
various possible permanent solutions.
The technique of soil extraction is currently being used as means of inducing
controlled subsidence on the north side of the Tower as a permanent solution. In
view of the uncertainty about the response to soil extraction of a Tower close to
leaning instability, it was essential to study the mechanisms of behaviour with the
numerical model. A method of simulating local soil extraction was developed.
Initial studies have demonstrated the existence of a Critical line north of which soil
extraction leads to a positive response. The location of this Critical line appears to
be about Vi a radius in from the north side.
The process of under excavation by means of an inclined drill was then
simulated. A significant reduction in inclination of the model Tower was achieved.
Moreover, it was shown that the changes in foundation contact stress distribution
were very small. By confining soil extraction to a short distance beneath the north
side, settlements beneath the south side can be kept to less than % of those on the
north side.
It can be concluded from the numerical studies that the technique of soil
extraction offers a very positive method of permanently reducing the inclination
of the Tower by as much as a V-i. This encouraging result led to a successful large
scale field trial of the soil extraction technique. The technique is now being applied
to permanently stabilise the Tower. At present (July 2000) a reduction of 800 arc
seconds has been achieved and the observed behaviour has been consistent with
that predicted by the analyses. It is doubtful that this very sensitive operation on
a Tower that is on the point of leaning instability would have been undertaken
without the positive results of the numerical analysis.
6.7 Summary
1. Shallow foundations can have different shapes in plane. If they are long in one
dimension, they are classified as strip foundations and can be analysed
assuming plane strain conditions. If they are circular, they can be analysed
either assuming axi-symmetric conditions, if the loading is vertical, or using
the Fourier series aided finite element method for general loading. For other
shapes a full 3D analysis is required.
2. When modelling surface foundations both the soil and the foundation should,
in general, be discretised into finite elements. However, if the loading is
vertical and the footing is assumed to be either very flexible or very stiff
compared to the soil, further approximations can be made in which it is no
longer necessary to include the foundation in the mesh.
3. The bearing capacity for a square surface footing is smaller than that of a
circular footing.
4. Finite element analyses enable theoretical shape factors to be determined for
surface foundations. These differ depending on the roughness of the footing,
but in general agree with the empirical formulae used in most design manuals.
5. For drained soils there are three coefficients in the general bearing capacity
equation, namely Nc , A^ and NY . For strip footings exact theoretical
expressions are available for Nc and Nq. Only approximate solutions (e.g.
stress fields) exist for Ny. It has been shown that finite element analyses can
recover the theoretical correct values of Nq , but more importantly such
analysis provide considerable insight in to the values of Ny. They have also
indicated some surprising differences between the behaviour of strip and
circular footings.
6. No design guidance is available for the undrained bearing capacity of pre-
loaded strip foundations on clay. It has been shown how finite element
analyses can be used to tackle this problem. It was concluded that for the
majority of real situations it is unlikely that pre-loading will give rise to a
substantial improvement in undrained bearing capacity.
7. Results of finite element analyses using the sophisticated MIT-E3 have been
presented to show how the effects of observed anisotropic soil behaviour can
be reproduced in finite element analyses. The effect of this anisotropic
behaviour on the bearing capacity of both strip and circular surface
foundations has been quantified by finite element analysis.
8. For shallow footings founded below the ground surface it is important to
correctly model the interface between the sides of the footing and the soil.
Interface elements are useful for this purpose, but must be given a zero tensile
stress capacity, as well as the appropriate shear strength.
9. Finite element analyses indicate that the depth factors for circular footings are
higher than for strip footings.
10. The example of the leaning Tower of Pisa has been used to demonstrate finite
element analyses of shallow foundations. The difference between bearing
capacity failure and leaning instability has been described. It was shown how
finite element analyses were used in the decision making process for the
temporary North weighting and permanent under-excavation schemes at Pisa.

Finite Element Analysis in Geotechnical Engineering Volume Two - Application 6. Shallow Foundations

Uploaded by

Copyright:

Available Formats

Finite Element Analysis in Geotechnical Engineering Volume Two - Application 6. Shallow Foundations

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Finite Element Analysis in Geotechnical Engineering Volume Two - Application 6. Shallow Foundations

Uploaded by

Copyright:

Available Formats

6.

6.3 Foundation types

6.3.1 Surface foundations

6.3.2 Shallow foundations

6.4 Choice of soil model Figure 6.2: Shallow foundations

6.5 Finite element analysis of surface foundations

required to analyse rectangular footings

6.5.2 Flexible foundations

6.5.3 Rigid foundations

Figure 6.7: Boundary condition options for rigid

6.5.4 Examples of vertical loading

6.5.4.2 Strip footings on undrained clay

Load-displacement curves for the

velocity field solutions). It is therefore

used. It can be seen from the inset on

the footing occurs. This is possible

Su = 100 kPa, E= 100 MPa, B = lm

Figure 6.13: Effect of Ko on failure mechanism and

6.5.4.3 Effect of footing shape on the bearing capacity of

foundations. The same soil

directions. Loading of the footing was 5m

6.2 for smooth and rough footing

bearing capacity, although there is no Figure 6.15: Load-displacement

meshes similar to that used for the

vertical normalised displacement of

factor, sc, by the following equation:

with LIB values of 1, 2 and 4. These

soil 1.25 Skempton's empirical expression

where B is the half width of the

footing, see Figure 6.20, and Nc, Nq and Ny

value of Nq = 10.7. The ultimate

dilation equal to the angle of shearing

rough, v = 0 rough, v = 25°

Figure 6.22: Effect of roughness and dilation on

6.5.4.5 Strip footings on a drained soil

Figure 6.24. The smooth footings

footings explains the difference in the Ny values.

Figure 6.25: Development of failure beneath a smooth

In all of the Ny analyses presented above Ko = 0.577. To investigate the effect

Figure 6.27: Effect of Ko on failure mechanism and

6.5.4.6 Circular footings on a weightless drained soil

footing the larger the ultimate Nq

rough, v = 0° rough, v = 25°

Figure 6.29: Failure mechanism for a circular

Table 6.1: Bearing capacity factor Nq

Smooth footing, q' = lOOkPa, v = 0° 10.03 15.57

Smooth footing, q' = lOOkPa, v = 25° 10.8 16.01

Rough footing, q' = lOOkPa, v = 0° 9.92 18

Rough footing, q' = lOOkPa, v = 25° 11.03 19.42

Smooth footing, q' = lOkPa, v = 0° 10

6.5.4.7 Circular footings on a drained soil

Table 6.2: Bearing capacity factor NY

Smooth, v = 0° 3.51 3 6.76 - 3.59 3.14

Smooth, v = 25° 3.51 3 6.76 - 3.74 3.31

Rough, v = 0° 11.6 13.5 6.76 - 6.74 6.62

Rough, v = 25° 11.6 13.5 6.76 - 6.72 6.64

The vectors of incremental displacements at failure are shown in Figure 6.31.

rough circle, v = 0° rough circle, v = 25°

Figure 6.31: Failure mechanism for a