You are on page 1of 29

Randolph, M. F. (2003). Geotechnique 53, No.

10, 847875
847
Science and empiricism in pile foundation design
M. F. RANDOLPH

Scientic approaches to pile design have advanced enor-


mously in recent decades and yet, still, the most funda-
mental aspect of pile designthat of estimating the axial
capacityrelies heavily upon empirical correlations. Im-
provements have been made in identifying the processes
that occur within the critical zone of soil immediately
surrounding the pile, but quantication of the changes in
stress and fabric is not straightforward. This paper
addresses the degree of condence we can now place (a)
on the conceptual and analytical frameworks for estimat-
ing pile capacity, and (b) on the quantitative parameters
required to achieve a design. The discussion is restricted
to driven piles in clays and siliceous sands, with particu-
lar attention given to extrapolating from design ap-
proaches derived for closed-ended piles of relatively small
diameter to the large-diameter open-ended piles that are
used routinely in the offshore industry. From a practical
viewpoint, we need design approaches that minimise
sensitivity to the estimated pile capacity. This may be
achieved partly through a greater reliance on pile load
testing, where signicant advances have been made in the
last decade, but also by adopting design approaches that
are focused more on guarding against unacceptable de-
formation of the complete foundation. Example applica-
tions in the paper are drawn both from offshore
applications, where current challenges include estimating
the axial capacity of ultra-thin-walled, large-diameter
caissons, and from onshore applications such as bridge
piers and piled raft foundations, where inelastic displace-
ment of the piles is not only acceptable, but often
essential for efcient design.
KEYWORDS: axial capacity; dynamic testing; pile driving; pile
foundations; pile groups; piled rafts
Les methodes scientiques servant a` la conception des
piles ont fait denormes progre`s pendant les dernie`res
decennies et pourtant laspect le plus fondamental de ce
travail de conceptionlestimation de la capacite axiale des
pilessappuie encore lourdement sur des correlations
empiriques. Des ameliorations pour identier les proces-
sus qui se produisent dans la zone critique de sol dans le
voisinage immediat de la pile ont ete faites, mais la
quantication des changements de contrainte et de struc-
ture nest pas simple. Cet expose sinterroge sur le degre
de conance que nous pouvons desormais accorder (a)
aux cadres de travail analytiques et conceptuels pour
lestimation de la capacite des piles et (b) aux parame`tres
quantitatifs requis pour leur conception. Cette etude se
limite aux piles enfoncees dans des argiles et sables
siliceux, et extrapole a` partir des methodes conceptuelles
derivees pour des piles fermees de diame`tre relativement
petit et des grosses piles ouvertes qui sont utilisees regu-
lie`rement dans lindustrie offshore. Dun point de vue
pratique, nous avons besoin de methodes conceptuelles qui
minimisent limportance de la capacite estimee de la pile.
On peut y arriver en partie en accordant une plus grande
abilite aux essais de chargement de pile qui ont fait des
progre`s signicatifs au cours des dix dernie`res annees,
mais aussi en adoptant des methodes conceptuelles qui
sattachent davantage a` empecher une deformation inac-
ceptable de toute la fondation. Les exemples donnes dans
cet expose sont tires des applications offshore ou` les
difcultes actuelles sont destimer la capacite axiale des
caissons de gros diame`tres aux parois ultra minces ; ces
exemples sont egalement tires dapplications sur terre
comme les piles de ponts et les fondations radeaux a` piles
pour lesquelles un deplacement non elastique des piles est
non seulement acceptable mais egalement, souvent, essen-
tiel a` la reussite de la construction.
INTRODUCTION
This paper provides an opportunity to reect on the consid-
erable advances that have been made over the last two
decades in the design of piles and pile groups, and to
identify those aspects of pile performance that may be
estimated by sound conceptual models and analysis, and
those aspects where we still need to rely on empirical
correlations. In the latter case, if we are to extrapolate to
pile geometries or soil conditions outside the current data-
base, we must take care to ensure that the correlations are
consistent with our understanding of mechanics and not
distorted by limitations in the database.
Much of the design of pile foundations is still dominated
by estimation of axial capacity, even in applications such as
pile groups for buildings and bridge piers, where the critical
issue is more likely to be the magnitude of displacements
under operating conditions. Indeed, one of the recommenda-
tions proposed later for onshore applications is to endeavour
to weight design criteria more towards limiting displace-
ments, even for the ultimate limit state, by means of non-
linear analysis of pile group response, rather than expressing
them solely in terms of the capacity of individual piles. By
contrast, in the offshore eld, particularly where individual
piles are used as anchors, axial capacity plays a necessarily
dominant role in design, and here the main challenge is
extrapolation to the extreme geometries now used, including
suction-installed caissons with diameters over 5 m and wall
thicknesses as low as 0
.
5% of the diameter.
In order to limit the scope to manageable proportions, this
paper is restricted to the following topics:
(a) axial capacity of displacement piles (driven or jacked)
in clay and sand
(b) the role of pile testing, and in particular interpretation
of dynamic pile tests
(c) performance of pile groups and piled rafts.
Manuscript received 19 March 2003; revised manuscript accepted 6
October 2003.
Discussion on this paper closes 1 June 2004, for further details see
p. ii.

Centre for Offshore Foundation Systems, The University of


Western Australia, Crawley, Australia.
The Centre for Offshore Foundation Systems is established and
supported under the Australian Research Councils research centres
programme.
This choice is consistent with my belief that we may never
be able to estimate axial pile capacity in many soil types
more accurately than about 30%. We therefore need to rely
on pile tests conducted early during the construction phase
to rene the nal design (generally in terms of varying the
embedded pile length, but possibly also the diameter or
number of piles). Hopefully, however, results from load tests
may allow adequate performance of the pile group to be
demonstrated, allowing for inelastic pile response, even
though extreme loads on individual piles exceed their nom-
inal design capacity.
In each of the three areas above, my aim will be to separate
the scientic and empirical components on which we rely
for design calculations, to identify any empirical correlations
that appear inconsistent with theoretical reasoning, and to
suggest areas where improvements may be possible, either by
new analysis or by gathering more specic data to resolve
current uncertainties. Each of the areas is illustrated by
practical examples based on case histories.
AXIAL CAPACITY OF DRIVEN PILES IN CLAY
Overview
Any scientic approach to predicting the limiting shaft
friction that may be mobilised along the shaft of a driven
pile must consider the changes that occur during installation,
equilibration of excess pore pressures, and loading of the
pile (Fig. 1). As the pile is driven, the soil immediately
adjacent to the pile will undergo severe distortion and
changes to the fabric, with a degree of remoulding and the
potential formation of residual shear planes (Bond & Jar-
dine, 1991). The soil outside the immediate vicinity of the
pile will be displaced outwards, with a strain eld that
resembles spherical cavity expansion ahead of the pile tip,
merging to cylindrical cavity expansion along the pile shaft.
In clay with moderate to low yield stress ratio, which is the
main focus here, the mean effective stress in the soil
adjacent to the pile will gradually reduce during the cyclic
shearing action as the pile is driven, and the interface
friction angle, , will reduce to a residual value consistent
with the high rates of shearing and relatively low level of
effective stress (Lehane & Jardine, 1994), both of which
moderate the degree of damage.
At the end of installation, an excess pore pressure eld
will exist around the pile, arising partly from changes in
mean effective stress due to shearing of the soil, but
primarily from increases in total stress as the soil is forced
outwards to accommodate the volume of the pile. As posi-
tive excess pore pressures dissipate, pore water will ow
radially away from the pile, and soil immediately around the
pile will undergo consolidation, with decrease in water
content and increase in mean effective stress. Outside this
zone, which may extend to a few times the diameter of the
pile, the radial strains are tensile during equilibration
(Randolph & Wroth, 1979). The timescale of equilibration
will be proportional to the square of the pile diameter, d,
and inversely proportional to a coefcient of consolidation,
c
h
, that reects (a) primarily horizontal drainage, and (b)
partial consolidation and partial unloading of the soil domain
(Fahey & Lee Goh, 1995).
The nal phase comprises loading of the pile, resisted by
shaft friction along the pile shaft, and end-bearing pressure
at the pile tip. The limiting shaft friction,
s
, will be
determined by the local radial effective stress at failure, 9
rf
,
and an interface friction angle, , according to

s
9
rf
tan (1)
The magnitudes of and, particularly, 9
rf
will depend on
the very complex processes that occur during pile installa-
tion and subsequent consolidation of the soil close to the
pile. Partial healing of any residual shear surfaces gener-
ated during pile installation may occur, although it is also
likely that will reduce to a residual value quite rapidly as
slip occurs between pile and soil.
The dependence of pile shaft capacity on conditions in a
very narrow zone in the immediate vicinity of the pile no
doubt contributes to the scatter in results from pile load
tests. Even on a single site, it is common for values of shaft
friction, normalised by the average shear strength, s
u
, or
vertical effective stress, 9
v0
, to vary quite widely, emphasis-
ing the sensitivity to details of the installation process. As
an extreme example, in the database of pile shaft friction
measured in nine separate tests at Pentre, Chow (1997)
quotes values for
s
/s
u
or
s
=9
v0
that range by more than
35% from the average values, with no apparent trend with
depth or other soil characteristic.
The complexity of the changes in stress and fabric in the
soil immediately adjacent to a driven pile has limited
analytical treatment of the processes involved, and most
practical design still relies on correlations (ONeill, 2001).
It is now accepted that the simple correlation parameters
(
s
/s
u
) and (
s
=9
v0
) are complex functions of soil para-
metersin particular the yield stress ratio and, more deba-
tably, plasticity index, sensitivity and so forth. As the
undrained strength ratio, s
u
=9
v0
, is also a function of the
yield stress ratio, correlations for shaft friction that are
functions of both shear strength and vertical effective stress
were introduced. Originally this was in the form of the
lambda coefcient (
s
= 2s
u
9
v0
; Vijayvergiya &
Focht, 1972), and more recently the American Petroleum
Institute (API, 1993) guidelines, based on Randolph &
Murphy (1985), have proposed estimating the shaft friction
as the larger from the following two expressions:

s
0
:
5

s
u
9
v0
_

s
0
:
5s
0
:
75
u

9
0
:
25
v0
(2)
In all these correlations, there appears to be an effect of
pile length, or embedment ratio, L/d, with the average
normalised shaft friction decreasing with increasing embed-
ment ratio. This has been addressed by incorporating correc-
tions for values of L/d above a certain threshold (Semple &
Rigden, 1984), or by using a power law correlation such as
that proposed by Kolk & van der Velde (1996):

s
0
:
55s
0
:
7
u

9
0
:
3
v0
40
L=d
_ _
0
:
2
(3)
(a) (b) (c)
Fig. 1. Three main phases during history of driven pile: (a)
installation; (b) equilibration; (c) loading
848 RANDOLPH
It is clear, however, that correlations of the type given in
equations (2) and (3) are entirely empirical, and coefcients
of variation mostly exceed 25%.
Length effect
The apparent decrease in normalised shaft friction with
increasing embedment ratio has been attributed to two main
mechanisms, associated respectively with the installation and
the loading phases. The latter has been addressed by
Randolph (1983), who showed that, where the load transfer
response along the shaft exhibits strain-softening, progressive
failure of a pile could lead to a signicant reduction in
capacity. Fig. 2 shows the displacement prole down a
typical pile, and the relative states along load transfer curves
at positions A, B and C. The design chart presented by
Randolph (1983) showed that, for piles where the end-
bearing capacity was much less than the shaft capacity, the
reduction factor, R
f
, dened as
R
f

Q
actual
Q
rigid
(4)
where Q
actual
is the actual pile capacity and Q
rigid
is the ideal
capacity of a rigid pile (calculated as the integrated peak
shaft friction), could be expressed as a function of (a) the
degree of strain softening,
residual
/
peak
, and (b) the
relative compressibility of the pile.
The pile compressibility may be expressed conveniently as
the ratio of the elastic shortening of the pile, treated as a
free-standing column subjected to a load equivalent to the
ideal shaft capacity, dL(
peak
)
average
, to the local displace-
ment, w
res
, required for degradation from peak to residual
shaft friction. Thus the compressibility factor, K, is dened
as
K
dL
2

peak
=(EA)
pile
w
res
(5)
where (EA)
pile
is the cross-sectional rigidity of the pile. The
reduction factor, R
f
, will also be affected to some degree by
the soil stiffness (or local displacement to peak shaft fric-
tion) and the precise shape of the load transfer curves.
Therefore the actual reduction should be evaluated for any
given case, by means of numerical analysis. However, to a
rst approximation for preliminary design calculations, the
reduction factor may be expressed as
R
f
% 1 (1 ) 1
1
2

K
p
_ _
2
for K . 0
:
25 (6)
with R
f
taken as (approximately) unity for smaller values
of K.
The strain-softening load transfer response arises from
reduction of the radial effective stress, 9
r
, at the pile shaft
and, more signicantly, the reduction in interface friction
angle, , to a residual value. Ring shear tests suggest that
the softening factor, , may lie in the range 0
.
50
.
8 (com-
pared with a recommendation of 0
.
7 in the American
Petroleum Institute guidelines: API, 1993), with the lower
range possible for high-plasticity clays at moderate to large
effective stress levels. Ring shear tests show that most
strain-softening occurs within relatively small displacements
(1030 mm), although it is possible that w
res
for full-scale
piles might be somewhat larger. For modern offshore pile
geometries, where the L/d ratio rarely exceeds 60, typical K
values would not exceed 510, giving rise to reduction
factors in the range 0
.
650
.
9. Progressive failure can there-
fore still lead to a signicant reduction in the ideal capacity.
The other source of length effect is that associated with
stress changes during installation. This has been quantied
by means of measurements from instrumented piles, particu-
larly the extensive research programme undertaken at Imper-
ial College (Jardine and co-workers Bond, Lehane and
Chow). A summary of radial stress changes measured at the
end of jacked pile installation from three different clay sites
has been presented by Lehane & Jardine (1994), as shown
in Fig. 3; each value of radial stress has been normalised by
the local cone resistance, q
c
. The three sites comprise
heavily overconsolidated London clay, a stiff glacial till
(Cowden), and a lightly overconsolidated silty clay or clayey
silt (Bothkennar).
The measured radial stresses have been tted by power
law curves of the form
A
B
C

w
w
w
Displacement profile
Fig. 2. Progressive failure of pile in strain-softening soil
0 0
.
2 0
.
4 0
.
6 0
.
8 1
.
0
0
5
10
15
20
25
30
D
i
s
t
a
n
c
e

f
r
o
m

p
i
l
e

t
i
p
,

h
/
d
Normalised radial stress,
ri
/q
c
n 0
.
6 0
.
35 0
.
2
Decay
curves
(d/h)
n
Bothkennar
Cowden
London
Profile from
strain path method
(Whittle, 1992)
Fig. 3. Radial stress changes during jacked pile installation in
clay (after Lehane & Jardine, 1994)
SCIENCE AND EMPIRICISM IN PILE FOUNDATION DESIGN 849

ri
q
c
/
d
h
_ _
n
(7)
where h is the distance from the pile tip (equivalent to
L z, where z is the depth and L is the embedded pile
length). Deduced values of n range from 0
.
2 to 0
.
6, although
reasons for the higher values of n have been discussed by
Coop & Wroth (1990). Also shown in Fig. 3 is a prediction
from Whittle (1992), using the strain path method with soil
parameters based on the Bothkennar site. By contrast with
the measured data, the analytical prediction shows a varia-
tion in normalised radial stress only in the lower few
diameters, where it reduces from a value close to unity down
to a value of 0
.
5, after which it remains constant.
The divergence between the science of the strain path
method prediction and the empiricism of the t to eld
data suggests that further study of the processes involved is
required, and we must explore what facet of soil behaviour,
or of experimental technique, may have led to this differ-
ence. For practical application, it will also be necessary to
decide how to extrapolate from the eld measurements,
which are on full-displacement, closed-ended piles, to allow
estimation of stress changes around partial-displacement,
open-ended pipe piles. These aspects may be explored con-
veniently through the analogy of cavity expansion.
Cavity expansion analogy for excess pore pressures and
equilibration times
The analogy of cylindrical cavity expansion to model the
installation of displacement piles formed the basis of early
attempts to quantify stress changes due to pile installation
(Kirby & Esrig, 1979; Randolph et al., 1979). Subsequently,
the strain path method, pioneered by Baligh at MIT (Baligh,
1985, 1986), provided more realistic and detailed predictions
for the strains and stress changes in the immediate vicinity
of the pile, particularly in respect of the zone of very high
stress gradients ahead of, and behind, the pile tip and the
transition to quasi steady-state conditions (in terms of nor-
malised stresses) along the pile shaft. Comparison of the two
approaches shows that, ignoring the few diameters close to
the pile tip, the radial displacement elds are extremely
similar apart from immediately adjacent to the pile shaft
(within a zone of thickness about 10% of the pile radius, for
a full-displacement pile).
Assuming that pile installation occurs under undrained
conditions, the radial displacement, r, for soil at nal
radius, r, may be deduced as
r
r
eq

r
r
eq

r
r
eq
_ _
2
1

(8)
where r
eq
is the pile radius for a closed-ended pile, and for
an open-ended pile is the radius of an equivalent solid pile
that gives the same volume of displaced soil. For thin-walled
piles of wall thickness t the equivalent pile radius and
diameter are
r
eq
%

dt
p
; d
eq
% 2

dt
p
(9)
where it is assumed implicitly that the pile is installed in an
unplugged manner, with the top of the internal soil plug
remaining (approximately) level with the external soil
surface.
The relationship in equation (8) is shown in Fig. 4 for a
closed-ended pile (r
eq
r
pile
), and also an open-ended pile
for a d/t ratio of 40, which is a typical value for steel pipe
piles. It is shown dashed in the region close to the (solid)
pile, where the cavity expansion solution is no longer
deemed accurate. The location of the open-ended pile is
indicated, and the thicker line for r/r
eq
greater than 3
.
2 is
applicable to the open-ended pile. The right-hand axis gives
the radial displacement for the open-ended pile, normalised
by the actual pile radius, r
pile
, rather than the equivalent
radius, r
eq
. Note that, for d/t of 40, the area ratio (of pile
wall to the gross cross-sectional area of the pile) is r 0
.
1
($ 4t/d), and the equivalent radius is 0
.
32 times the actual
radius.
The assumed radial expansion for the open-ended pile
shown schematically in Fig. 4 is such as to accommodate
the full wall thickness, essentially modelling the pile as a
perfect sampling tube. Support for this assumption comes,
experimentally, from the observation that, under the dynamic
conditions of pile driving, the soil plug does indeed appear
to progress up the pile, with only small variations in the
position of the top of the soil plug relative to the original
ground surface.
The excess pore pressures generated by pile installation
arise from two sources: changes in mean effective stress
during shearing and partial remoulding of the soil (which
will give rise to positive excess pore pressures for lightly
overconsolidated clay, and negative pore pressures for poten-
tially dilatant, heavily overconsolidated clay), and increases
in mean total stress due to outward expansion of the soil
to accommodate the pile volume.
Simple cavity expansion theory, applied to an elastic,
perfectly plastic soil with shear modulus G and undrained
shear strength s
u
would give rise to an excess pore pressure
distribution of (Gibson & Anderson, 1961)
u
s
u
ln
rG
s
u
_ _
2 ln
r
r
pile
_ _
> 0 (10)
Although this expression does not account for changes in
mean effective stress as the soil is sheared and remoulded,
these may be accounted for approximately for lightly over-
consolidated clays by adjustment of the rigidity index, I
r

G/s
u
. The main features of a logarithmic decay with radius,
and typical values of maximum pore pressure adjacent to
full displacement piles of 4s
u
to 6s
u
(in lightly overconsoli-
dated soils) agree well with results from the strain path
method (Baligh, 1986). An important feature of equation
(10) is the term accounting for the area ratio, r, for open-
ended piles, where the reduction in excess pore pressure
compared with a full-displacement (solid or closed-ended)
pile is s
u
ln(r).
The excess pore pressure elds around closed-ended and
open-ended piles (with d/t 40) based on cylindrical cavity
expansion, taking G/s
u
100, are shown in Fig. 5, together
with isochrones during dissipation. Baligh (1986) has com-
Open-ended
Closed-ended
Open-ended pile (d/t 40)
Normalised final radius, r/r
eq
3 2 1 5 10 7
0
0
.
1
0
.
2
0
.
3
0
.
4
0
.
5
0
.
6
0
.
7
0
.
8
0
.
9
1
.
0

r
/
r
e
q

r
/
r
p
i
l
e
0
0
.
05
0
.
10
0
.
15
0
.
20
0
.
25
0
.
30
Fig. 4. Radial displacement eld for closed- and open-ended
piles
850 RANDOLPH
mented that the excess pore pressures predicted by cavity
expansion may be overestimated, as a result of not following
the correct strain path, but for lightly overconsolidated soils
this will be offset by ignoring the excess pore pressures due
to shearing of the clay (with corresponding reduction in
mean effective stress). The isochrones of excess pore pres-
sure shown in Fig. 5 have been derived using the radial
consolidation solution of Randolph & Wroth (1979), with
the non-dimensional times expressed as
T
c
h
t
d
2
; T
eq

c
h
t
d
2
eq
(11)
where t is the time and c
h
is an appropriate coefcient of
consolidation for horizontal drainage. During the consolida-
tion process, the outer soil (beyond 35 pile radii) under-
goes swelling, while the inner soil consolidates: hence the
coefcient of consolidation must reect this fact (Fahey &
Lee Goh, 1995), and is most easily assessed through piezo-
cone dissipation tests.
An interesting (and somewhat surprising) feature of Fig. 5
is that isochrones for equal proportions of excess pore
pressure, u/u
max
, occur at very similar non-dimensional
times, T and T
eq
for the two pile types, as remarked on by
Whittle (1992). This is illustrated in Fig. 6, where the
normalised excess pore pressure is plotted against the two
alternative time factors, for closed and open-ended piles of
different wall thickness ratios. For comparison, a dissipation
curve based on the strain path method, as presented by
Whittle (1992), is also shown. Whittles original curve was
presented using a time factor expressed in terms of vertical
pre-consolidation pressure, 9
p
, and horizontal permeability,
k
h
, and the results in Fig. 6 have been scaled by assuming
c
h
% 159
p
k
h
=
w
, where
w
is the unit weight of water. The
gure shows that the dissipation curves for all the piles of
different wall thickness fall in a narrow band, when ex-
pressed in terms of T
eq
, based on the equivalent pile
diameter, rather than the true diameter. Note also that the
timescale of consolidation is affected by the original magni-
tude of the excess pore pressure ratio, u
max
/s
u
(and hence
the lateral extent of the pore pressure eld), and the results
shown in Fig. 6 are for an initial excess pore pressure ratio
of 4
.
6, corresponding to G/s
u
100 for the cavity expansion
analogy.
In passing, it may be noted that, in their analysis of
dissipation around a piezocone, Teh & Houlsby (1991)
proposed a generalised time factor, T

, given by
T


4c
h
t
d
2

I
r
p (12)
in order to bring together dissipation curves for different soil
rigidity indices. This contrasts with the normalisation using
T
eq
in Fig. 6(b), where normalisation using d
eq
can be shown
to be equivalent to taking T
eq
as inversely proportional to I
r
(rather than to the square root of I
r
). In fact, the optimal
normalisation depends on (a) the range of I
r
values that need
to be considered, and (b) whether the focus is on the early
dissipation response (up to T
50
) or the later response (times
greater than T
50
). For the interpretation of piezocone tests,
with a likely range for I
r
between 50 and 500 and with the
focus on the early dissipation response, the normalisation
proposed by Teh & Houlsby (1991) may be optimal. How-
ever, dissipation around open-ended piles, where the focus is
more on the times for 5090% dissipation, and the range of
rI
r
(in the light of equation (10)) that need to be considered
is very broad, the normalisation shown in Fig. 6(b) using T
eq
appears more useful.
Despite the approximations involved in the cylindrical
cavity analogue for pile installation, it appears that
the general pattern of excess pore pressure, and the
T 0
T 0
.
01
T 0
.
8
T 2
T 10
T 0
.
1
T 0
T
eq
0
.
03
T
eq
0
.
2
T
eq
0
.
8
T
eq
2
T
eq
10
Pile wall
1 3 5 7 9
0
1
2
3
4
5

u
/
s
u
1 3 5 7 9
0
1
2
3
4
5

u
/
s
u
0
.
9
0
.
7
0
.
5
0
.
3
u/u
max

0
.
9
0
.
7
0
.
5
0
.
3
0
.
1
d/t 40, r
pile
32r
eq
u/u
max
0
.
1
Normalised radius, r/r
eq
Normalised radius, r/r
eq
Fig. 5. Excess pore pressures generated by pile installation: (a)
closed-ended pile; (b) open-ended pile
G/s
u
100
G/s
u
100
Closed
d/t 20
d/t 40
d/t 80
d/t 160
Closed
d/t 20
d/t 40
d/t 80
d/t 160
SPM: Whittle
T c
h
t/d
2
10 1 0
.
1 0
.
01 0
.
001
0
01
02
03
04
05
06
07
08
09
10
u
max
u
u
max
u
T
eq
c
h
t/d
eq
2
0
01
02
03
04
05
06
07
08
09
10
10 1 0
.
1 0
.
01 0
.
001
Fig. 6. Dissipation of excess pore pressures at pile shaft
SCIENCE AND EMPIRICISM IN PILE FOUNDATION DESIGN 851
consolidation response, can be predicted reasonably for both
piezocones and driven piles. Two examples illustrating this
are shown in Fig. 7. The rst example is for a closed-ended
driven pile, which was subjected to dynamic tests at differ-
ent times after driving (data kindly supplied by Mr Antonio
Alvez, PhD student at COPPE, Federal University of Rio de
Janeiro). Dissipation from a piezocone test, from which
appropriate values of I
r
and c
h
were deduced, is compared
with theoretical dissipation curves and also the measured
increase of shaft resistance, Q
s
(where Q
s,1
and Q
s0
repre-
sent long-term and initial shaft resistance respectively). Two
different dynamic pilesoil interaction models were used in
analysing the pile tests, a continuum model and the Smith
(1960) model, both of which are described later. Although
the increase in shaft friction is not precisely proportional to
the decrease in excess pore pressure, because of stress
relaxation effects (discussed later) and changes in radial
effective stress during loading, it seems that consolidation
theory gives a sufciently accurate estimate of the timescale
of increase in shaft resistance.
The second example, in Fig. 7(b), is from centrifuge
model tests on very thin-walled suction caissons, reported by
Cao et al. (2002). The prototype dimensions of the caisson
are shown, with a d/t ratio of 80. However, as the caisson
was installed using suction, the outward soil movement may
be less than for a driven pile, as more soil is drawn inside
the caisson (Andersen & Jostad, 2002). Hence the operative
d/t ratio may be nearer 160 than 80 in terms of outward
movement of soil.
In both of these examples, the experimental data are
matched reasonably well by the theoretical dissipation
curves. The rst example, in Fig. 7 (a), also demonstrates
that simple scaling of piezocone dissipation times, by the
square of the diameter ratio (equivalent pile diameter di-
vided by piezocone diameter), should give a reasonable
estimate of the consolidation times for a pile.
Referring to Fig. 6, two important observations may be
made. The rst is that dissipation times for typical open-
ended piles (d/t % 40) and suction caissons (d/t % 200) will
be respectively one and two orders of magnitude shorter
than for a closed-ended pile of the same diameter. The
second observation is that signicant dissipation, with 20%
reduction in pore pressure, occurs for T
eq
% 0
.
1. For typical
values of consolidation coefcient in the range 330 m
2
/yr,
this value of T
eq
corresponds to 0
.
55 days for an open-
ended offshore pile 2 m in diameter, or 0
.
33 days for a
closed-ended onshore pile 0
.
5 m in diameter. These times
are longer than most installation times (except in the case of
equipment breakdown). However, for the 0
.
1 m diameter
instrumented pile used to obtain the data in Fig. 3, 20%
pore pressure dissipation would occur in 0
.
33 h, compared
with total jacking periods of 15 h (Lehane & Jardine,
1994).
It appears, therefore, that partial pore pressure dissipation
during installation may account, at least in part, for the h/d
effect deduced from radial stress measurements, and the
divergence between the trends in the data and theoretical
predictions from the strain path method. Rapid initial pore
pressure dissipation may also account for the low values
reported by Karlsrud (1999) in low-plasticity clays. Such
clays, with high silt content, are likely to show shorter
consolidation times, comparable with pile installation times,
leading to greater damage to the soil (lower residual inter-
face friction angles, because of the higher effective stress
levels during installation), less set-up following installation,
and thus lower shaft friction values than for higher plasticity
clays.
Radial stress changes during installation, equalisation and
loading
The pile shaft friction depends on the radial effective
stress acting around the shaft, according to equation (1), and
this may be estimated by considering the sequential changes
during pile installation, consolidation and loading. Measure-
ments of radial total stress,
ri
(less the in situ pore pressure,
u
0
) immediately after installation, and radial effective stress,
9
rc
, at the end of consolidation, both normalised by the in
situ vertical effective stress, 9
v0
, are shown in Fig. 8 for
values of h/d . 10. The data were assembled by Lehane
(1992), and Fig. 8(a) shows his proposed trend lines (see
also Lehane et al., 1994).
During installation, the trend of radial total stress ratio,

ri
u
0 =9
v0
, increases in proportion to the yield stress
ratio to the power of about 0
.
4, from a value of $2 for
normally consolidated soil, to just under 10 at very high
yield stress ratio. As Lehane (1992) observed, the gradient is
approximately parallel to the correlation of K
0
with over-
consolidation ratio proposed by Mayne & Kulhawy (1982),
with a radial total stress ratio of 33
.
5 times K
0
.
Although the trend in the data on Fig. 8(a) is evident, the
logarithmic scales can lead to quite signicant deviation
from the mean line. At present, the only viable analytical
approach for quantifying detailed stress changes during pile
installation and consolidation appears to be the strain path
method, but simpler quasi-analytical approaches are needed
for routine design. Potential approaches, admittedly some-
what speculative, are discussed here.
After installation, the radial total stress (less the in situ
pore pressure) may be expressed as
u
u
max
10
08
06
04
02
0
0001 001 01 1 10 100
10
08
06
04
02
0
Theory: G/s
u
100
Theory: G/s
u
50
Cone: Mid-face
Cone: Shoulder
Pile: Continuum
Pile: Smith model
Theory: d/t 160
Theory: d/t 80
Test SAT06
Test SAT08
Normalised time, T c
h
t/d
2
65 mm
24 m
52 m
10
08
06
04
02
0
0001 001 01 1
Normalised time after installation , T c
h
t/d
2
00001
(a)
(b)
u
u
max
Q
s
Q
s0
Q
s
Q
s0
Fig. 7. Measured dissipation around closed and open-ended
piles: (a) piezocone dissipation and pile shaft resistance in
high-plasticity clay (data provided by Mr Antonio Alvez); (b)
pore pressure dissipation around thin-walled caisson (data from
Cao et al., 2002)
852 RANDOLPH

ri
u
0
9
ri
u
max
(9
ri
p9
i
) p9
0
p (13)
where u
max
is the maximum excess pore pressure and p9
0
and p9
i
are respectively the original in situ mean effective
stress and the value just after pile installation (adjacent to
the pile shaft). The bracketed term has a relatively narrow
range [negative, owing to the slight unloading strains next to
the pile according to the strain path method (Baligh, 1986;
Whittle, 1992), but limited in magnitude to the current
undrained shear strength, allowing for any remoulding that
may have occurred as the pile is installed]. The in situ mean
effective stress, p9
0
, may be estimated through K
0
, and the
increase in mean total stress, p, required to accommodate
the pile should prove amenable to estimation through numer-
ical analysis (strain path or cavity expansion methods).
Estimating these quantities with any accuracy at present is
not straightforward, but the approach represents a possible
scientic way forward.
During equilibration, the excess pore pressure reduces to
zero and the radial effective stress increases to a nal value
denoted by 9
rc
. The data for the nal radial effective stress
ratio, 9
rc
=9
v0
, in Fig. 8(a) have been correlated with lines
that lie nearly parallel to the trend of the installation
stresses, but are offset by varying amounts, depending on
the sensitivity of the clay (Lehane, 1992; Jardine & Chow,
1996). During consolidation there is some relaxation in total
stress (so that the nal radial effective stress is less than the
initial radial total stress), and the data suggest that the
degree of relaxation is high for low yield stress ratios (also
high sensitivity) and reduces as the yield stress ratio in-
creases (and sensitivity reduces).
Such a trend is consistent with radial consolidation mod-
els, which show the outer soil (beyond 35 times the pile
radius) swelling, while the inner core consolidates (Fahey &
Lee Goh, 1995). It is the difference in stiffness of these two
zones that gives rise to the relaxation in total radial stress
(with no relaxation in classical solutions where the soil is
assumed elastic and homogeneous). The relaxation gradient
at any stage during consolidation is d9
r
=du, and it may be
argued that this quantity will become progressively less than
unity the softer the inner soil is relative to the outer
(swelling material), and will therefore be a function of the
relative magnitude of the current radial effective stress, 9
r
,
and the preconsolidation or yield stress, 9
vc
. This effect may
be captured by a function such as

d9
r
du
e
( 9
r
9
ri
)= 9
vc
(14)
where and are adjustable parameters.
Integrating this expression over the change in excess pore
pressure from u
max
down to zero, the nal radial effective
stress is given by
9
rc
9
v0

9
ri
9
v0

ln 1

R
u
max
9
v0
_ _
(15)
where R is the yield stress ratio, 9
vc
=9
v0
. This expression is
plotted in Fig. 8(b), adjusting to unity and to a value of
5, in order to give a reasonable t to the data (identical to
the data in Fig. 8(a)). Although this approach is speculative,
and signicant further work is needed before it might be
useful in design, the concept of a relaxation gradient that
varies during consolidation is consistent with physical argu-
ments of the conditions around the pile, and also with eld
measurements by Lehane (1992), which indicate a gradual
reduction of jd9
r
=duj during consolidation. One important
consequence is that the net relaxation ratio, (
ri
u
0
)=9
rc
,
will be higher (for a given soil) for an open-ended pile than
for a closed-ended pile.
The nal phase of the piles history to consider is the
loading phase. By the end of consolidation, the radial effec-
tive stress will have become the largest of the three normal
stresses (vertical, radial and circumferential) close to the
pile. During loading of the pile, a reduction in the radial
stress is therefore expected. Lehane (1992) and the design
approach proposed by Jardine & Chow (1996) suggest that
the reduction may be taken as about 20%, independent of
the yield stress ratio, so that 9
rf
in equation (1) is then
0
:
89
rc
.
Example design calculations: new horizons
The offshore industry continues to face new challenges as
it moves into deeper waters and new regions of the world.
Currently, one of the most active offshore areas is off the
west coast of Africa, where very high-plasticity clays have
been encountered in water depths of 1000 m. Typical gener-
ic soil properties, based on data from a number of sites, are
summarised in Table 1.
The combination of high plasticity index with high fric-
tion angles measured in triaxial compression and simple
shear is unusual and, in a similar fashion to Mexico City
clay, lies well outside common correlations of friction angle
with PI (Mesri et al., 1975). Unlike Mexico City clay,
however, interface friction angles are signicantly lower,
particularly at residual. This characteristic poses a particular
challenge in estimating the shaft capacity of driven pipe
piles and thin-walled suction caissons in these clays, as
traditional approaches based on correlations with s
u
and 9
v0
will diverge from more fundamental approaches based on
equation (1).
10
After installation
(
ir
u
0
)/
v

0
R
a
d
i
a
l

s
t
r
e
s
s

c
o
e
f
f
i
c
i
e
n
t
s
(

i
r


u
0
)
/

v
0

a
n
d

r
c
/

v
0
1
01
Increasing
sensitivity
After consolidation

c
/
v

0
1 10 100
Yield stress ratio, R
After installation
(
ir
u
o
)/
v

0
After consolidation

c
/
v

0
R
a
d
i
a
l

s
t
r
e
s
s

c
o
e
f
f
i
c
i
e
n
t
s
(

i
r


u
0
)
/

v
0

a
n
d

r
c
/

v
0
10
1
01
1 10 100
Yield stress ratio, R
Equation (15)
(a)
(b)
Fig. 8. Radial stress coefcients after installation and consolida-
tion (data from Chow, 1997): (a) relaxation ratio as function of
soil sensitivity; (b) relaxation ratio derived from function of
current yield stress ratio
SCIENCE AND EMPIRICISM IN PILE FOUNDATION DESIGN 853
To illustrate the ideas discussed earlier, two different
geometries of offshore piles will be considered:
(a) a conventional pipe pile, 2 m in diameter and with
50 mm wall thickness (d/t 40, r 0
.
1) embedded
100 m (L/d 50)
(b) a suction caisson, 6 m in diameter with wall thickness
of 30 mm (d/t 200, effective area ratio, allowing
for suction installation, of r 0
.
01) embedded 20 m
(L/d 3
.
3).
The shaft capacity of these piles will be estimated using the
approach described here (equations (10), (13) and (15)) and
also the method of Jardine & Chow (1996) (referred to here
as the MTD method, as it is known in the offshore industry).
The MTD method for piles in clay is based on the empirical
correlations of Lehane (1992) and Lehane et al. (1994). In
contrast to the MTD method, the effect of h/d is ignored in
the present approach apart from for h/t , 10, where the
normalised radial total stress is assumed to increase gradu-
ally by a maximum factor of 2 at the pile tip (as suggested
by the strain path method results shown in Fig. 3). Simplis-
tically, the radial effective stress just after installation (9
ri
)
has been estimated using equation (1), assuming that the
shaft friction during installation is equal to the remoulded
shear strength, and the maximum excess pore pressure gen-
erated by a solid pile has been taken as 4
.
6s
u
.
The proles of peak shaft friction obtained from these
two approaches are compared with that estimated using the
API guidelines (API, 1993) in Fig. 9 for each pile geometry.
In Fig. 9(a), the strong h/d effect from the MTD method is
evident, with lower shaft friction over most of the pile shaft,
apart from close to the tip. For the particular combination of
soil properties, it turns out that the approach described here
gives a shaft friction prole that is remarkably similar to
that obtained from API (1993), although this is something of
a coincidence and will be affected by the interface friction
angle, . Average values of shaft friction from the different
methods are quite close, with the MTD method about 10%
lower than the other two methods.
For the caisson (Fig. 9(b)), there is a much greater
divergence of shaft friction proles. The API (1993) prole
is identical to that for the driven pile, whereas the approach
suggested here gives lower shaft friction, largely because of
the low area ratio of the caisson and hence lower excess
pore pressures generated during installation and lower nal
radial effective stresses. The MTD method was not intended
to apply to piles with such low L/d or high d/t ratios, which
fall well outside the database used to calibrate the method.
It is an instructive comparison, however, reminding us that
extrapolation of any design method must be carried out with
care, particularly where the method is based on empirical
correlations.
Whereas the suction caisson may be considered as effec-
tively rigid, in terms of strain-softening effects during axial
loading, the driven pile is relatively exible. The calculated
loaddisplacement responses, assuming strain softening by
40% ( decreasing from 208 to 128) over relative pilesoil
slip of 50 mm, are shown in Fig. 10. The reduction factor
due to strain-softening is around 10%, which is somewhat
less than the value of 15% estimated from equations (5) and
(6), mainly because of the triangular distribution of shaft
friction arising from the linearly increasing shear strength
with depth.
Summary
The science in estimating driven pile capacity in clay
provides the framework within which the different phases of
the installation, consolidation and loading history of the pile
are considered. It also extends to different analytical ap-
proaches, such as the strain path method, and cavity expan-
sion, which allow quantication of certain aspects of each
process. Magnitudes of total stress increase, quantication of
the differences between full and partial displacement piles,
and estimation of the timescale for consolidation may all be
treated analytically. However, design calculations still rely on
Table 1. Clay properties offshore West Africa
Parameter Typical values
Shear strength, s
u
: kPa 1
.
5z (with z the depth in m)
Effective unit weight, 9: kN/m
3
3
.
5
Yield stress ratio, R 1
.
8
Sensitivity, S
t
4
Plasticity index, PI: % 100
Friction angle, 9: degrees 35
Interface friction angle, : degrees 20 (residual value $12)
Shaft friction: kPa
160 120 80 40 0
0
20
40
60
80
100
120
(a)
Shear strength profile
MTD
approach
Present
approach
API (1993)
D
e
p
t
h
:

m
D
e
p
t
h
:

m
(b)
Shaft friction: kPa
30 25 20 15 10 5 0
0
5
10
15
20
25
Present
approach
Shear strength
MTD
approach
API (1993)
Fig. 9. Proles of peak shaft friction for offshore piles: (a)
driven pipe pile (L/d 50, rr 0
.
1); (b) suction caisson (L/d
3
.
3, rr 0
.
01)
854 RANDOLPH
empirical correlations in order to quantify those aspects that
are dominated by the complexities of soil response, such as
reduction in effective stresses and degree of remoulding
during pile installation, relaxation of radial total stress
during consolidation, and reduction in radial effective stress
during loading.
There appears to be divergence between analysis and eld
measurements in respect of the h/d effect during pile instal-
lation, although partial consolidation appears partly respon-
sible. Resolution of this is important, and requires careful
review of what fundamental mechanisms might lead to an
h/d effect. An improved model to quantify stress relaxation
during consolidation is also needed, perhaps through numer-
ical parametric studies, as this is an area where considerable
scatter in the database exists. The concept of a relaxation
gradient that changes as consolidation proceeds, as the
relative stiffness of the inner and outer soil zones evolves,
has been proposed as a possible way forward.
The challenge of providing anchors in deepwater, using
suction-installed caissons, requires extension of our current
design approaches to low aspect ratio (L/d , 6) and low
area ratio (r % 0
.
01) pile geometries. A rational scientic
basis is essential for this.
AXIAL CAPACITY OF DRIVEN PILES IN SAND
Overview
Over the last decade there have been two major advances
in design approaches for driven piles in sand. The rst of
these is the capturing, through instrumented pile tests, of the
gradual degradation of shaft friction at any given depth as
the pile is driven progressively deeper (Lehane et al., 1993),
and the second is the linking of key parameters such as base
resistance and maximum shaft friction to the cone resistance,
q
c
, which has evolved from the early correlations of
Bustamante & Gianeselli (1982). Both of these advances are
empirical in nature, but they embody principles that could,
in due course, be quantied more scientically.
Historically, pile design in sand has been based on simple
linear relationships for both shaft friction and base resis-
tance, but with limiting values at some critical depth
expressed either in absolute terms or normalised by the pile
diameter (Vesic, 1967, 1970; Coyle & Castello, 1981). The
rationale behind this approach has been challenged
(Kulhawy, 1984), and alternative explanations offered for the
experimental nding that increasing lengths of piles driven
into sand do not yield proportional increases in capacity.
For base resistance, the inuence of decreasing friction
angle with increasing stress level, and the non-linear rela-
tionship between stiffness and stress, both contribute to a
decreasing gradient of base resistance with depth (Randolph
et al., 1994). For shaft friction, although equation (1) still
provides the physical basis, the normal effective stress, 9
rf
,
at any given depth has been found to degrade as the pile is
installed, owing to gradual densication of the surrounding
material. These components of the axial capacity of driven
piles in sand, and the necessary adjustments for open-ended
piles, are explored here in the context of recent design
recommendations (Jardine & Chow, 1996).
Base resistance
Although it is natural to correlate the end-bearing resis-
tance of a pile with the cone resistance, consideration must
be given to the displacement needed to mobilise a given
proportion of cone resistance. Fleming (1992) proposed a
hyperbolic relationship for bored piles, relating the end-
bearing pressure, q
b
, and the base displacement, w
b
, giving a
normalised end-bearing resistance, q
b
/q
c
, expressed as
q
b
q
c
%
w
b
=d
w
b
=d 0
:
5q
c
=E
b
(16)
where E
b
is the Youngs modulus of the soil below the pile
base.
For a bored pile, with initially zero base pressure at zero
displacement, this relationship will lead to end-bearing pres-
sures mobilised at a base displacement of 0
.
1d of around
1520% of q
c
(Lee & Salgado, 1999). However, for driven
and jacked piles, signicant residual pressures are locked in
at the pile base during installation (equilibrated by negative
shear stresses along the pile shaft, as if the pile were loaded
in tension). This will lead to a stiffer overall pile response in
compression, and signicantly higher end-bearing stresses
mobilised at small displacements.
The magnitude of residual base stress will depend on the
relative magnitudes of shaft and base capacity, as well as on
the method of installation. For jacked piles the residual base
stress can be as high as 7080% of the lesser of shaft or
(ultimate) base capacity (Poulos, 1987). For driven closed-
ended piles the residual stress will be lower, but may still be
as high as 75% of the base capacity (Maiorano et al., 1996).
The lowest residual base stress is likely to be for open-ended
piles, unless they become fully plugged during driving.
Equation (16) can be generalised to allow for a residual
pressure, q
b0
, locked in below the pile base at the start of
loading, to give
q
b
q
c
%
w
b
=d 0
:
5q
b0
=E
b
w
b
=d 0
:
5q
c
=E
b
(17)
The resulting end-bearing responses are illustrated in Fig. 11
for E
b
/q
c
1
.
25 (lower set of curves for each value of
q
bo
/q
c
) and E
b
/q
c
5 (upper set of curves for q
b0
/q
c
0
.
3
and 0
.
7). This range of E
b
/q
c
reects conservative values
suggested for bored and driven piles (Poulos, 1989; Fleming,
1992).
The exact form of the end-bearing response is of course
debatable. However, the main principles illustrated in Fig. 11
are as follows:
(a) Steady-state conditions are reached after large displace-
ment (410 diameters for zero residual stress),with the
end-bearing resistance of a pile approaching the cone
resistance, after appropriate averaging of the latter
quantity to reect the larger size of the pile.
(b) At limited displacements, such as 10% of the pile
diameter as is often taken as the practical denition of
ultimate, the end-bearing resistance will be signi-
cantly lower than the cone resistance, and will also
010 008 006 004 002 0
0
5000
10000
15000
20000
25000
30000
35000
40000
P
i
l
e

h
e
a
d

l
o
a
d
:

k
N
MTD method
Pile head displacement: m
Present
approach
Ideal
capacities
Fig. 10. Loaddisplacement response of driven pipe pile
SCIENCE AND EMPIRICISM IN PILE FOUNDATION DESIGN 855
depend strongly on any residual stresses locked in at
the pile base at zero displacement.
The large displacements necessary to mobilise a true ulti-
mate end-bearing capacity lead to a form of scale effect in
comparing pile end-bearing and cone resistance, as the
(much greater diameter) pile will react more slowly to
changes in stratigraphy than the cone. To overcome this, it is
essential to average the cone resistance over a number of
diameters above and below the pile base level. Distances
recommended range up to 8 diameters above the pile base
(to allow for the gradual development of residual stresses),
and 2 diameters below the pile base, with a weighting
towards the minimum envelope of the cone prole (Fleming
& Thorburn, 1983). In practice, averaging over a shorter
length, between 1 and 2 diameters above and below the pile
base, is often acceptable, provided there are no strong
stratigraphic changes within the wider range.
Chow (1997) assembled a database of high-quality pile
load tests, and her data for the end-bearing resistance of
closed-ended piles driven into sand are shown in Fig. 12.
The values of cone resistance have been obtained by aver-
aging over 1
.
5d relative to the pile base, and the ultimate
end-bearing resistance, q
bu
, is that mobilised at a pile base
displacement of 0
.
1d. The design curve proposed by Jardine
& Chow (1996) is indicated, and is expressed as
q
bu
q
c
1 0
:
5 log
d
d
cone
_ _
> 0
:
13 (18)
At rst glance, the design curve appears a reasonable t to
the data, in spite of some scatter. However, the data for
small pile diameters are dominated by jacked piles, where
the full cone resistance (appropriately averaged according to
the pile diameter) would be mobilised at each stroke, and
high residual stresses (or at least a high reloading stiffness)
will be retained. An annotated version of the database is
shown in Fig. 13(a), with jacked piles indicated and also a
vibro-driven pile, where the normalised end-bearing capacity
falls below the other data. The driven pile result from the
Akasaka (AK: BCP Committee, 1971: see legend in Fig. 12)
pile tests plots above the jacked pile data, but the reported
loaddisplacement plot (see their gure 9) is anomalous,
with a base resistance that suddenly falls after a displace-
ment of one pile diameter, with a corresponding jump in the
shaft friction (the total load remaining largely unchanged).
Correction for that anomaly would result in a normalised
end-bearing (q
bu
/q
c
) of $0
.
4 for the driven pile.
Load cells or strain gauges in instrumented driven piles
tend to undergo zero shifts during installation, due to
changes in fabrication strains within the pile caused by the
high dynamic stresses. It is therefore usually necessary to
zero strain gauges prior to testing the pile statically, and to
rely on alternative means to estimate any residual base loads.
A common approach is to assume equal shaft capacity in
tension and compression, although this will tend to over-
estimate residual base loads (see later). Correction for
residual loads varies among the pile tests, but examples of
how this may change the deduced end-bearing capacity are
shown in Fig. 13(a) for the Baghdad (BG: Altaee et al.,
1992, 1993) and Hunters Point (HP: Briaud et al., 1989) pile
tests (the arrows linking uncorrected and corrected data).
From the current database, however, and notwithstanding
some inconsistency in respect of allowing for residual base
loads, a design end-bearing capacity of around 0
.
4q
c
, inde-
pendent of diameter, appears reasonable. This may turn out
to be conservative in cases where high residual stresses can
be justied, for example for jacked piles or where the
transient base pressures mobilised during pile driving can be
shown to be a high proportion of q
c
.
The vagaries of data interpretation, and the inevitable
subjectivity involved, are well illustrated in Fig. 13(b), which
shows a recent reinterpretation of the same database by
White (2003). The reinterpretation includes:
q
b
/
q
c
0 1 2 3 4
0
0
.
2
0
.
4
0
.
6
0
.
8
10
Normalised displacement, w
b
/d
0
.
9
0
.
6
0
.
3
q
bo
/q
c
0
.
7
q
bo
/q
c
0
.
3
0
0 0
.
1 0
.
2
E
b
/q
c
5
E
b
/q
c
1
.
25
Fig. 11. Development of end-bearing resistance
0 0
.
2 0
.
4 0
.
6 0
.
8
1.0
Pile diameter: m
Key to individual pile tests
from Chow (1997)
0
0
.
1
0
.
2
0
.
3
0
.
4
0
.
5
0
.
6
0
.
7
0
.
8
0
.
9
1
.
0
q
b
u
/
q
c
Design curve from
Jardine & Chow (1986)
KA (Franki)
KA (Cone)
D (C,d)
DK (C,j)
HP (C,d)
AK (C,j)
S (C,d)
E (O,d)
A (C,d)
G (C,d)
LB (C,j)
HT (C,d)
AK (C,d)
BG (C,d)
Fig. 12. Normalised end-bearing capacities for closed-ended piles from Chow (1997)
856 RANDOLPH
(a) adjustment of design cone resistance values to allow
for partial penetration into a dense sand layer [Kallo
(KA: de Beer et al., 1979); Lower Arrow Lake
(E: McCammon & Golder, 1970)]
(b) correction to include residual base load [Drammen (D:
Gregersen et al., 1973)]
(c) reservations on the quality of the data, such as
estimation of q
c
from SPT data, and insufcient base
displacements to estimate q
bu
.
In the light of these caveats and adjustments, evidence for a
signicant diameter effect is unconvincing, provided appro-
priate averaging of the cone resistance is undertaken and
ultimate base capacity is assessed in terms of relative
displacement (proportion of pile diameter) not absolute dis-
placement. Particular care should be taken in strongly strati-
ed soils, for example where piles are driven through weak
material and penetrate only 1 or 2 diameters into a dense
sand layer. For such cases the design cone resistance needs
to be weighted to reect the overlying weaker material
(Meyerhof, 1976; Meyerhof & Valsangkar, 1977).
Corresponding end-bearing data for open-ended piles are
shown in Fig. 14. There are many fewer data points, and
they are very sparse for diameters in excess of 1 m, which is
the main area of interest for offshore applications. Again
there appears to be a decreasing trend of normalised end-
bearing resistance with increasing pile diameter. However,
scrutiny of the data reveals that:
(a) the data for piles of diameter 1 m (HO: Kusakabe et
al., 1989) and 1
.
2 m (K: Ishihara et al., 1977) are
projected from tests where the base movement was only
$0
.
5% of the pile diameter
(b) the data point for the pile of 2 m diameter (T: Shioi et
al., 1992) has been normalised using a cone resistance
of 35 MPa, whereas the pile tip was very close to the
top of a much softer stratum (see Fig. 15).
Certainly for design a much more conservative value of q
c
would be adopted in this case, possibly as low as 10 MPa.
In order to arrive at an acceptable design approach for
large diameter open-ended piles, it is necessary to consider
the mechanics of the soil plug (Fig. 16). If the soil plug
starts to slip relative to the pile, then the shear stresses
around the plug, which are themselves a function of the
average vertical effective stress in the plug, will lead to an
exponential growth in the vertical stress within the soil plug.
It may be shown (Randolph et al., 1991) that the available
end-bearing resistance may be expressed as
q
bplug
9
vbase
% e
4h
p
=d
i
(19)
where h
p
is the height of the soil plug, d
i
is the internal pile
diameter, and 9
vbase
is the ambient vertical effective stress at
the base of the plug (taken as 9h
p
). As for the external
shaft friction, the ratio
s
=9
v
may be expressed as
Ktan , where is the interface friction angle. Although the

q
b
u
/
q
c
0 0
.
2 0
.
4 0
.
6 0
.
8 1
.
0
Pile diameter: m
0
0
.
1
0
.
2
0
.
3
0
.
4
0
.
5
0
.
6
0
.
7
0
.
8
0
.
9
1
.
0
1
.
1
1
.
2

q
b
u
/
q
c
0 0
.
2 0
.
4 0
.
6 0
.
8 1
.
0
Pile diameter: m
0
0
.
1
0
.
2
0
.
3
0
.
4
0
.
5
0
.
6
0
.
7
0
.
9
1
.
0
Jacked piles
Vibro-driven
Suggested design value
(diameter independent)
Residual load
corrections
Driven pile (suspect data point)
q
c
reassessed (shallow
penetration of sand layer)
Possible clay
layer at base
q
c
estimated from SPT
Zero residual
load observed
(a)
(b)
0
.
8
Fig. 13. Commentary on database of pile end-bearing capacity:
(a) annotated database from Chow (1997); (b) alternative
interpretation of data by White (2003)
0 0
.
5 1
.
0 1
.
5 2
.
0
Pile diameter: m
0
0
.
1
0
.
2
0
.
3
0
.
4
0
.
5
0
.
6
q
b
u
/
q
c
Limited base
movement (0
.
5 % of d)
Overestimated q
c
DK (O,d)
DK (O.d)
CH (O,d)
T (O,d)
G (O,d)
HO (O,d)
K (O,d)
CR (O,d)
Key to individual pile tests
from Chow (1997)
Fig. 14. Normalised end-bearing capacities for open-ended piles from Chow (1997)
SCIENCE AND EMPIRICISM IN PILE FOUNDATION DESIGN 857
value of K may be as high as unity close to the pile tip
(Paik & Lee, 1993; De Nicola & Randolph, 1997; Lehane &
Gavin, 2001), it has been found to decay rapidly along the
length of the soil plug. Minimum values of may be
deduced from Mohrs circle considerations, and lie in the
range 0
.
150
.
2 for typical soil friction angles (Randolph et
al., 1991).
From equation (19), the available end-bearing pressure
rises rapidly with the plug length, so that lengths of only a
few diameters can provide sufcient internal resistance to
ensure plugged failure mode under static loading, regard-
less of the pile diameter. This contrasts with recommenda-
tions of Hight et al. (1996) and Jardine & Chow (1996),
where driven piles of diameter greater than
2(D
r
0
.
3) metres (with D
r
the relative density, expressed
as a fraction) are assumed not to plug.
Part of this divergence of opinion revolves around the
semantics of plugged or unplugged. Here, the term
plugged is restricted to the situation where, during static
loading, the top of the soil plug does not slip relative to the
pile wall. Near the tip, however, relative slip will occur
during loading, owing to compression of the soil plug.
Lehane & Randolph (2002) have considered the separate
response of the soil plug, soil below the pile base, and pile
soil interaction around the annular tip, in order to establish
minimum values for the end-bearing capacity of open-ended
piles in sand. Their recommendations are shown in Fig. 17,
based on conservative assumptions that ignore any increase
in the stress ratio, K, near the pile tip or densication of the
soil within the plug or beneath the pile base (together with
any associated residual stress systems). Even moderate re-
laxation of these assumptions suggests that a design end-
bearing capacity of q
b
/q
c
% 0
.
2 is reasonable, and such a
value is broadly consistent with the database in Fig. 17,
taking account of limitations in the data plotted for the piles
of diameter greater than 1 m. Results of centrifuge model
tests in dense sand also support this as a lower bound design
value (Bruno, 1999; De Nicola & Randolph, 1999).
Shaft friction
Since the pioneering work of Vesic in the 1960s (Vesic,
1967, 1970), it has been realised that in sand and other soils
of high permeability the magnitude of shaft friction at a
given depth can reduce as the pile is driven further, with the
net effect that the average friction along the pile shaft can
reach a limit and even reduce as the pile embedment in-
creases. However, this effect has only recently been quanti-
ed, through the carefully instrumented pile tests undertaken
by the research group at Imperial College. The phenomenon
of friction degradation is illustrated in Fig. 18 (Lehane
et al., 1993), with proles of shaft friction measured in the
three instrument clusters at different distances (h) from the
tip of a pile 6 m long and 0
.
1 m in diameter, as it is jacked
into the ground. For comparison, the cone prole is plotted
on the same scale, but with q
c
factored down by 100. The
shaft friction measured at h/d 4, in particular, follows the
shape of the q
c
prole closely, allowing for differences in
cone and pile diameter. Comparison of the proles from the
instrument clusters at h/d 4 and h/d 25 shows that the
friction measured at the latter position is generally less than
50% of that measured close to the pile tip.
The physical basis for friction degradation is the gradual
densication of soil adjacent to the pile shaft under the
cyclic shearing action of installation. This process is en-
hanced by the presence of crushed particles from the pas-
sage of the pile tip, which gradually migrate through the
matrix of uncrushed material (White & Bolton, 2002). The
far-eld soil acts as a spring, with stiffness proportional to
G/d (where G is the soil shear modulus), so that any
0 10 20 30 40 50 60
TP
24
.
4m
TP
55
.
0m
Pile tip
35 MPa assumed
but could be 10 MPa
40
30
20
10
D
e
p
t
h

b
e
l
o
w

s
e
a
b
e
d
:

m
Fig. 15 Cone resistance prole for Tokyo Bay pile test (after
Shioi et al., 1992)
Soil plug
Pile wall

v

v
d
v

Fig. 16. Equilibrium of soil element within soil plug


Driven piles
w
b
/d 02
w
b
/d 015
w
b
/d 01
04
03
02
01
0
q
b
u
/
q
c
Bored pile (Lee & Salgado, 1999)
0 02 04 06 08 10
Relative density, D
r
Fig. 17. Normalised end-bearing capacity for open-ended piles
(after Lehane & Randolph, 2002)
858 RANDOLPH
densication close to the pile results in reduced radial effec-
tive stress. The operative value of G will be high, as the soil
is heavily overconsolidated having moved through the zone
of high stress close to the pile tip during installation and is
being unloaded.
The incremental volume change, and hence reduction in
radial effective stress, is likely to depend on the current
stress level, with greater changes at higher stress levels. This
suggests an exponential variation of radial stress along the
pile shaft of the form (Randolph et al., 1994)
K
9
r
9
v0


s
9
v0
tan
cv
K
min
(K
max
K
min
)e
h=d
(20)
where K
max
may be taken as a proportion of the normalised
cone resistance, typically 12% of q
c
=9
v0
, and K
min
lies in
the range 0
.
20
.
4, giving a minimum friction ratio,
s
=9
v0
,
of 0
.
10
.
25 (Toolan et al., 1990). The coefcient may be
taken in the region of 0
.
05 for typical pile diameters,
although there are some indications that the value decreases
as the pile diameter increases and vice versa. Indeed, much
higher values of are needed to match data from centrifuge
model tests (Bruno, 1999), although scaling problems related
to the spring stiffness of the surrounding soil may occur for
centrifuge modelling of piles in sand (Fioravante et al.,
1999; Fioravante, 2002).
Other key variables that affect the rate of degradation
include
(a) the unloading modulus of the soil (probably close to
the small strain value, G
0
), with higher unloading
modulus leading to more rapid degradation
(b) the number of shearing cycles per diameter of
advance (or blow count for driven piles).
Assuming that cyclic stress reversal is the major trigger for
compressive volumetric strain in the shearing zone, the rate
of degradation should be very low for continuous jacking
(De Jong & Frost, 2002), and maximum for driven piles
with high blow count. For intermittently jacked piles such as
the Imperial College instrumented pile tests, the rate of
degradation would be intermediate, lower than most driven
piles, although this may be compensated for by scale effects
associated with the small diameter of the pile. As noted by
Fioravante (2002), the unloading stiffness of the surrounding
soil annulus is proportional to G/d, and hence the reduction
in radial stress resulting from any contraction of soil in the
interface layer will be higher for smaller-diameter piles.
The MTD method, derived from the Imperial College eld
studies and database of high-quality pile tests, expresses the
shaft friction for driven piles in sand as

s

q
c
45
9
v0
p
a
_ _
0
:
13
d
h
_ _
0
:
38
9
rd
_ _
tan
cv
(21)
where p
a
is atmospheric pressure (100 kPa) and 9
rd
is a
(relatively small) stress increase due to dilation during load-
ing (Jardine & Chow, 1996). The minimum h/d is taken
conservatively as 4 (in the absence of data at lower h/d
ratios), and for open-ended piles the diameter, d, is replaced
by the equivalent diameter, d
eq
. The method adopts a power
law degradation, rather than an exponential decay, but this
leads to similar shapes of shaft friction proles.
A comparison between the MTD and the present method,
using equation (20) with K
max
0
:
01q
c
=9
v0
, is provided in
Fig. 19, for a 1 m diameter open-ended pile driven 40 m into
sand. The main difference is close to the pile tip, where the
MTD method yields identical values of shaft friction for
open- and closed-ended piles (for h/d
eq
, 4). The present
method gives different maximum values of shaft friction,
dictated by K
max
, and it is suggested that K
max
is increased
to 0
:
015q
c
=9
v0
for closed-ended piles in view of the higher
normalised end-bearing resistance. The shaft friction ratio
between open and closed-ended piles implied by the two
methods is quite similar, with an average ratio of around
0
.
7, although the MTD method gives a ratio that decreases
from unity at the pile tip down to (d
eq
/d)
0
:
38
(typically
$0
.
65) whereas the present method gives an increasing ratio
as K approaches K
min
for both geometries. The average ratio
of 0
.
7 may be compared with the API (1993) design
recommendation of 0
.
8, but also with recent experimental
studies that show a much lower ratio of just under 0
.
5 (Paik
et al., 2003).
Shaft friction in tension and compression
The tensile capacity of piles in sand has been found to be
less than the shaft capacity measured in compression, and
most design guidelines include a reduction of 1030% to
allow for this (API, 1993). Two factors were identied by
De Nicola & Randolph (1993) that contributed to lower
tensile shaft friction: the rst was a reduction in effective
stress levels adjacent to the pile compared with loading in
compression (even for a rigid pile), and the second was the
Poissons ratio reduction in diameter (and consequential
Local shear stress: kPa
0 10 20 30 40 50 60
0
1
2
3
4
5
6
D
e
p
t
h

o
f

i
n
s
t
r
u
m
e
n
t
:

m
Cone resistance
q
c
/100
h/d 25
h/d 14
h/d 4
Fig. 18. Measured proles of shaft friction (Lehane et al., 1993)
Shaft friction: kPa
0 100 200 300 400
0
5
10
15
20
25
30
35
40
45
D
e
p
t
h
:

m
MTD method
Present method:
003, 005 and 007
Open-ended pile: L 40 m, d 1 m, 0.1
q
c
5 1z MPa, 11 kN/m
3
,
cv
28
K
max
001q
c
/
v0
, K
min
03
Fig. 19. Example proles of shaft friction for driven pile in sand
SCIENCE AND EMPIRICISM IN PILE FOUNDATION DESIGN 859
reduction in radial effective stress). These two effects were
quantied for piles fully embedded in sand, by the expres-
sion
(Q
s
)
tens
(Q
s
)
comp
% 1 0
:
2 log
10
100
L=d
_ _ _ _
(1 8 25
2
) (22)
where Q
s
is the shaft capacity and
p
(L/d)(G
ave
/E
p
)tan ,
with G
ave
, E
p
and
p
being respectively the average soil
shear modulus, Youngs modulus of an equivalent solid pile
and Poissons ratio for the pile.
The two factors that contribute to reduced tensile capacity
tend to compensate as the pile aspect ratio increases, with
the average change in effective stress level decreasing and
the effect of Poissons ratio contraction increasing. This is
shown in Fig. 20 where, for a typical modulus ratio of
E
p
/G
ave
400, the shaft capacity ratio is $0
.
8 for a range
of L/d. Even for quite wide extremes of E
p
/G
ave
, the shaft
capacity ratio remains within 0
.
70
.
85.
Although other effects, such as local stress changes due to
dilation, will inuence the shaft capacity ratio, the expres-
sion in equation (22) provides a reasonable design basis for
assessing the reduced shaft capacity for loading in tension,
compared with that for loading in compression.
Euripides pile test
A major eld investigation of driven pile capacity in
dense sand, the Euripides pile test, was undertaken in the
1990s, funded by a number of companies operating in the
offshore area and managed by Fugro BV. The data are now
in the public domain and are in the process of being
published: a brief comparison of measured and predicted
pile resistance is presented here. The pile test comprised a
heavily instrumented pile, 0
.
76 m in diameter and 35 mm
wall thickness, driven open-ended into very dense sand.
Cone resistance proles are shown in Fig. 21, and it may be
seen that, in spite of some variability in the cone resistance
below 30 m, the average q
c
rises to between 60 and 70 MPa.
The cone resistance over the upper 22 m is very low, and
most of the test pile capacity was generated below that
level.
The pile resistance mobilised in the dense sand has been
estimated using the MTD method, and also that presented
here, adopting the design q
c
prole shown in Fig. 21. The
MTD method gives an end-bearing ratio of q
b
/q
c
0
.
17,
whereas a ratio of 0
.
2 has been adopted for the alternative
method. Values of K
max
, K
min
and have been taken as
0
:
01q
c
=9
v0
, 0
.
3 and 0
.
05 respectively, whereas the value of

cv
measured in ring shear tests was 308.
The resulting proles of pile resistance are shown in Fig.
22, compared with the measured loads at a pile displacement
of 0
.
1d. The test pile was initially driven to depths of
30
.
5 m, 38
.
7 m and 47
.
0 m, with compression and tension
tests being conducted at each level. The pile was then
extracted and driven without pause to 46
.
7 m penetration,
after repairing some of the damaged instrumentation. Over-
all, the agreement between either prediction method and the
test data appears reasonable over the range where the
instrumentation survived the driving process. However,
the measurements show very low friction mobilised over the
depth range 2230 m, and a need is indicated for some
renement of the average cone resistance (possibly averaging
over a greater distance above pile base level). At greater
depths, the MTD and exponential decay methods give shaft
friction values (or gradients of axial load) that lie respec-
tively slightly above and below the measurements. Predic-
tions of end-bearing resistance appear reasonable, with close
agreement between the MTD method and measured base
load at the intermediate depth.
The average shaft friction ratio between tensile and com-
pressive load tests ranged between 0
.
6 and 0
.
9, but with no
clear pattern among the four separate load test depths.
Further assessment of any residual loads (or load cell zeros)
may lead to some revision of these estimates, but the range
spans that shown in Fig. 20 based on the approach of De
Nicola & Randolph (1993).
09
08
07
06
05
0 10 20 30 40 50 60
Pile aspect ratio, L/d
S
h
a
f
t

c
a
p
a
c
i
t
y

r
a
t
i
o
400
200
E
p
/G
ave
800
tan = 05

p
= 03
Fig. 20. Ratio of shaft capacity in tension and compression (De
Nicola & Randolph, 1993)
55
50
45
40
35
30
25
20
0 20 40 60 80 100
D
e
p
t
h
:

m
Simplified design profile
Cone resistance, q
c
: MPa
Fig. 21. Cone resistance proles from Euripides site
50
40
30
20
10
0
0 5 10 15 20
Field data: location 1
Field data: location 2
MTD method
Exponential decay: 0
.
05
Pile load: compression: MN
D
e
p
t
h
:

m
Fig. 22. Measured and calculated load distributions for Eur-
ipides pile tests
860 RANDOLPH
Summary
We now have a much clearer picture of changes that
occur in the soil around piles driven into sand, even though
our design calculations rely heavily on empirical correla-
tions. Use of the cone resistance provides a better quantita-
tive basis for these correlations, but we need to review how
best to average q
c
, taking account of the pile diameter, the
displacement needed to reach steady-state conditions, and
the soil stratigraphy. The effect of pile diameter on design
end-bearing resistance, or on the plugging of pipe piles, is
an area of apparent divergence between science and empiri-
cism, which needs to be resolved. Analysis of the soil plug
response suggests that the (static) end-bearing resistance of
pipe piles that do not plug during driving may be taken
conservatively as about half that for a comparable closed-
ended pile, but that design values of q
bu
/q
c
may be strongly
affected by the magnitude of residual base pressure, q
b0
, or
the extent to which the base response has been pre-stiffened
during the installation process.
Modern design methods must take account of friction
degradation, but further work is needed in order to explore
how the rate of degradation is affected by pile diameter,
method of installation (particularly blow counts during driv-
ing), and soil modulus.
The design approaches considered here are conservative in
two respects. The true ultimate base resistance will exceed
the design value based on a limited displacement, with the
average end-bearing resistance ultimately approaching the
cone resistance. Secondly, recent studies have suggested that
the shaft friction of piles in sand shows signicant increase
with time (Chow et al., 1998), with gains of 50100%
possible. Although the resulting shaft friction may prove
somewhat brittle, and so should not be considered in con-
junction with large displacements to mobilise the base
resistance, further understanding and quantication of this
phenomenon would be valuable.
DYNAMIC PILE TESTING
Overview
Load tests to verify capacity are an essential part of most
piling contracts, reecting the relatively high level of uncer-
tainty in predictive methods. Traditional static load tests,
using kentledge or reaction systems, undoubtedly provide the
most precise method of evaluating the loaddisplacement
response, with minor limitations in terms of interaction
(mainly affecting the pile stiffness) and loading rate effects.
However, static pile tests are relatively expensive, and also
give limited information on the distribution of resistance
along the pile unless it is instrumented. For large diameter
cast-in situ piles, external reaction systems become prohibi-
tively expensive, and alternative devices such as Osterberg
cells (Osterberg, 1989), which use part of the pile itself to
provide reaction, offer a more effective means of measuring
capacity. An example application of this technique is pre-
sented later.
Dynamic pile tests, using high-energy piling hammers,
provide an alternative to static loading tests, at a cost that is
typically two orders of magnitude lower, but require sophis-
ticated numerical simulation of the measured stress waves in
order to interpret the test. Alternative methods of testing, at
intermediate loading rates (and costs) have appeared over
the last decade, in particular the Statnamic test (Berming-
ham & Janes, 1989), in which a fast-burning fuel is used to
accelerate a mass away from the pile, thereby loading it in
reaction. Reviews of various alternative methods for pile
load tests have been given by England & Fleming (1994),
who focus on different procedures for static load tests, and
Poulos (1998), who compares the relative merits of static,
Statnamic and dynamic tests
Further discussion here is conned to the interpretation of
dynamic pile tests, which have the potential to be extremely
cost-effective, but where commercial practice has fallen
behind signicant advances in modelling the dynamic inter-
action between pile and soil. This has contributed to a
growing scepticism regarding the condence that may be
placed in such tests. The discussion will focus on modelling
the inertial (or radiation) damping from the soil surrounding
the pile, and on treatment of open-ended piles, but a brief
description of the main principles is included rst.
A schematic diagram of the main features of dynamic pile
testing is shown in Fig. 23. The test comprises monitoring
of strain and acceleration by means of instruments attached
to the pile shaft above ground, and preferably at least two
diameters below the pile head, during blows applied by
standard piledriving hammers or large drop-weights. Nor-
mally pairs of instruments are located at opposite sides of
the pile in order to minimise effects of bending, and the raw
data are converted to force, F (by multiplying the measured
strain by the cross-sectional rigidity, EA, of the pile) and
velocity, v (by integrating the accelerometer data).
Initially, the stress wave travels down the pile unimpeded,
during which time the measured force and velocity are
directly proportional to each other: that is, F Zv, where Z
is the pile impedance, EA/c, with c being the wave speed in
the pile. Any resistance to movement of the pile, for
example due to shaft friction, or change in pile section
(more precisely, impedance) including at the pile tip, will
cause an upward-travelling wave to be propagated back up
towards the pile head. Monitoring two properties (force and
velocity) allows separation of downward (subscript, d) and
upward (subscript, u) components of the stress wave, accord-
ing to
F
d
0
:
5(F Z)
F
u
0
:
5(F Z) (23)
It may be shown that, for sufcient magnitude of dynamic
force, the total soil resistance encountered by the pile during
the passage of the stress wave down and up the pile is equal
to the algebraic sum of the initial downward-travelling
(maximum) force plus the upward-travelling force that ar-
rives back at the instrumentation at a time 2L/c after the
initial maximum (Rausche et al., 1985; Randolph, 1990).
Simplied methods have been proposed for estimating the
static component of the total soil resistance mobilised during
a dynamic test. However, the most reliable approach to
analyse dynamic tests is by computer simulation of the
Instrumentation:
accelerometers
strain cells
0
2h/c 2L/c
Time
Reflections
from shaft
resistance
Wave
speed,
c
1
Downward
travelling
Depth
Pile
impedance:
Z EA/c
Shock
wave
L
h
v
0
Upward
travelling
(reflection
from base)
Fig. 23. Schematic of stress wave travel down pile
SCIENCE AND EMPIRICISM IN PILE FOUNDATION DESIGN 861
pilesoil interaction, endeavouring to match computed and
measured stress-wave signals. The computer model is
the dynamic equivalent of the load transfer approach for
static (non-linear) analysis, with the pile modelled as a one-
dimensional elastic column and static load transfer springs
replaced by an appropriate dynamic model.
Dynamic pilesoil interaction models
Worldwide, the majority of dynamic pile tests are ana-
lysed using pilesoil interaction models based on Smith
(1960), with the soil resistance expressed as
Min 1,
w
p
Q
_ _
(1 J)
s
(24)
where w
p
is the local pile displacement, and
s
are the
dynamic and static shaft friction values (or equivalent in
end-bearing pressure), Q is the displacement, or quake, to
mobilise the limiting static resistance, and J is a damping
parameter multiplying the local pile velocity, v.
In the Smith model, separate viscous and inertial effects
are lumped into a single parameter, and the model does not
differentiate properly between conditions prior to pilesoil
slip and those after slip. A more rational model is shown in
Fig. 24, as proposed by various researchers in the 1980s
(Simons & Randolph, 1985; Lee et al., 1988) and subse-
quently implemented in many research-oriented, but few
commercial, codes. In what will be referred to here as the
continuum model, the far-eld soil response may be mod-
elled accurately using an elastic spring in parallel with a
dashpot representing inertial damping (Novak, 1977), with
the shaft response expressed as
% G
w
d

s
_ _
<
lim
(25)
where v
s
is the shear wave velocity in the soil. Here, w and
v are the displacement and velocity of the soil immediately
adjacent to the pilesoil interface, rather than the pile,
which is an important distinction.
The interface itself is modelled using a limiting shaft
friction that is velocity dependent, such as

lim

s
1 m
v

0
_ _
n
_ _
(26)
where m and n are viscous parameters and v is the relative
(or slip) velocity between pile and soil, normalised by v
0
(taken for convenience as 1 m/s). Studies by Litkouhi &
Poskitt (1980) suggest that the exponent n typically lies in
the range 0
.
20
.
5, and m is 0
.
30
.
5 for sand, and as high as
2 or 3 for clays.
A typical response of the model described by equations
(25) and (26) is shown in Fig. 25(a), compared with a
corresponding result for the Smith (1960) model in (b)
note that both results are from the example application
described later. Various features are evident from this com-
parison:
(a) The initial response of the continuum model is
extremely stiff, and dominated by the inertial dashpot
term in equation (25). As noted by Randolph & Deeks
(1992), for typical acceleration levels the dashpot term
will be at least a factor of 10 larger than the spring
term during the initial response, and here slip starts to
occur after a pile displacement of 0
.
3 mm, when the
dashpot accounts for 97% of the resistance.
(b) By contrast, the Smith model with a standard quake
of 2
.
5 mm and damping value, J, of 0
.
2 s/m shows a
much more gradual development of resistance: the
immediate consequence of this is that the prole of
shaft friction deduced using the continuum model will
be offset to signicantly greater depths, compared with
the Smith model (as the pile will take 13 ms to move
2
.
5 mm, during which time the stress wave will have
travelled 515 m).
(c) The dynamic amplication of the static shaft friction is
affected by the velocity exponent, n (here taken as 0
.
2
in the continuum model, and unity in the Smith model);
a non-linear version of the Smith model is straightfor-
Pilesoil interface
Far-field
response
Plastic
slider
Elastic
spring
Pile node
Viscous
dashpot
Inertial
dashpot
Fig. 24. Model for dynamic pilesoil interaction along shaft
S
h
e
a
r

s
t
r
e
s
s
,

s
Total
Static limit
Inertial dashpot
Elastic spring
Viscous dashpot
0 5 10 15 20 25
Pile displacement: mm
(a)
Total
Static response
Viscous
component
0 5 10 15 20 25
Pile displacement: mm
(b)
1
.
5
1
.
0
0
.
5
0
.
5
1
.
0
1
.
5
1
.
5
1
.
0
0
.
5
0
0
.
5
1
.
0
1
.
5
2
.
5
2
.
0
0
S
h
e
a
r

s
t
r
e
s
s
,

s
Fig. 25. Example responses of (a) continuum and (b) Smith
shaft models
862 RANDOLPH
ward to implement, although it is rarely adopted. Even
so, the continuum model tends to give a more sustained
maximum resistance, as the spring component continues
to rise as the inertial component decreases, whereas the
formulation of the Smith model automatically gives a
decline in resistance as the pile velocity decreases.
A similar continuum model exists for the dynamic inter-
action at the base of the pile (Deeks & Randolph, 1995),
and again the inertial component dominates the initial
response, although not to quite to the same extent as for the
shaft (Randolph, 2000).
Open-ended piles
The driving response of open-ended piles is usually
evaluated by lumping internal and external friction together
(as external friction) and considering the end-bearing resis-
tance on the annular base. However, the response of the soil
plug is very different from that of the external soil, as
energy cannot be propagated to the far-eld. A way of
modelling the soil plug, as a column of nodal masses
interconnected by damped springs, and its interaction with
the pile was rst proposed by Heerema & de Jong (1979)
and extended using continuum-style soil models by
Randolph (1987), as shown in Fig. 26. Recent studies by
Liyanapathirana et al. (2001) have shown that this approach
works well for conditions where the incremental lling ratio,
dened as the ratio of incremental plug movement within
the pile, to the penetration of the pile (Brucy et al., 1991)
exceeds 5070%.
Example application
Examples of dynamic testing of steel pipe piles are
presented here, drawn from a recent project in Perth, Wes-
tern Australia, involving duplication of the main freeway
bridge across the Swan River. Fig. 27 shows a typical
geometry of the steel pipe piles, with diameter of 0
.
61 m
and wall thickness of 12
.
5 mm. In most piles, an annular
diaphragm was incorporated a few metres back from the pile
tip, in order to encourage the development of a fully
plugged condition. Without the diaphragm, it was found that
the piles could be driven to large penetrations with relatively
low energy, and although the pile would undoubtedly have
responded in a plugged manner to static loading, it was not
possible to verify the plugged capacity dynamically. The soil
stratigraphy for the pile groups supporting the approach
piers and abutments is indicated in Fig. 27: it comprised
sand ll overlying alluvial soft clays and sand lenses above
a dense sand base in which the piles were to be founded.
Matching of computed and measured stress waves can be
achieved by using one of the signals (such as the force) as
input, and matching the other signal (velocity). However, a
better approach is to combine the signals together in the
manner of equation (23), using the downward-travelling
wave (the externally applied force) as input, and matching
the upward-travelling waves (due to soil resistance and tip
reection).
Figure 28 shows examples of matching the upward-travel-
ling force wave using three approaches:
(a) explicit modelling of the soil plug using the continuum
soil model
(b) internal and external shaft friction lumped together,
using the continuum soil model
(c) internal and external shaft friction lumped together,
using the Smith (1960) model.
All results are for identical distribution of friction down the
inside and outside of the pile, and identical base resistance
on the pile annulus. It is clear that lumping the internal shaft
friction with the external friction has had a signicant effect
on the shape of the computed stress wave, even using an
identical continuum model. Also, the response computed
using the Smith (1960) model is very different from that
using the continuum model. In fact, for this particular open-
ended pile test, no good match could be obtained without
explicit modelling of the soil plug, even by varying the
distribution of soil resistance.
The best (although not particularly close) t to the
Shaft
Soil response
model
Soil plug
Base
Base response
model
Pile wall
External
soil
Fig. 26. Modelling dynamic response of soil plug
0
.
61 m
Sand
fill
Soft
clay
Dense
sand
6m
33 m
Steel
pipe
pile
Annular steel
diaphragm
Soil
plug
Fig. 27. Partially plugged pipe pile
SCIENCE AND EMPIRICISM IN PILE FOUNDATION DESIGN 863
measured upward-travelling wave in Fig. 28 was obtained
with the capacities summarised in Table 2 (Pile 1). The
quality of t was relatively sensitive to the internal shaft
friction, and reasonable ts were obtained for mobilised
resistances in the range 6
.
57
.
5 MN. The static response of
the pile was evaluated by lumping the internal shaft capacity
into the base resistance, and assuming that the resulting base
resistance of 4
.
91 MN would be mobilised at a displacement
of 10% of the pile diameter (0
.
06 m). The distribution of
external shaft friction (and shear modulus) deduced from the
dynamic test was incorporated unaltered into the static
analysis. The resulting pile response is compared with the
results of a static load test on the pile in Fig. 29 and shows
surprisingly good agreement considering the rather arbitrary
manner of incorporating the dynamic soil plug resistance in
with the base response.
One further example of predicting the static pile response
from a dynamic test is shown in Fig. 30, for a tension test
carried out as part of the same project. In this case the
open-ended pile was driven (with no internal diaphragm)
through the upper sand ll to tip into the underlying clay,
and a plug length of 10 m (compared with total pile penetra-
tion of 16
.
5 m) was measured. The stress-wave matching has
been discussed by Randolph (2000), and the resulting capa-
cities are summarised in Table 2.
The computed response under (static) tensile loading,
taking just the external shaft resistance, is compared with
the measured response in Fig. 30. The measured capacity in
tension is about 15% lower than the external shaft capacity
deduced from the dynamic test, which is consistent with the
ratio of tensile and compressive shaft capacities from Fig.
20. However, it should also be pointed out that the level of
condence in the back-analysed external shaft capacity is
poor, with a possible range of 11
.
5 MN (mainly by redis-
tributing internal friction to external friction).
Summary
Dynamic pile tests are arguably the most cost-effective of
all pile-testing methods, although they rely on relatively
sophisticated numerical modelling for back-analysis. Theor-
etical advances in modelling the dynamic pilesoil inter-
action have been available since the mid-1980s, but have
been slow to be implemented by commercial codes, most of
which still use the empirical parameters of the Smith (1960)
model. In order to allow an appropriate level of condence
in the interpretation of dynamic pile tests, and estimation of
the static response, it is high time that appropriate scientic
models were used for pilesoil interaction, including explicit
modelling of the soil plug for open-ended piles.
There will still be a need for some empiricism, mainly in
adjusting the static shaft friction to allow for the very high
shearing rates during a dynamic test. Modelling viscous and
Computed
Smith model
closed-ended
Computed
continuum model
closed-ended
Computed
open-ended
Measured
Time: ms
10 20 30 40 0
1
.
0
0
.
5
0
0
.
5
1
.
0
1
.
5
2
.
0
2
.
5
3
.
0
F
o
r
c
e
:

M
N
Fig. 28. Matching of measured and computed upward-travelling waves
Table 2. Summary of mobilised soil resistances from dynamic tests
External shaft
resistance:
MN
Internal shaft
resistance:
MN
Annular base
resistance:
MN
Total pile
resistance:
MN
Pile 1 (compression) 2
.
43 3
.
97 0
.
94 7
.
34
Pile 2 (tension) 1
.
02 0
.
98 0
.
06 2
.
06
Computed
Range of reasonable
fits to dynamic test
Measured
0 20 40 60 80 100 120
Displacement: mm
0
1
2
3
4
5
6
7
8
F
o
r
c
e
:

M
N
Fig. 29. Comparison of measured and computed pile response
864 RANDOLPH
other effects needs to be placed on a more rational basis, as
approaches such as equation (26) are far from ideal. Finally,
it should be realised that the deduced soil resistance para-
meters will not be unique. It is therefore important when
back-analysing dynamic pile tests to adopt appropriately
conservative values (or distributions) of soil resistance that
still give a reasonable t to the stress wave data, rather than
necessarily adopting the best t set of parameters. Given
the empirical nature of the soil models, the concept of best
t is of limited value, and a philosophy of conservative,
while still consistent is a more prudent approach.
DESIGN OF PILE GROUPS
Design philosophy
The main objective of pile group design is to ensure that
the foundation does not undergo excessive displacements
during its design life. This principle is represented by the
serviceability limit state in design codes, but in addition
there is normally a requirement to ensure that there is an
adequate material factor against collapse, or ultimate limit
state. Where piles exhibit strain-hardening, the ultimate limit
state for an individual pile is customarily (but somewhat
arbitrarily) taken as the load to cause either an absolute
displacement (such as 50 or 100 mm), or a relative displace-
ment (such as 10% or 15% of the pile diameter).
In practice, however, for most applications the real ulti-
mate limit state should be determined by considering the
interaction of the foundation with the superstructure, and
must consider the complete foundation system (pile group,
with or without ground-contacting cap) rather than indivi-
dual piles. As for serviceability limit states, the ultimate
limit state then also reverts to a displacement criterion, but
one that is determined largely by structural considerations.
The advantage of moving more towards displacement
criteria, reducing the emphasis on the capacity of individual
piles, is illustrated in Fig. 31: axial pile capacity depends
critically on effective stress and fabric conditions at the
pilesoil interface, which are difcult to determine accu-
rately, whereas the deformation response is inuenced pri-
marily by soil conditions away from the pile, which are only
slightly affected by the installation process. Simple, but
robust, analytical approaches for estimating the vertical
stiffness of piles and pile groups are readily available, as
outlined below. Hence, provided the in situ stiffness of the
soil strata can be determined, the pile group loaddisplace-
ment response up to working load can be estimated with
greater accuracy than the ultimate capacity.
The discussion here will be restricted to vertically loaded
pile groups, and the trend towards piled rafts where raft
foundations are supported by relatively low numbers of piles,
placed strategically in order to minimise differential settle-
ments and bending moments in the raft. However, a case
study involving foundations subjected to general vertical,
horizontal and moment loading will also be presented,
illustrating the importance of considering the overall founda-
tion response, rather than safety factors on individual piles.
Analysis of vertically loaded pile groups
Many different approaches may be used to analyse the
response of single piles and pile groups subjected to vertical
loading. The most widespread, and most rigorous for homo-
geneous soil conditions, is the boundary element method
(Poulos, 1968; Banerjee & Buttereld, 1981; Basile, 1999),
which has the advantage that only the pilesoil interfaces
need to discretised, rather than the full continuum. A much
simpler approach, leading to analytical solutions, is to adopt
a Winkler approximation, where values of shear stress down
the pile shaft, and end-bearing pressure at the base, are
taken as proportional to the local pile displacement. This
approach has been shown to lead to good agreement with
results from more rigorous numerical analysis (Randolph &
Wroth, 1978), and recent analytical studies have evaluated
the relationship between Winkler spring constant and elastic
shear modulus in closed form (Mylonakis, 2001).
The original solution of Randolph & Wroth (1978)
adopted the idealised assumption of shear stresses around
the pile that decayed inversely with radius, leading to a
logarithmic decay in vertical displacements, together with a
maximum radius of inuence, r
m
. The shear stress at the
pile wall is then expressed in terms of the local displace-
ment, w, as

2Gw
d
(27)
where
ln
2r
m
d
_ _
The Winkler spring constant, k, relating the force per unit
length transferred from pile to soil to the local displacement,
is then given by
k
d
w

2G

G (28)
The value of is typically around 1
.
5 for oating piles, but
increases for end-bearing piles (with w
b
0) where it
0 10
20
30
40
50
Measured
Computed
Range of reasonable
fit for external shaft
friction from
dynamic test
165 m
061 m
10 m
Soil
plug
F
o
r
c
e
:

M
N
0
0
.
2
0
.
4
0
.
6
0
.
8
1
.
0
1
.
2
1
.
4
1
.
6
Steel
pipe
pile
Displacement: mm
Fig. 30. Comparison of measured and computed pile response
in tension
(a) (b)
Fig. 31 Pile group capacity and stiffness: (a) capacity dependent
on conditions at pilesoil interface; (b) stiffness determined
primarily by far-eld conditions
SCIENCE AND EMPIRICISM IN PILE FOUNDATION DESIGN 865
varies with aspect ratio, L/d, and stiffness ratio, E
p
/G,
according to (Mylonakis, 2001)

k
G
% 1
:
3
2(1 )G
E
p
_ _
0
:
025
1 7
L
d
_ _
0
:
6
_ _
(29)
The resulting pile head stiffness, K P/w
t
, where P and w
t
are the pile-head load and ground-level displacement respec-
tively, may then be expressed as
K E
p
A
p

tanh(L)
1 tanh(L)
(30)
The parameters and L represent non-dimensional base
stiffness and slenderness ratio for the pile, expressed as

P
b
w
b
E
p
A
p

and L

k
E
p
A
p

L (31)
where P
b
and w
b
are respectively the load and displacement
at the pile base, and E
p
A
p
is the cross-sectional rigidity of
the pile.
The above expression may be used to calculate the
stiffness of single piles in homogeneous soil, or in multi-
layered soil (essentially using equation (30) and the asso-
ciated value of as a transfer function; Mylonakis &
Gazetas, 1998). For soils where the stiffness varies with
depth, alternative solutions may be developed, involving
Bessel functions of fractional order (Guo & Randolph,
1997). Such solutions are readily evaluated using modern
mathematical packages.
For pile groups, the stiffness of each pile is reduced
because of interaction effects, as indicated in Fig. 32.
Mylonakis & Gazetas (1998) demonstrated that the inter-
action factor, (as dened by Poulos, 1968), must reect
not only the (assumed) logarithmic decay in displacements,
but also the reinforcing effect of the neighbouring pile. This
leads to a reduction in the interaction factor, as indicated in
Fig. 33. For piles of the same length and diameter, the
interaction factor for a given spacing, s, may then be ex-
pressed as the product of two terms representing the loga-
rithmic decay and a diffraction factor, (Mylonakis &
Gazetas, 1998), giving

ln(r
m
=s)
ln(2r
m
=d)
_ _
(32)
where the diffraction factor, , is a function of and L,
according to

2L sinh(2L)

2
[sinh(2L) 2L] 2[cosh(2L) 1]
2 sinh(2L) 2
2
sinh(2L) 4cosh(2L)
(33)
This elegant capturing of the reinforcing effect of each pile
in a group is a seminal advance in quantifying interaction
between piles. The form of equation (33) is illustrated in
Fig. 34, where it may be seen that converges to 0
.
5 for
long (or compressible) piles, is less than 0
.
5 for end-bearing
piles (high ), and is between 0
.
5 and 1 for most oating
piles (low ).
An equivalent expression to equation (33) may be derived
for piles of the same length, but different diameters d
1
and
d
2
(and corresponding values of , and L), leading to an
interaction factor,
21
(proportional settlement of Pile 2
compared with Pile 1), given by

21

21

2
2

2
1

sh
1

1
ch
1
sh
2

2
ch
2
_
_

(
1
tanh
1
L)
2
sh
2
ch
2

1
sh
1
ch
1
_ _
(1
1
tanh
1
L)(sh
2

2
ch
2
)
_

_
(34)
P
w
1t
w
1t
1 2
S
d
L
Fig. 32. Interaction between two piles for axial loading
0 02 04 06 08 10
0
.
376 0
.
578
Pile 2
adjusted
Pile 2
logarithmic
decay
Pile 1
L/d 20
E
p
/G 500
0
.
3
s/d 3
0
0
.
2
0
.
4
0
.
6
0
.
8
1
.
0
Normalised displacement, w/w
1o
D
e
p
t
h
,

z
/
L
Fig. 33. Displacement proles down loaded and adjacent piles
0 0
.
5 1
.
0 1
.
5 2
.
0 2
.
5
0
0
.
05
0
.
2
1
0
.
1
1
.
0
0
.
8
0
.
6
0
.
4
0
.
2
0
Dimensionless pile length, L
End-bearing pile
D
i
f
f
r
a
c
t
i
o
n

f
a
c
t
o
r
,

Fig. 34. Diffraction factor, (Mylonakis & Gazetas, 1998)


866 RANDOLPH
where

21

ln(r
m
=s)

1
; sh
i
sinh
i
L and ch
i
cosh
i
L (35)
It may be shown that the reciprocal theorem is satised by
the above expression, as
21
/K
1
is identical to
12
/K
2
, and
that the above expression reduces to that from Mylonakis &
Gazetas (1998) as the pile diameters converge.
The effect of non-homogeneous soil conditions may be
incorporated in an approximate manner, following Randolph
& Wroth (1978), whereby the load transfer parameter, , is
reduced by ln(r) (where r G
average
/G
L
, the ratio of
average shear modulus to the value at pile tip level, ranging
between 0
.
5 and 1), and also the tanh(L) term in the
numerator of equation (30) is multiplied by r.
The above approach has been used to evaluate the stiff-
ness of square groups of piles, from 2 3 2 up to 30 3 30,
for L/d 25, E
p
/G
L
1000, r 0
.
75 and 0
.
3. The
results are shown in Fig. 35(a), where the pile group
stiffness, K
g
(ratio of total applied load to average settle-
ment), has been normalised by G
L
B, where B is the width of
the pile group. Plotting the normalised stiffness against the
normalised width, B/L, leads to an envelope of curves that
tends to the stiffness of a surface raft as B/L becomes large.
The stiffness envelope may also be matched closely by using
an equivalent pier approximation of the pile group (Poulos
& Davis, 1980; Randolph, 1994), demonstrating the robust-
ness of calculations of pile group stiffness even with quite
approximate models (Fig. 35(b)).
From a practical viewpoint, it is also useful to link the
average settlement of a pile group to the dimension, B, of
the foundation and the factor of safety against ultimate
capacity. This has been explored by Cooke (1986) and more
recently by Mandolini (2003), who suggest that for ne-
grained soils the average settlement is typically around 0
.
3
0
.
6% of the foundation width, B, for a factor of safety of 3.
Piled rafts
As may be seen from Fig. 35, for pile groups where
B/L , 1 the pile group stiffness is signicantly greater than
the stiffness of a raft foundation. Therefore, even if the pile
cap rests directly on competent ground, it will contribute
little to the response of the overall foundation. Viggiani
(2001) has referred to such foundations as small pile
groups, where piles are needed to ensure adequate bearing
capacity, and the pile cap (or raft) can easily be made
sufciently stiff to eliminate differential settlements. By
contrast, for large pile groups, with B/L . 1, the pile cap
will often provide sufcient margin against bearing failure,
and will contribute signicantly in terms of transferring load
directly to the ground. The design of such foundations
hinges more on limiting the average and differential settle-
ments to a acceptable level. As, for large rafts, the exural
stiffness will be low, the location and length of any pile
support should be chosen in order to minimise differential
settlements. Design strategies for small and large pile
groups have been discussed by Viggiani (2001) and
Mandolini (2003).
The concept of using a limited number of piles, poten-
tially loaded to near their ultimate capacity, beneath a raft
foundation in order to reduce differential settlements was
mooted by Burland et al. (1977), and also explored further
by Randolph (1994) and Horikoshi & Randolph (1998). The
principle behind the design approach is illustrated in Fig. 36.
For large rafts (B/L . 1) the stiffness of a pile group
occupying the full area of the raft will be quite similar to
that of the raft alone, and a more effective approach to
reducing differential settlements is to place a few piles over
the central region of the raft. Where a layer of soft soil
exists beneath the raft, then it may be necessary to install
short piles extending through that layer over the full raft
area, but then to use longer piles in the central 2540% of
the raft area.
Piled rafts of this nature require careful analysis, but there
are a variety of analytical approaches of varying complexity
that are now available (Franke et al., 1994; Poulos, 1994,
2001; Clancy & Randolph, 1996; Russo, 1998; Katzenbach
et al., 2000). Studies on optimising pile geometry have
mainly been restricted to uniform loading of the raft, and
L/d 25
E
p
/G
L
1000
075
0
.
3
s/d 2
s/d 3
s/d 5
s/d 10
Raft stiffness
G
r
o
u
p

s
t
i
f
f
n
e
s
s
,

K
g
/
G
L
B
0 1 10
Normalised width of pile group, B/L
0
2
4
6
8
10
12
14
16
18
20
G
r
o
u
p

s
t
i
f
f
n
e
s
s
,

K
g
/
G
L
B
01 1 10
Normalised width of pile group, B/L
0
2
4
6
8
10
12
14
16
18
20
Stiffness of
incompressible pier
80% of stiffness of
incompressible pier
Equivalent pier
(same area and
length as pile group)
(a)
(b)
Fig. 35. Comparison of pile group and equivalent pier stiff-
nesses: (a) normalised stiffness of pile groups; (b) comparison of
equivalent pier stiffness
Piled foundation Raft foundation Piled raft foundation
Fig. 36 Transition from raft to pile group to piled raft
foundation
SCIENCE AND EMPIRICISM IN PILE FOUNDATION DESIGN 867
these have suggested that piles of length greater than about
70% of the width of the raft are required, situated over the
central 2540% of the raft area (Horikoshi & Randolph,
1998; Prakoso & Kulhawy, 2001; Viggiani, 2001).
In practice, buildings often concentrate a signicant pro-
portion of the total load towards the outer edges of the raft,
giving rise to concern over potential hogging of the raft. In
order to address that concern, a recent study by Reul &
Randolph (2004) has also considered the case where half of
the applied load is distributed uniformly over the central
25% of the raft, and the other half is transmitted around the
raft edge. Fig. 37 shows the raft and pile geometry analysed,
together with the two alternative load distributions.
Horikoshi & Randolph (1997) proposed a raftsoil stiffness
ratio dened as
K
rs
5
:
57
E
r
2G
1
1
2
r

B
r
L
r
_
t
r
L
r
_ _
3
(36)
where the subscript r denotes the raft properties of Youngs
modulus, E, Poissons ratio, , breadth, B, length, L, and
thickness, t. The two different raft thicknesses shown in Fig.
37, of 1 m and 3 m, lead to raftsoil stiffness ratios of 0
.
06
and 1
.
5, representing practical limits of exible and stiff
rafts.
The resulting settlement patterns for piled and unpiled
rafts are presented in Fig. 38(a) for the exible raft (where
the bending moments are negligible), whereas Fig. 38(b)
shows bending moment proles for the stiff raft (where the
differential settlements are negligible, less than 1% of the
average settlement). It may be seen from this gure that
the central pile support reduces differential settlements for
the exible raft case by a factor of between 2
.
7 (uniform
loading) and 3
.
5 (core-edge loading), with corresponding
ratios of differential settlement to average settlement for the
piled rafts of 0
.
15 and 0
.
24. For the stiff raft, maximum
(absolute) bending moments are reduced by a factor of about
2
.
4, at the expense of introducing greater hogging moments.
Thus, even for situations where a signicant fraction of the
applied load is concentrated towards the edges of the
foundation, central pile support appears a benecial design
approach.
For any given design, the precise layout and geometry of
the pile support would need to be ne-tuned by numerical
analysis. However, the principle is to optimise the design by
locating the pile support in such a way as to minimise
differential settlements and bending moments. In this way,
the foundation costs may be minimised without compromis-
ing performance. Although few, if any, piled rafts have been
optimised in this way, several studies have been published
showing how conventionally designed piled foundations
would have performed adequately, and with smaller differen-
tial settlements, with as little as 3050% of the actual
cumulative length (number times average length) of piles
used, but with the piles concentrated in the central part of
the raft (Horikoshi & Randolph, 1998; Prakoso & Kulhawy,
2001; Viggiani, 2001; Mandolini, 2003).
General loading: a case study
The nal topic covered in this paper concerns the founda-
tions for a major bridge in Vietnam, the construction of
which was funded jointly by the Vietnamese and Australian
governments. The My Thuan bridge crosses one of the two
branches of the Mekong delta, and is a cable-stayed bridge
with a central span of 350 m and clearance of 37
.
5 m (Fig.
39). Maunsell Australia Pty Ltd were the consulting engi-
neers for the project, and the author was a member of the
technical advisory group for the Australian Agency for
International Development (AusAID). The case study is
instructive in two ways, rst for its use of sophisticated pile
tests in order to nalise the design of the main pile support
for the bridge towers (Chandler, 1998), and second as an
illustration of the effects of redistributing load away from
heavily loaded piles, effectively allowing the ultimate limit
state to be considered in terms of allowable deformations.
The main towers of the bridge are supported on founda-
30 m
38 m
38 m
1 m
6 m
Core:edge (50:50)
Uniform
1 or 3 m
Fig. 37. Example piled raft showing two alternative load
distributions
0
.
5 0
.
3 0
.
1 0
.
1 0
.
3 0
.
5
K
rs
0
.
06
Piled raft
uniform loading
Piled raft
core-edge loading
Unpiled raft
uniform loading
Unpiled raft
core-edge loading
(a)
0
.
50
0
.
45
0
.
40
0
.
35
0
.
30
0
.
25
0
.
20
0
.
15
0
.
10
0
0
.
05
N
o
r
m
a
l
i
s
e
d

s
e
t
t
l
e
m
e
n
t
,

w
G
B
/
P
t
o
t
a
l
0
.
5 0
.
3 0
.
1 0
.
1 0
.
3 0
.
5
0
.
010
0
.
005
0
0
.
005
0
.
010
0
.
015
0
.
020
0
.
025
K
rs
1
.
5
Piled raft
uniform loading
Piled raft
core-edge loading
Unpiled raft
uniform loading
Unpiled raft
core-edge loading
Position across centreline of raft, x/B
Position across centreline of raft, x/B
N
o
r
m
a
l
i
s
e
d

s
e
t
t
l
e
m
e
n
t
,

w
G
B
/
P
t
o
t
a
l
(b)
Fig. 38. Settlement and moment proles across raft centreline:
(a) settlement proles; (b) proles of bending moment per unit
length
868 RANDOLPH
tions comprising 16 bored piles, arranged as two groups of
eight beneath each leg of the tower, and linked by a 6 m
thick pile cap as shown in Fig. 40. The piles are 2
.
4 m
(nominal) diameter, around 95 m long, and were constructed
under bentonite; base-grouting to 5 MPa was also performed.
Under normal conditions the water depth is around 23 m,
but a deep scour hole exists upstream of the bridge where
two branches of the river converge. The ultimate design
conditions allow for migration of the scour hole, resulting in
scouring to a depth of 47 m at the tower locations. The most
critical design loads correspond to ship impact on one of the
tower foundations, either parallel to the river (08) or at 458
to the river, with design loads as given in Fig. 40. The
design vertical load of 315 MN corresponds to an average
load of just under 20 MN per pile.
A simplied soil stratigraphy is shown in Fig. 41 for the
south tower and also for the rst piers on the southern bank.
Only limited strength data were obtained from the site
investigation, and instead the nal pile length was based on
the results of two load tests on piles for the south bank pier,
which were deliberately extended to reach the founding
stratum of dense sand deemed necessary for the tower piles.
Two pile load tests were conducted on the south bank pier,
each with a pair of Osterberg cells, as indicated. Further
load tests were conducted on the south and north tower
foundations, using a single Osterberg cell.
The use of Osterberg cells at two levels within a test pile
allows measurement of base capacity, shaft capacity over the
pile section between the two cells, and shaft capacity of the
upper pile section, by means of three test phases conducted
in sequence (Osterberg, 1989). The results of the two south
bank load tests are presented in Fig. 42. The measured base
response was consistent in both tests and showed no evi-
dence of any accumulation of soft sediments from the
construction process. The base response has been modelled
using a hyperbolic response with an ultimate end-bearing
pressure of 4
.
5 MPa, and an initial stiffness that corresponds
to a shear modulus of about 450 MPa.
The values of shaft friction measured in the rst load test
were below expectations, and before the second pile was
constructed, modications were made to the construction
procedures. These included steps to reduce the delay be-
tween excavation and concreting, a reduction in the head of
bentonite above river level, and mechanical scarifying of the
Fig. 39. My Thuan bridge in Vietnam
~ 60 m
2
.
4 m
x
5
.
5 m
V 315 MN
M 320 MNm
H 20 MN
95 m
Fig. 40. Pile layout for tower foundations of My Thuan bridge
Silty clay
(s
u
~ 200 kPa)
Water
Clayey sand
( 38)
Silty clay
(s
u
300 kPa)
Dense sand
( 40)
South bank test
piles (864 m)
South pier pile (96 m)
93 m
75 m
56 m
42 m
23 m
0 m
Osterberg cells
83 m
68 m
51 m
40 m
0 m
Fig. 41. Soil stratigraphy and location of Osterberg cells
Test 2
20 40 60 80 100
Test 1
Displacement: mm
Simulation (RATZ)
Base
Middle section
(between cells)
Test 1
Upper section
(above top cell)
Test 2
100
75
50
25
0
25
5
A
v
e
r
a
g
e

s
h
a
f
t

f
r
i
c
t
i
o
n
:

k
P
a
E
n
d
-
b
e
a
r
i
n
g
p
r
e
s
s
u
r
e
:

M
P
a
Shaft
Fig. 42. Response measured from load tests on south bank piles
SCIENCE AND EMPIRICISM IN PILE FOUNDATION DESIGN 869
borehole shaft prior to casting the pile. The last step was
achieved by welding short strands of wire to the callipering
tool (Fig. 43), which was then raised and lowered within the
borehole, rotating it as required in order to cover the full
circumference of the shaft.
As may be seen from Fig. 42, these measures led to an
improvement of about 30% in the shaft friction measured in
the second load test, with design values of 55 kPa and
90 kPa adopted for the upper and middle zones respectively.
Making allowance for slight differences in stratigraphy be-
tween the south bank and tower locations, and also for the
design scour depth of 47 m, the ultimate capacity of the
95 m long tower piles was determined as 42
.
1 MN. Adopting
a material factor of 0
.
72 (Australian Standards, 1995), and
allowing for the buoyant pile weight of 5
.
4 MN, a design
geotechnical ultimate capacity of 24
.
9 MN was arrived at for
the 95 m long piles. Subsequent load tests on the tower
piles, using a single Osterberg cell, resulted in a measured
ultimate capacity of the upper section of pile of 2627 MN,
and conrmed an ultimate capacity well in excess of
30 MN.
Analysis of the pile group under the design load condi-
tions, using the software PIGLET (Randolph, 2003), leads to
a distribution of loads among the piles as shown in Fig. 44.
Assuming elastic response of the piles, it is found that three
piles under load case 1, and six piles under load case 2,
exceed the design capacity of 24
.
9 MN. However, non-linear
analysis, with the axial load limited to a notional design
capacity, allows the loads to be redistributed among the piles
at a cost of increased deformations. The load distributions
for the case where the axial loads have been limited to
22 MN are shown in Fig. 44. For each load case, only four
piles remain with axial loads less than the imposed limit of
22 MN, suggesting that the pile group is close to failure in
the sense of no longer being able to nd a load distribution
in equilibrium with the applied loading.
Figure 45 shows the development of vertical displacement
at the pile group centroid for the non-linear analyses. The
lower two curves are conventional loaddisplacement re-
sponse for the two load cases, as the applied loads are
factored up proportionally to their full values and with an
imposed axial load limit of 22 MN on individual piles. Final
displacements are 280 mm (load case 1) and 250 mm (load
case 2), compared with the elastic value of 74 mm for both
cases.
From a design perspective, more useful information is
given by the upper two curves in Fig. 45, which give the
nal vertical displacements for different magnitudes of the
limiting axial load. This shows the more critical nature of
load case 2, where displacements start to increase signi-
cantly once the axial load limit is reduced below 28 MN.
For the actual design limit of 24
.
9 MN the vertical displace-
ment is 140 mm for this load case (and 76 mm for load case
1: little more than the elastic value).
An appropriate design strategy for bridge foundations
involves an iterative interaction with the structural design
engineers, in order to determine what displacement of the
foundation constitutes an ultimate limit state. In the present
case this allows conrmation that the geotechnical design
limit of 24
.
9 MN (and the resulting displacement of
140 mm) is indeed adequate, and essentially weights the
denition of ultimate limit state more towards a deformation
criterion, rather than the geotechnical capacity of individual
piles in the group.
Fig. 43. Tool to scarify borehole shaft prior to concreting
Load case 1:
ship impact at 0
40
30
20
10
0
10
20
30
40
30 20 10 0 10 20 30
Load case 2:
ship impact at 45
Non-linear analysis
axial limit: 22 MN
Distance from pile group centroid: m
A
x
i
a
l

l
o
a
d
:

M
N
Upload
Elastic
analysis
Download
Fig. 44. Computed axial loads for critical load cases
Load
case 2
Final displacement as
function of axial
capacity
Load-displacement
response with axial
capacity of 22 MN
Load
case 1
Load
case 2
Elastic
030 025 020 015 010 005 0
0
5
10
15
20
25
30
35
Vertical displacement of pile group centriod: m
A
x
i
a
l

c
a
p
a
c
i
t
y
:
M
N
A
v
e
r
a
g
e

a
x
i
a
l
l
o
a
d
:

M
N
Fig. 45. Non-linear response of pile group under design loads
870 RANDOLPH
Summary
In contrast to other areas of pile design, the science of
predicting the deformation response of pile groups is well
advanced, with a variety of methods available, differing in
complexity and degree of rigour. Under vertical loading,
quantication of the reinforcing effect of piles in moderating
interaction factors is a major improvement, and the approach
leads to estimates of pile group stiffness that are consistent
with simple analogues based on equivalent pier or raft
approximations.
For piled rafts, the benets of central pile support for
(primarily) raft foundations have been documented widely
both for uniformly loaded pile groups and for the more
demanding case where a signicant proportion of the loading
is concentrated at the raft edges. As illustrated here, even in
that case, substantial reduction in differential settlements (for
exible rafts) and absolute magnitude of bending moments
(in stiffer rafts) is possible using central pile support.
The trend towards design based on allowable deformations
may also be appropriate for small pile groups, such as for
supporting bridge piers, which are subjected to combined
vertical and horizontal loading. Non-linear analysis allows
consideration of the consequences of limiting axial (or
lateral) loads on individual piles to a notional design value,
and hence, in conjunction with structural considerations,
permits even the ultimate limit state to be weighted more
towards allowable deformations.
The key soil parameter required for predicting the defor-
mation response of pile groups is the shear modulus, and
back-analysis of the performance of large pile groups sug-
gests that the relevant value is close to the small-strain
modulus, G
0
(Mandolini & Viggiani, 1997). This may be
measured relatively easily using modern methods such as
seismic cone tests or the small strain response of high-
quality samples. For piled rafts, with fewer piles and higher
average shear stresses in the soil, some factoring of G
o
will
be necessary, and further analysis and eld evidence is
needed in order to quantify the reduction. The use of a
modied hyperbolic response to bridge between the initial
stiffness (derived using G
0
) and the ultimate capacity ap-
pears a promising approach (Mayne & Poulos, 2001; Mayne,
2003). Pile load tests carried out near the start of the piling
contract provide a useful means of verifying, or ne-tuning,
the design, not just in terms of pile capacity, but also in
respect of shear modulus values deduced from the load
displacement response at different load levels.
CONCLUDING REMARKS
Einstein once commented along the lines that:
Experimental data are believed by everyone, except the
person who did the experiment; while theory is believed by
nobody, except the person who developed it.
This comment is particular pertinent to pile design, where
we need to be vigilant over the quality of data relied upon
in empirical approaches and to continue to improve our
analytical methods. With this in mind, the aim of this paper
has been to explore advances in scientic approaches to pile
design, and the extent to which we still need to rely on
empirical correlations.
The scope has been deliberately restricted, partly because
of space limitations, but also reecting the authors interests
and experience. Summary comments have been provided at
the end of each topic and will not be repeated here.
In estimating the axial capacity of piles, areas have been
identied where empirical trends from measured perform-
ance do not appear to be supported by science in the form
of conceptual or analytical models. These areas therefore
deserve closer scrutiny, particularly where design requires
extrapolation outside the existing database.
Overall, signicant uncertainty exists in our ability to
estimate the axial capacity of individual piles, and hence we
need to adapt our design and contracting strategies to en-
courage the testing of piles in order to ne-tune designs,
allowing more optimistic material factors as more precise
estimates of pile capacity become possible. Dynamic pile
testing has an important role in this respect, but inertia
within the industry has restricted the adoption of more
scientic models of pilesoil interaction.
We should capitalise on the ductile response of most piled
foundation systems, weighting design criteria more towards
allowable deformations, with reduced dependence on the
capacity of individual piles. Efcient design of piled rafts
may allow the piles to operate close to their geotechnical
design capacity under working conditions, with their main
purpose being to minimise differential settlements and bend-
ing moments in what is primarily a raft foundation.
Despite adverse comments by some of the pioneers of soil
mechanics, there is a signicant role for scientic methods
in pile design. If we are to continue to attract the best of
each new generation of engineers, then we must incorporate
such science in our teaching and our practice, using empiri-
cal approaches to validate and calibrate, but not replace,
scientic theory.
ACKNOWLEDGEMENTS
I would like to gratefully acknowledge support and con-
tributions from a wide range of people during preparation of
this paper: rst and foremost my wife, Cherry, and sons,
Nick and Tom; mentors at key stages of my career, includ-
ing John Burland, Peter Wroth, Andrew Schoeld and John
Booker; colleagues Martin Fahey and Barry Lehane from
UWA, and Carl Erbrich from Advanced Geomechanics;
collaborators who have supplied data and taken time to
correspond on some of the issues, particularly David White
but also Antonio Alvez, Fiona Chow, George Mylonakis and
Oliver Reul; and nally past and present staff and students
in the Geomechanics Group at UWA who have made it such
a fun place to work. I am also grateful to AusAID for
permission to use data from the My Thuan bridge project.
NOTATION
A area (A
p
for cross-sectional area of pile)
B
r
width of raft foundation
c wave speed in pile
c
h
coefcient of consolidation for horizontal ow
D
r
relative density
d diameter of pile (subscripts eq for equivalent, i for
internal)
E Youngs modulus (subscripts b for soil at pile base, p for
pile, r for raft)
F dynamic axial force in pile
G shear modulus of soil
h distance measured upwards from pile tip (subscript p
for soil plug)
I
r
rigidity index, G/s
u
J Smith damping factor
K compressibility factor, horizontal stress ratio (subscript
0 for in situ value), pile head stiffness
K
rs
non-dimensional raft-soil stiffness ratio
k load transfer factor, permeability (with subscript h for
horizontal ow)
L pile length, (subscript r for raft length)
m non-dimensional rate factor
n exponent
P load at pile head
PI plasticity index
SCIENCE AND EMPIRICISM IN PILE FOUNDATION DESIGN 871
p mean stress (subscripts i for initial, o for in situ, a for
atmospheric pressure)
Q Smith quake, capacity (subscript s for shaft)
q pressure (subscripts b for pile base, c for cone
resistance)
R yield stress ratio
R
f
pile capacity reduction factor
r radial coordinate (subscript m for maximum radius of
inuence)
r
eq
equivalent pile radius
S
t
soil sensitivity
s
u
undrained shear strength
T, T

non-dimensional time factors (subscript eq for


equivalent)
t time
t
r
thickness of raft
u pore pressure (subscripts o for in situ, max for
maximum value)
v velocity (subscript s for shear wave velocity in soil)
w settlement (subscript t for pile head)
z depth co-ordinate
shaft friction correlation factor, interaction factor
between piles
shaft friction correlation factor
unit weight (9 for effective, subscript w for water)
indicating change
interface friction angle, load transfer ratio k/G, also
indicating increment
load transfer parameter
non-dimensional pile compressibility parameter
shaft compressibility inverse length, parameter in radial
consolidation model
parameter in radial consolidation model and radial
stress degradation
Poissons ratio (subscripts p for pile, r for raft)
ratio of residual to peak shaft friction, diffraction
parameter
mathematical constant
r area ratio
stress (subscripts p for vertical yield stress, rc for radial
after consolidation, rf for radial at failure, ri for initial
radial, v
0
for in situ vertical)
shear stress (subscripts peak for peak, res for residual, s
for shaft friction)
9 soil friction angle
geometry factor for pile interaction
non-dimensional base stiffness parameter
REFERENCES
Altaee, A., Fellenius, B. H. & Evgin, E. (1992). Axial load transfer
for piles in sand: I. Tests on an instrumented precast pile. Can.
Geotech. J. 29, No. 1, 1120.
Altaee, A., Fellenius, B. H. & Evgin, E. (1993). Axial load transfer
for piles in sand and the critical depth. Can. Geotech. J., 30,
No. 3, 455463.
Andersen, K. H. & Jostad, H. P. (2002). Shear strength against
outside wall of suction anchors in clay after installation. Pro-
ceedings of the international conference on offshore and polar
engineering, Kyushu, paper 2002PCW02.
API (1993). RP2A: Recommended practice for planning, designing
and constructing xed offshore platforms. Washington, DC:
American Petroleum Institute.
Australian Standards (1995). Piling Design and installation,
AS2159-1995. Sydney: Standards Australia.
Baligh, M. M. (1985). Strain path method. J. Soil Mech. Found.
Div., ASCE 111, No. 9, 11801136.
Baligh, M. M. (1986). Undrained deep penetration. Geotechnique
36, No. 4, 471485; 487501.
Banerjee, P. K. & Buttereld, R. (1981). Boundary element method
in engineering science. New York: McGraw-Hill.
Basile, F. (1999). Non-linear analysis of pile groups. Proc. ICE
Geotech. Engng 137, 105115.
BCP Committee (1971). Field tests on pipe piles in sand. Soils
Found. 11, No. 2, 2949.
Bermingham, P. & Janes, M. (1989). An innovative approach to
load testing of high capacity piles. Proceedings of the interna-
tional conference on piling and deep foundations, London, pp.
409413.
Bond, A. J. & Jardine, R. J. (1991). Effects of installing displace-
ment piles in a high OCR clay. Geotechnique, 41, No. 3, 341
363.
Briaud, J.-L., Tucker, L. M. & Eng, E. (1989). Axially loaded 5 pile
group and single pile in sand. Proc. 12th Int. Conf. Soil Mech.
Found. Engng, Rio de Janeiro 2, 11211124.
Brucy, F., Meunier, J. & Nauroy, J.-F. (1991). Behaviour of pile plug
in sandy soils during and after driving. Proc. 23rd Annual
Offshore Technology Conf., Houston, 145154.
Bruno, D. (1999). Dynamic and static load testing of driven piles in
sand. PhD thesis, The University of Western Australia.
Burland, J. B., Broms, B. B. & De Mello, V. F. B. (1977). Behav-
iour of foundations and structures. Proc. 9th Int. Conf. Soil
Mech. Found. Engng, Tokyo 2, 495546.
Bustamante, M. & Gianeselli, L. (1982). Pile bearing capacity by
means of static penetrometer CPT. Proc. 2nd Eur. Symp.
Penetration Testing, Amsterdam, 493499.
Cao, J., Phillips, R., Popescu, R., Audibert, J. & Al-Khafaji, Z.
(2002). Excess pore pressures induced by installation of suction
caissons in NC clays. Proceedings of the international confer-
ence on offshore site investigation and geotechnics, London, pp.
405412.
Chandler, B. C. (1998). My Thuan Bridge: update on bored pile
foundations. Proceedings of the Australasian bridge conference,
Sydney.
Chow, F. C. (1997). Investigations in the behaviour of displacement
piles for offshore foundations. PhD thesis, Imperial College,
London.
Chow, F. C., Jardine, R. J., Brucy, F. & Nauroy, J. F. (1998). Effects
of time on capacity of pipe piles in dense marine sand. J.
Geotech. Geoenviron. Engng Div., ASCE 124, No. 3, 254264.
Clancy, P. & Randolph, M. F. (1996). Simple design tools for piled
raft foundations. Geotechnique 46, No. 2, 313328.
Cooke, R. W. (1986). Piled raft foundations on stiff clays: a
contribution to design philosophy. Geotechnique 36, No. 2,
169203.
Coop, M. R. & Wroth, C. P. (1990). Discussion of M. R. Coop &
C. P. Wroth (1989): Field studies of an instrumented model pile
in clay. Geotechnique 39, No. 4, 679696; 40, No. 4, 669672.
Coyle, H. M. & Castello, R. R. (1981). New design correlations for
piles in sand. J. Geotech. Engng Div., ASCE 197, No. GT7,
965985.
De Beer, E., de Jonghe, A., Carpentier, R. & Wallays, M. (1979).
Analysis of the results of loading tests on displacement piles
penetrating into a very dense sand layer. Proceedings of the
conference on recent developments in the design and construc-
tion of piles, London, pp. 199211.
Deeks, A. J. & Randolph, M. F. (1995). A simple model for
inelastic footing response to transient loading. Int. J. Numer.
Anal. Methods Geomech 19, No. 5, 307329.
De Jong, J. T. & Frost, J. D. (2002). A multisleeve friction
attachment for the cone penetrometer. Geotech. Test. J., ASTM
25, No. 2, 111127.
De Nicola, A. & Randolph, M. F. (1993). Tensile and compressive
shaft capacity of piles in sand. J. Geotech. Engng Div., ASCE
119, No. 12, 19521973.
De Nicola, A. & Randolph, M. F. (1997). The plugging behaviour
of driven and jacked piles in sand. Geotechnique 47, No. 4,
841856.
De Nicola, A. & Randolph, M. F. (1999). Centrifuge modelling of
pipe piles in sand under axial loads. Geotechnique 49, No. 3,
295318.
England, M. & Fleming, W. G. K. (1994). Review of foundation
testing methods and procedures. Proc. ICE Geotech. Engng 107,
No. 3, 135142.
Fahey, M. & Lee Goh, A. (1995). A comparison of pressuremeter
and piezocone methods of determining the coefcient of con-
solidation, Proc. 4th Int. Symp. on the Pressuremeter and Its
New Avenues, Quebec, 153160.
Fioravante, V. (2002). On the shaft friction modelling of non-
displacement piles in sand. Soils Found. 42, No. 2, 2333.
Fioravante, V., Ghionna, V. N., Jamiolkowski, M. & Sarri, H.
872 RANDOLPH
(1999). Shaft friction modelling of non-displacement piles in
sand. Proceedings of the international conference on analysis,
design, construction and testing of deep foundations, Austin,
TX.
Fleming, W. G. K. (1992). A new method for single pile settlement
prediction and analysis. Geotechnique 42, No. 3, 411425.
Fleming, W. G. K. & Thorburn, S. (1983). Recent piling advances:
state of the art. Proceedings of the conference on recent ad-
vances in piling and ground treatment, London.
Franke, E., Lutz, B. & El-Mossallamy, Y. (1994). Measurements
and numerical modelling of high rise building foundations on
Frankfurt clay. Proceedings of the conference on vertical and
horizontal deformations of foundations an embankments, Texas,
ASCE Geotechnical Special Publication No. 40, Vol. 2, pp.
13251336.
Gibson, R. E. & Anderson, W.F. (1961). In situ measurement of soil
properties with the pressuremeter, Civ. Engng Public Works Rev.
56, 615618.
Gregersen, O. S., Aas, G. & Dibagio, E. (1973). Load tests on
friction piles in loose sand. Proc. 8th Int. Conf. Soil Mech.
Found. Engng, Moscow 2, 109117.
Guo, W. D. & Randolph, M. F. (1997). Vertically loaded piles in
non-homogeneous media. Int. J. Numer. Anal. Methods Geo-
mech., 21, No. 8, 507532.
Heerema, E. P. & de Jong, A. (1979). An advanced wave equation
computer program which simulates dynamic pile plugging
through a coupled mass-spring system. Proceedings of the
International conference on numerical methods in offshore
piling, London, pp. 3742.
Hight, D. W., Lawrence, D. M., Farquhar, G. B., Milligan, G. W.,
Gue, S. S. & Potts, D. M. (1996). Evidence for scale effects in
the bearing capacity of open-ended piles in sand. Proc. 28th
Annual Offshore Technology Conf., Houston, 181192.
Horikoshi, K. & Randolph, M. F. (1997). On the denition of raft-
soil stiffness ratio. Geotechnique 47, No. 5, 10551061.
Horikoshi, K. & Randolph, M. F. (1998). Optimum design of piled
rafts. Geotechnique 48, No. 3, 301317.
Ishihara, K., Saito, A., Shimmi, Y., Miura, Y. & Tominaga, M.
(1977). Blast furnace foundations in Japan. Proc. 9th Int. Conf.
Soil Mech. Found. Engng, Tokyo Case history volume, 157236.
Jardine, R. J. & Chow, F. C. (1996). New design methods for
offshore piles, MTD Publication 96/103. London: Marine Tech-
nology Directorate.
Karlsrud, K. (1999). Lessons learned from instrumented pile load
tests in clay. Proceedings of the international conference on
analysis, design, construction and testing of deep foundations,
Austin, TX.
Katzenbach, R., Arslan, U. & Moormann, C. (2000). Piled raft
foundation projects in Germany. In Design applications of raft
foundations (ed. J. A. Hemsley), pp. 323391, London: Thomas
Telford.
Kirby, R. C. & Esrig, M. I. (1979). Further development of a
general effective stress method for prediction of axial capacity
for driven piles in clay. Proceedings of the conference on recent
developments in the design and construction of piles, London,
pp. 335344.
Kolk, H. J. & van der Velde, E. (1996). A reliable method to
determine friction capacity of piles driven into clays. Proc.
Offshore Technology Conf., Houston, Paper OTC 7993.
Kulhawy, F. H. (1984). Limiting tip and side resistance: fact or
fallacy? In Analysis and design of pile foundations (ed. J. R.
Meyer), pp. 8098. New York: ASCE.
Kusakabe, O., Matsumoto, T., Sandanbata, I., Kosuge, S. &
Nishimura, S. (1989). Report on questionnaire: predictions of
bearing capacity and driveability of piles. Proc. 12th Int. Conf.
Soil Mech. Found. Engng, Rio de Janeiro 5, 29572963.
Lee, J. H. & Salgado, R. (1999). Determination of pile base
resistance in sands. J. Geotech. Geoenviron. Engng, ASCE 125,
No. 8, 673683.
Lee, S. L., Chow, Y. K., Karunaratne, G. P. & Wong, K. Y. (1988).
Rational wave equation model for pile-driving analysis. J. Geo-
tech. Engng, ASCE 114, No. 3, 306325.
Lehane, B. M. (1992). Experimental investigations of pile behaviour
using instrumented eld piles. PhD thesis, Imperial College,
London.
Lehane, B. M. & Gavin, K. G. (2001). The base resistance of
jacked pipe piles in sand. J. Geotech. Geoenviron. Engng, ASCE
127, No. 6, 473480.
Lehane, B. M. & Jardine, R. J. (1994). Displacementpile behav-
iour in a soft marine clay. Can. Geotech. J. 31, No. 2, 181191.
Lehane, B. M. & Randolph, M. F. (2002). Evaluation of a minimum
base resistance for driven pipe piles in siliceous sand. J.
Geotech. Geoenviron. Engng Div., ASCE 128, No. 3, 198205.
Lehane, B. M., Jardine, R. J., Bond, A. J. & Frank, R. (1993).
Mechanisms of shaft friction in sand from instrumented pile
tests. J. Geotech. Engng Div., ASCE 119, No. 1, 1935.
Lehane, B. M., Jardine, R. J., Bond, A. J. & Chow, F. C. (1994).
The development of shaft friction on displacement piles in clay.
Proc. 13th Int. Conf. Soil Mech. Found. Engng, New Delhi 2,
473476.
Litkouhi, S. & Poskitt, T. J. (1980). Damping constant for pile
driveability calculations. Geotechnique 30, No. 1, 7786.
Liyanapathirana, D. S., Deeks, A. J. & Randolph, M. F. (2001).
Numerical modelling of the driving response of thin-walled
open-ended piles. Int. J. Numer. Anal. Methods Geomech. 25,
No. 9, 933953.
Maiorano, R. M. S., Viggiani, C. & Randolph, M. F. (1996).
Residual stress system arising from different methods of pile
installation. Proc. 5th Int. Conf. on Application of Stress-Wave
Theory to Piles, Orlando, 518528.
Mandolini, A. (2003). Design of piled raft foundations: practice and
development. Proc. 4th Int. Sem. on Deep Foundations on Bored
and Auger Piles, Ghent, 5980.
Mandolini, A. & Viggiani, C. (1997). Settlement of piled founda-
tions. Geotechnique 47, No. 4, 791816.
Mayne, P. W. (2003). Class A footing response prediction from
seismic cone tests. Proc. 3rd Int. Symp. on Deformation Charac-
teristics of Geomaterials, Lyon 1, 883888.
Mayne, P. W. & Kulhawy, F. H. (1982). K
0
OCR relationships in
soils. J. Geotech. Engng Div., ASCE 108, No. GT6, 851872.
Mayne, P. W. & Poulos, H. G. (2001). Closure to Approximate
displacement inuence factors for elastic shallow foundations.
J. Geotech. Geoenviron. Engng Div., ASCE 127, No. 1, 100
102.
McCammon, N. R. & Golder, H. Q. (1970). Some loading tests on
long pipe piles. Geotechnique 20, No. 2, 171184.
Mesri, G., Rokhsar, A. & Bohor, B. F. (1975). Composition and
compressibility of typical samples of Mexico Clay. Geotechni-
que 25, No. 3, 527554.
Meyerhof, G. G. (1976). Bearing capacity and settlement of pile
foundations. J. Geotech. Engng Div., ASCE102. No. GT3, 197228.
Meyerhof, G. G. & Valsangkar, A. J. (1977). Bearing capacity of
piles in layered soils. Proc. 8th Int. Conf. Soil Mech. Found.
Engng, Moscow 1, 645650.
Mylonakis, G. (2001). Winkler modulus for axially loaded piles.
Geotechnique 51, No. 5, 455461.
Mylonakis, G. & Gazetas, G. (1998). Settlement and additional
internal forces of grouped piles in layered soil. Geotechnique
48, No. 1, 5572.
Novak, M. (1977). Vertical vibration of oating piles. J. Engng
Mech. Div., ASCE 103, No. EM1, 153168.
ONeill, M. W. (2001). Side resistance in piles and drilled shafts. J.
Geotech. Geoenviron. Engng Div., ASCE 127, No. 1, 116.
Osterberg, J. (1989). New device for load testing driven piles and
drilled shafts separates friction and end-bearing. Proceedings of
the international conference on piling and deep foundations,
London, Vol. 1, pp. 421427.
Paik, K. H. & Lee, S. R. (1993). Behaviour of soil plugs in open-
ended model piles driven into sands. Mar. Georesources Geo-
technol. 11, 353373.
Paik, K. H., Salgado, R., Lee, J. & Kim, B. (2003). Behaviour of
open and closed-ended piles driven into sand. J. Geotech.
Geoenviron. Engng Div., ASCE 129, No. 4, 296306.
Poulos, H. G. (1968). Analysis of settlement of pile groups.
Geotechnique 18, No. 3, 449471.
Poulos, H. G. (1987). Analysis of residual effects in piles. J.
Geotech. Engng Div., ASCE 113, No. 3, 216219.
Poulos H. G. (1989). Pile behaviour: theory and application.
Geotechnique 39, No. 3, 365415.
Poulos, H. G. (1994). An approximate numerical analysis of pile-
raft interaction. Int. J. Numer. Anal. Methods Geomech. 18,
7392.
SCIENCE AND EMPIRICISM IN PILE FOUNDATION DESIGN 873
Poulos, H. G. (1998). Pile testing: from the designers viewpoint.
Proc. 2nd Int. Statnamic Sem., Tokyo, 321.
Poulos, H. G. (2001). Piled-raft foundation: design and applications.
Geotechnique 51, No. 2, 95113.
Poulos, H. G. & Davis, E. H. (1980). Pile foundation analysis and
design. New York: John Wiley & Sons.
Prakoso, W. A. & Kulhawy, F. H. (2001). Contribution to piled raft
optimum design. J. Geotech. Geoenviron. Engng, ASCE 127,
No. 1, 1724.
Randolph, M. F. (1983). Design considerations for offshore piles.
Proceedings of the conference on geotechnical practice in off-
shore engineering, Austin, pp. 422439.
Randolph, M. F. (1987). Modelling of the soil plug response during
pile driving. Proc. 9th SE Asian Geotech. Conf., Bangkok 2,
6.16.14.
Randolph, M. F. (1990). Analysis of the dynamics of pile driving.
In Developments in soil mechanics IV: Advanced geotechnical
analyses (eds P. K. Banerjee and R. Buttereld). Elsevier
Applied Science, pp. 223272.
Randolph, M. F. (1994). Design methods for pile groups and piled
rafts. Proc. 13th Int. Conf. Soil Mech. Found. Engng, New Delhi
5, 6182.
Randolph, M. F. (2000). Pilesoil interaction for dynamic and static
loading. Proc. 6th Int. Conf. on Application of Stress-Wave
Theory to Piles, Sao Paulo Appendix, 311.
Randolph, M. F. (2003). PIGLET: Analysis and design of pile
groups. Users Manual, Version 4-2. Perth.
Randolph, M. F. & Deeks, A. J. (1992). Dynamic and static soil
models for axial pile response dynamics. Proc. 4th Int. Conf. on
Application of Stress-Wave Theory to Piles, The Hague, 114.
Randolph, M. F. & Murphy, B. S. (1985). Shaft capacity of driven
piles in clay, Proc. 17th Ann. Offshore Technol. Conf., Houston, 1,
371378.
Randolph, M. F. & Wroth, C. P. (1978). Analysis of deformation of
vertically loaded piles. J. Geotech. Engng. Div., ASCE 104, No.
GT12, 14651488.
Randolph, M. F. & Wroth, C. P. (1979). An analytical solution for
the consolidation around a driven pile. Int. J. Numer. Anal.
Methods Geomech. 3, No. 3, 217229.
Randolph, M. F., Carter, J. P. & Wroth, C. P. (1979). Driven piles in
clay: the effects of installation and subsequent consolidation.
Geotechnique 29, No. 4, 361393.
Randolph, M. F., Leong, E. C. & Houlsby, G. T. (1991). One-
dimensional analysis of soil plugs in pipe piles. Geotechnique
41, No. 4, 587598.
Randolph, M. F., Dolwin, J. & Beck, R. D. (1994). Design of driven
piles in sand. Geotechnique 44, No. 3, 427448.
Rausche, F., Goble, G. G. & Likins, G. E. (1985). Dynamic
determination of pile capacity. J. Geotech. Engng Div., ASCE
111, 367383.
Reul, O. & Randolph, M. F. (2004). Design strategies for piled rafts
subjected to non-uniform vertical loading. J. Geotech. Geoenvir-
on. Engng Div, ASCE 130. (1), (in press).
Russo, G. (1998). Numerical analysis of piled rafts. Int. J. Anal.
Numer. Methods Geomech. 22, No. 6, 477493.
Semple, R. M. & Rigden, W. J. (1984). Shaft capacity of driven
piles in clay, Proceedings of the symposium on analysis and
design of pile foundations, San Francisco, pp. 5979.
Shioi, Y., Yoshida, O., Meta, T. & Homma, M. (1992). Estimation
of bearing capacity of steel pipe pile by static loading test and
stress-wave theory. Proc. 4th Int. Conf. on Application of Stress-
Wave Theory to Piles, The Hauge, 325330.
Simons, H. A. & Randolph, M. F. (1985). A new approach to one-
dimensional pile driving analysis. Proc. 5th Int. Conf. Numerical
Methods in Geomechanics, Nagoya 3, 14571464.
Smith, E. A. L. (1960). Pile driving analysis by the wave equation.
J. Soil Mech., ASCE 86, 3561.
Teh, C. I. & Houlsby, G. T. (1991). An analytical study of the cone
penetration test in clay. Geotechnique 41, No. 1, 1734.
Toolan, F. E., Lings, M. L. & Mirza, U. A. (1990). An appraisal of
API RP2A recommendations for determining skin friction of
piles in sand. Proc. 22nd Annual Offshore Technol. Conf.,
Houston, 3342.
Vesic, A. S. (1967). A study of bearing capacity of deep founda-
tions, Final Report, Project B-189. Atlanta: Georgia Institute of
Technology.
Vesic, A. S. (1970). Tests on instrumented piles, Ogeechee River
site. J. Soil Mech. Found. Div., ASCE 96, No. SM2, 561584.
Viggiani, C. (2001). Analysis and design of piled foundations: First
Arrigo Croce Lecture. Riv. Ital. di Geotecnica 35, No. 1, 4775.
Vijayvergiya, V. N. & Focht, J. A. (1972). A new way to predict
capacity of piles in clay. Proc. 4th Annual Offshore Technol.
Conf., Houston 2, 865874.
White, D. J. (2003). Field measurements of CPT and pile base
resistance in sand, Technical Report CUED/D-SOILS/TR327.
Cambridge University Engineering Dept.
White, D. J. & Bolton, M. D. (2002). Observing friction fatigue on
a jacked pile. In Centrifuge and constitutive modelling: two
extremes (ed. S. M. Springman), pp. 347354. Rotterdam: Swets
& Zeitlinger.
Whittle, A. J. (1992). Assessment of an effective stress analysis for
predicting the performance of driven piles in clays. Proceedings
of the conference on offshore site investigation and foundation
behaviour, London, Vol. 28, pp. 607643.
VOTE OF THANKS
HUGH ST JOHN, Director, Geotechnical Consulting Group
When I contacted Mark to ask him to write his own
introduction (which at the time I thought I was going to
do), he said that he would prefer me not to be sycophantic
but to refer to his return from exile after 16 years. I ended
up taking responsibility for the vote of thanks, which gives
me a lot more license, because he wouldnt write that
either.
I would like to start off with a complaint about his
Australian accent. This was supposed to be an overseas
lecture, Mark. You made no effort to make it sound like
one.
I rst met Mark around 30 years ago when he turned up
at BRE, the bright young graduate fresh from Oxford who
seemed a little bewildered about what he was supposed to
be doing. I had preceded him by a few years but was
immediately aware of the presence of a superior intellect. I
gave up competing after we both decided to go in for the
Cooling Prize. At least I am a quick learner in some
respects. I think that these early years were very formative
for him. He found himself in a situation which suited him; a
group of people looking for theory to t their excellent eld
data, and a leadership which encouraged self development.
Mark soon found his problems to solve and after a short
while a subject that he could develop as his own, the
behaviour of piles. Tonight we have seen how this has
blossomed, initially under the mentorship of John Burland
and the late Peter Wroth, of course, two former Rankine
Lecturers. I think that both of them instilled in him what he
has so ably demonstrated tonight, the importance of the
why? when deciding on the how?, but also the art of
distilling what is a very complex problem through a series
of logical steps, to something that can be understood.
Not once this evening have we seen an equation that is
more than half a line long. Not once have we seen a formula
with a multitude of variables. Marks hallmark is that he
breaks things down into a series of logical steps and then
uses the building blocks he creates to examine the problem
in its entirety. Although this is an academic approach, he
uses such models to solve practical problems, and applies
his engineering judgement and observation to challenge the
logic of both other peoples theories and the way in which
design is carried out. This is a very powerful combination
which he has obviously used to good effect in advising a
wide range of clients.
Mark expressed some concern to me that in part of his
lecture he is challenging some of the assumptions made by
others in the derivation of design methods, and that they
may take umbrage as they are so close to home. I assured
874 RANDOLPH
him, probably unwisely that, as true scientists, trying to
rene their own understanding of their data, they would
welcome such a side swipe. But maybe, as a non-academic I
havent understood how things really work. There are many
more PhDs to be had from the subject, and Im sure that
even those closest to the subject have seen something here
this evening which may stimulate a further thought.
I think that Einstein was right though about psyche of the
Scientist. I hadnt realised before that even Einstein, like
everyone else, only presented his best data. . .otherwise he
wouldnt be sceptical about it.
I tried to nd a quote to match Marks, and decided that I
should look for something from another adopted son of
Australia, the late Spike Milligan, who had a theory about
the origin of rain. He wrote. . ..
There are holes in the sky where the rain gets in, but they
are ever so small. Thats why the rain is thin.
This is an acute observation, but not backed up by the
scientic theory. However, it is an empirical relationship
between the sky and the size of the rain drops which always
works.
Tonight Mark has more than achieved what we have come
to expect from a Rankine lecture. He has shown us the fruits
of his lifetimes work, challenged us to think further about
the assumptions that we make when we select a particular
design method and shown us ways of doing better. He has
also demonstrated through some fascinating examples how
an understanding of the science and thus an appreciation of
the mechanisms controlling soilstructure interaction can
result in nding better solutions to geotechnical problems,
backed up of course by a means of verifying the result. He
has been and will continue to be an inspiration to many both
within his own highly successful teams and to the geotechni-
cal profession worldwide.
I would like you now to join me in giving our heartfelt
thanks to Mark for all the blood, sweat and tears (probably
largely sweat) that he has put in to preparing and delivering
the 43
rd
Rankine Lecture.
SCIENCE AND EMPIRICISM IN PILE FOUNDATION DESIGN 875

You might also like