Professional Documents
Culture Documents
Poulos (1989) Geot.1989.39.3.365 PDF
Poulos (1989) Geot.1989.39.3.365 PDF
Poulos (1989) Geot.1989.39.3.365 PDF
3, 36S415
H. G. POULOS*
This paper discusses the application of theory to L’article discute I’application de la thbrie i
the analysis of pile foundation behaviour under l’analyse du comportement des fondations par
axial loading. A classification system is suggested pieux sous chargement axial. On propose un
for pile analysis and design procedures, based on systeme de classification ba& sur la rigueur de la
the rigour of the underlying theory. It is shown that thborie associb pour I’analyse des pieux et les pro-
a number of the analyses in use have a common &d&s de construction. On dCmontre que quelques-
underlying theoretical basis founded on the bound- unes des analyses employ&es ont une base thborique
ary element method and that such methods give commune dbivant de la m6thode des &ments
solutions which are consistent with those from limites et que de telles mbthodes donnent des solu-
other methods such as the finite element method. tions qui s’accordent bien avec celles obtenues a
The main characteristics of pile behaviour are partir d’autres mbthodes, telles que lea &ments
reviewed for single piles and pile groups subjected finis. Les caractitristiques principales du com-
to static loading, cyclic loading, and to loading portement des pieux soot examink pour lea pieux
arising from externally-imposed soil movements. et les groupes de pieux soumis aux chargements
Correlations are then summarized for the geotech- statiques et cycliques aussi bien qu’au chargement
nical parameters required for the prediction of r&&ant des mouvements du sol impo&s de
axial pile behaviour. Using characteristic values of I’extbrieur. Puis on rbume les corrklations pour les
these parameters, design charts are developed for param&res gbtechniques &ceasaires pour la pr8
the load capacity and settlement of piles and pile diction du comportement axial des pieux. A I’aide
groups. Finally, three case studies are described de valeurs caract&istiques de ces parametres on
which demonstrate the sensitivity of pile per- propose des abaques de construction pour la force
formance predictions to the method of analysis, the portante et le tassement des pieux et des groupes
idealization of the soil profile, and the selection of de pieux. Trois cas rMs sont p&en&. IIs illus-
soil parameters. It is shown that the method of trent la faqon dont les p&lictlons du com-
analysis is likely to have less effect on the predict- portement des pieux dCpendent de la mbthode
ed performance than does the geotechnical charac- d’analyse, de la sckmatisation du profil du sol et
terization of the site. de la &lection des parametres du sol. On dkmontre
que la performance prkvue est probablement moins
influencb par la mbthode d’analyse que par la di+
KEYWORDS: analysis; case history; design; piles; termination des caract&istiques gbtechniques du
repeated loading; settlement. site.
365
Downloaded by [] on [10/01/19]. Copyright © ICE Publishing, all rights reserved.
366 POULOS
cedures has been the rapid development of (b) to outline a versatile analytical framework
powerful numerical techniques, in particular the which can incorporate several existing
finite element method and the boundary element methods of analysis
method. These, in conjunction with the now com- to compare alternative analytical procedures
monplace availability of powerful desktop com- to examine the characteristics of behaviour of
puters, have made feasible methods of pile single piles and pile groups under static axial
analysis which even a decade ago would not have loading, cyclic axial loading, and when sub-
been considered practical for foundation design. jected to externally-imposed soil movements
Several such analysis methods have been pro- (4 to review methods of obtaining the soil
posed and various claims have been made regard- parameters required for the prediction
ing the superiority of one or other of these (f) to present examples of simplified design
methods over alternative approaches. It would charts based on theoretical analysis and char-
therefore appear appropriate at this time to acteristic soil parameters
review some of these methods, examine simi- (9) to discuss the application of theory to practi-
larities and differences among them, and obtain cal cases, and to examine the sensitivity of
some insight into the significance of such differ- pile performance predictions to a variety of
ences when compared with uncertainties which factors over which the geotechnical analyst
may arise in the practical application of the has control, including the method of analysis,
methods. and the geotechnical characterization of the
This Paper therefore has the following objec- site.
tives
Attention will be confined to problems involving
(a) to present a classification of methods for pile static or quasi-static axial loading of piles and
foundation analysis and design pile groups in soil. Both load capacity and defor-
mation are considered, but it is emphasized at the Such procedures involve the use of simple compu-
outset that load capacity and deformation are not tational methods or design charts, and generally
independent and, although they can be separated do not demand the use of a computer. Category 3
with reasonable justification for many problems procedures involve the use of a site-specific
involving conventional direct loading, their inter- analysis based on relatively advanced analytical
dependence may be very important for problems or numerical techniques such as the finite element
involving cyclic loading or external soil move- method or the boundary element method. In
ments. most cases, such procedures require the use of a
computer. Category 3 procedures are frequently
used to carry out the necessary parametric solu-
CATEGORIES OF ANALYSIS AND DESIGN tions and develop the design charts which can
PROCEDURES then be used as category 2 solutions.
Analysis and design procedures can be divided Typical examples of the various categories of
into three broad categories, depending on the procedures for axially loaded piles are shown in
level of sophistication and rigour. An extended Table 2. In choosing an appropriate category of
classification system of these procedures has been design for a practical problem, the following
proposed by Poulos & Hull (1989) and is shown factors need to be considered
in Table 1. Category 1 procedures probably (a) the significance and scale of the problem
account for most pile design done throughout the (b) the available budget for foundation design
world. Category 2 procedures have a proper theo- (c) the geotechnical data available
retical basis, albeit simplified, and are being (d) the complexity of both the geotechnical
increasingly used for pile deflexion calculations. profile and the design loading conditions
(e) the stage of the design process (i.e. whether a variety of constitutive soil models can be uti-
feasibility, preliminary or final design is being lized, and such factors as soil non-
carried out) homogeneity and anisotrophy can be taken
(f) the experience of the designer with the into account.
methods being considered.
Finite element methods offer the most powerful
Small projects with a limited budget for geotech- analytical approach in that, not only can non-
nical work rarely justify more than a category 1 linear soil behaviour be modelled, but the com-
or 2A approach. However, for final foundation plete history of the pile can be simulated, i.e. the
design in a major project for which considerable processes of installation, reconsolidation of the
geotechnical data has been obtained, a category 3 soil following installation, and subsequent
approach would be appropriate. If, for the same loading of the pile (e.g. Nystrom, 1984; Withiam
project, a preliminary design was required based & Kulhawy, 1979). Such analyses are valuable in
only on limited in situ or laboratory data, a cate- leading to a better understanding of the details of
gory 2 approach might be useful for carrying out pile behaviour, but are unlikely to be readily
sensitivity studies to identify those parameters applicable to practical piling problems because of
which are most significant and need to be deter- their complexity and the considerable number of
mined with greatest care. geotechnical parameters required.
There will often be occasions in practice where A reasonable compromise between excessive
a category 3 analysis is required to account for complexity and unacceptable simplicity is provid-
some unusual aspect of the soil profile or the ed by boundary element methods, in which the
unconventional nature of the design loading (e.g. pile-soil interface is discretized and the character-
if it involves a significant cyclic component). It is istics of the soil response are represented in a
not unlikely that the parameters which are lumped form by ascribing the behavioural fea-
required for such a category 3 analysis will have tures of the soil to the interface elements. This
to be estimated rather crudely, e.g. from empirical method has been developed by a number of
correlations which would be more appropriate to research workers and is widely used in practice,
category 1 or 2 procedures. Consequently, it and attention will therefore be focused on this
cannot be expected that solutions from a category method. Although a considerable number of for-
3 analysis will always be superior to those from a mulations exist, most appear to have a common
simpler analysis. As will be illustrated in this underlying basis. A convenient means of
Paper, the method of analysis frequently is of developing a unified analysis is to employ a sub-
much less importance than the geotechnical structuring technique in which the pile (or piles)
parameters which are selected and the way in and the surrounding soil are considered separa-
which the geotechnical profile is idealized. tely and then compatibility conditions are
imposed.
The behaviour of each element is considered at
ANALYTICAL FRAMEWORK FOR AXIALLY a node which is located at the centre of that
LOADED PILES element and along a common vertical plane
In the past two decades, a number of methods through the pile axis. Fig. 1 shows the division of
have been developed for analysing axially loaded a single pile into elements, the distribution of free-
piles, many of which fall into category 3. Most of field soil movements due to some external cause
these methods involve the use of one (or more) of (e.g. swelling or consolidation of the soil mass due
the following analytical techniques to moisture changes or external loading) and the
specified distributions of
(a) simplified analytical methods involving the
consideration of independent horizontal (a) the limiting pile-soil stresses, f, for compres-
‘slices’ of pile and soil, e.g. Randolph & Wroth sing loading, and f, for tensile loading; for
(1978). shaft elements, the limiting pile-soil resistance
(b) boundary element methods, employing either will be termed here the shaft resistance, while
load-transfer functions to represent the inter- for base elements, the term end-bearing resist-
face response (e.g. Coyle & Reese, 1966; Kraft ance will be used
et al., 1981) or elastic continuum theory to (b) the local stiffness k, of the soil.
represent the soil mass response (e.g. Butter-
field & Banerjee, 1971; Banerjee, 1978; At this stage, no assumptions are made regarding
Banerjee & Davies, 1978; Poulos & Davis, the nature off,, f, and k,; these may vary with
1980). depth, stress (or displacement) level and with
(c) Finite element methods (e.g. Desai, 1974; Val- time.
liappan et al., 1974; Balaam et al., 1975; Otta- The responses of the pile and soil elements to
viani, 1975; Jardine et al., 1986), in which a an increment of axial load AP are analysed in
-I-c-
M
1
I 4 1 1 PJ
:sj
si
i
111
4db--
la) Geometry of lb1 TypIcal (cl Typical IdI ~~tSr$ution (el Dlstrlbutlon (fl Oistrlbution
Problem stresses Stresses of Free-Field of Limiting
Actmg on Actmg on Stiffness Soil Movement Pile-SOII
Pile Elements Soil Elements with Depth wth Depth
Fig. 1. Basic problem of single axially loaded pile: (a) geometry; (b) stresses acting on pile elements; (c) stresses acting
on soil elements: Cd1distribution of soil stiffness with deoth; (e) distribution of free-field soil movement with depth; (f)
distribution of liki&g pilesoil stress
turn. Details of the equation that describes incre- at element i, and fci is the limiting pile-soil inter-
mental displacements of the pile are given by action stress in compression, at element i.
Poulos (1979b). For the soil elements, the incre- Equations (l), (2a) and (2b) describe the behav-
mental displacement of the soil may arise from iour of the soil, and may be written for all ele-
two sources, the pile-soil interaction stresses ments of the system. By equating the expressions
{Ap}, and the free-field soil movements {AS,}. At for incremental pile and soil displacements,
any element i, the soil will be in one of three invoking the vertical equilibrium condition, and
states; non-failure, failure in ‘compression’, or checking the state of each element during the
failure in ‘tension’. loading increment, the incremental pile-soil
In the non-failure state, the incremental dis- stresses can be determined, and hence the pile dis-
placement of the soil at an element i, ApSi, can placement at each element can be evaluated.
then be expressed as Further details of the numerical procedure are
given by Poulos & Davis (1980). By successive
application of loading increments, the entire
ApSi = f !!L . Apj + Asei (1) load-displacement relationship for the pile may
j=l k,ij
be determined, including any unloading and
where lij is displacement influence factor for reloading sequences.
element i due to element j, ksij is soil stiffness for This analysis becomes less satisfactory as the
element i due to element j, Apj is the increment of pile becomes very short or its stiffness relative to
pile-soil interaction stress, A& is the increment the surrounding soil decreases. For pile sockets in
of external soil movement at element i and m is rock, the finite element method provides more
the total number of elements into which pile is reliable solutions, e.g. Pells & Turner (1979). The
divided. analysis in this section can be used for both the
In the ‘compressive’ failure state, the conditions load-transfer analysis (commonly termed the t-z
at element i may be expressed as analysis), in which the soil resistance is modelled
by a discrete spring at each element, and the con-
APi =.Li- Pi (24 tinuum analysis, in which the soil is modelled as
an elastic continuum.
and for the ‘tensile’ failure state, For the load-transfer or t-z analysis, the dis-
placement influence factor Iij in equation (1) is
APi=.Li-Pi (2b) zero, except when i = j, in which case Iii = 1; in
other words, the soil deflexion of a node i
where 6pi is incremental pile-soil stress on depends only on the stress at that node. The soil
element i, pi is the existing pile-soil interaction stiffness in equation (2) can be evaluated either
stress on element i before the load increment,& is empirically (e.g. Coyle & Reese, 1966) or theoreti-
the limiting pile-soil interaction stress in tension, cally (Baguelin & Frank, 1980; Kraft et al., 1981;
Ha & O’Neill, 1983; Chow, 1986; Ooi et al., IMPLEMENTATION OF THE THEORY
1989). Such features as soil non-homogeneity, In this Paper, attention will be concentrated on
non-linearity, post-peak softening, and interface the continuum analysis. With this approach, it is
dilation or contraction, may be incorporated. The only possible to derive proper results for quite
main limitation of the load-transfer analysis is idealized problems that involve linear soil behav-
that it can only properly consider a single pile, iour and one- or two-layer elastic soil masses.
and cannot be used directly to analyse pile The range of problems for which proper results
groups. It is possible to modify the single pile t-z can be derived can be extended by using the finite
curves to allow for group effects (Randolph, 1986) element method to evaluate the soil displacement
but it is usually necessary to evaluate these group influence factors. However, it is also possible to
effects using elastic continuum theory (e.g. Chow, utilize continuum theory in an approximate
1986). manner to obtain solutions for a much wider
In the continuum analysis the displacement range of problems of practical interest.
influence factor lij (in equation (1)) for all ele- Figure 2 gives an indication of the extensions
ments is non-zero, and is usually evaluated from to the theory which may be accomplished by
elastic continuum theory. The theory of Mindlin using the continuum analysis. The extensions are
(1936) is frequently used for this evaluation (e.g. classified into six groups, two of which involve
Poulos & Davis, 1968), although other basic the modelling of the soil, two of which involve the
elastic solutions may be employed, for example, modelling of the piles, and the last two which
the equations of Chan et al. (1974) for a layered involve the modelling of the loading on the piles.
half-space, as utilized by Chin (1988).
An implicit assumption in the elastic contin-
uum analysis is that, if incremental tensile stresses Modelling the soil projile
are developed in the soil mass by the loading of It is convenient to use the theory of elasticity
the pile, the soil continues to respond elastically. for modelling the soil behaviour. Despite the
Such an assumption is reasonable if the overall gross simplification which this model involves
stress conditions remain compressive (due to the when applied to real soil, it provides a useful
compressive overburden stresses), but may be basis for the prediction of pile behaviour, provid-
questionable if the overall stresses become tensile. ed that appropriate equivalent elastic parameters
Unease about this point has led to some criticism are selected for the soil. A significant advantage of
of elastic continuum theory (e.g. Leonards & using an elastic model for soil is that it provides a
Darrag, 1989) and a preference for the load- rational means of analysis of pile groups and
transfer approach, as the latter involves no con- evaluation of immediate and final movements of a
sideration of the stress state in the soil other than pile. In determining immediate movements, the
at the pile-soil interface. However, load-transfer undrained elastic parameters of the soil are used
and continuum solutions for a single pile gener- in the theory, whereas for final movements the
ally agree well and suggest that any effects of pos- drained parameters are used.
sible tension in the soil are unlikely to influence Mindlin’s equations may be used to obtain
pile behaviour significantly. approximate solutions for a layer of finite thick-
BASIC ANALYSIS
-Single pile in homogeneous
Modifications
Soil Behaviour
Soil Profile l non-linear Pde
Group Effects Loading Conditions
* finite layer rWpCl”W Characteristics
* two piles + residual loads
l bearing stratum * strain-softening * non-uniform * negative friction
* general l cyclic
* non-homogeneity * creep shaft diameter l expansive soil
groups l dynamic
* layering * cyclic loading * enlarged base
response
Fig. 2. Basic analysis, and extensions to allow for more realistic modelling
ness by using the Steinbrenner approximation between soil shear stress and shear strain, in
(Steinbrenner, 1934) to allow for the effect of the which case the tangent Young’s modulus of the
underlying rigid base in reducing the soil dis- soil E,,, is given by
placements. For piles bearing directly on a stiffer
stratum, the reduction in soil displacement may E
be analysed approximately by using the mirror-
image technique described in Poulos & Davis
(1980). In both these cases, the soil displacement where E, is the initial tangent Young’s modulus,
factors lij in equation (1) are modified. R, is the hyperbolic curve-fitting constant, p is the
In the elastic continuum theory, the soil pile-soil stress, pr is the limiting value of pile-soil
stiffness kij in equation (1) is directly related to stress (Pr = f,in compression, 0rJ in tension).
the Young’s modulus of the soil. In a homoge- It is also possible to use alternative models to
neous soil mass, ksij = EJdj where E, is the describe non-linear behaviour, for example, multi-
(constant) Young’s modulus and dj is the dia- linear models, or the Ramberg-Osgood model.
meter of element j. If a non-homogeneous soil Regardless of which non-linear model is used, it is
mass is considered and Mindlin’s equation is used important to consider unloading and reloading,
to evaluate the displacement influence factors, an as well as initial loading. The simplest assumption
approximation is required to determine ksij. is to assume that on unloading or reloading, the
Poulos (1979a) has examined alternative assump- behaviour is elastic until the previous greatest
tions and concluded that the following approx- stress level at that element is reached; thereafter,
imation is generally satisfactory non-linear initial loading resumes. However,
alternative assumptions are possible in which
ksij = 03(E,, + E,,)/dj (3) non-linear behaviour recommences at stress levels
less than the previous greatest level (Randolph,
where Esi and Esj are the values of soil Young’s
1986). Such behaviour can give rise to the accu-
modulus at elements i and j respectively.
mulation of permanent displacements if repeated
Equation (3) becomes inaccurate if large differ-
loading is applied, but additional parameters are
ences in soil modulus exist between adjacent ele-
then required in the analysis.
ments or if a soil layer is overlain by a much
In analysing the behaviour of piles subjected to
stiffer layer. In a subsequent investigation, Yama-
cyclic loading, at least three aspects of soil
shita et al. (1987) have developed a more general
response should be considered,
approximation in which Esi and Esj in equation
(3) are taken as weighted average values of soil (a) the degradation of pile-soil resistances (and
modulus, involving the value of soil modulus at possibly soil modulus) under repeated cyclic
all elements along and beneath the pile. loading
(b) loading rate effects
(c) the accumulation of permanent displacements.
Modelling of soil behaviour
One possible approach to incorporating cyclic
One of the more important features of pile-soil
loading effects has been described elsewhere by
behaviour is the limited stress which can be devel-
the Author (Poulos, 1988b). The effects of cyclic
oped between the soil and the pile. This feature is
degradation are most significant for shaft resist-
already incorporated into the analysis in that ance, and may be conveniently quantified by
checks are made to ensure that the pile-soil shear
means of a degradation factor Dr, where
stress does not exceed the specified limiting value.
If the interface behaviour at an element is shaft resistance at an
assumed to be perfectly linear until the limiting element after cyclic loading (5)
stress is reached, the interface model will be D, =
termed (conveniently, but not strictly correctly*) shaft resistance at that
an ‘elastic-plastic’ interface model. element for static loading
It is possible to incorporate non-linear behav-
iour of the soil, in an approximate manner, by The model of Matlock & Foo (1980) is useful for
assuming that the soil Young’s modulus varies quantifying the change in Dr as cycling proceeds,
with either stress or strain level. The simplest and involves only two parameters, a minimum
assumption is of a hyperbolic relationship degradation factor Dlim and a degradation rate
parameter 1.
The effects of loading rate are incorporating by
* Such a model does not necessarily imply that the
relationship between interface stress and displacement means of a loading rate factor D,, while per-
at an element is elastic-plastic; because of the contin- manent displacements developed in the soil
uum nature of the soil, the slope of this relationship will during cyclic loading are considered to be equiva-
change as other interface elements reach a failure state. lent to cyclic loading-dependent external soil
movements. Other aspects of soil behaviour may (4 A complete analysis of the group can be per-
also be included, for example, strain-softening formed, using a load-transfer analysis to
interface behaviour (Randolph, 1983; Kraft et al., determine the response of a pile to its own
1981) and creep (Booker & Poulos, 1976). load, and continuum theory to determine the
influence of the elements of the other piles-
this has been termed the hybrid method and
has been used by O’Neill et nl. (1977) and
Pile characteristics
Chow (1986).
The analysis can readily accommodate piles of
(4 Solutions for a two-pile group (obtained from
varying diameter or varying stiffness within the
a continuum analysis) can be used to obtain
formulation of the pile displacement equation. In
interaction factors that express the increase in
the case of varying diameter, additional annular
head settlement of a pile due to the presence
elements are introduced at the diameter discon-
of another pile (Poulos, 1968; Randolph &
tinuities, and values of the limiting pile-soil resist-
Wroth, 1979). The interaction factor o! is
ance at these elements, for both compression and
defined as
tension loading, must be specified.
It is also possible to allow for failure of the pile u = AS/S, (6)
material itself (either in compression or tension)
by limiting the axial stress that can be developed where AS is the increase in settlement of a pile
in the pile section. The procedures for making due to the presence of another equally loaded
this modification are described in Poulos & Davis pile and S, is the settlement of a single pile
(1980). under its own load. For a group of piles, the
interaction factors may be superposed to
develop a set of equations relating the settle-
ment of each pile to the single pile settlement,
Pile groups the load on each pile head, and the inter-
The continuum-based analysis can readily be action factors.
extended to analyse a group of axially loaded A modified form of the interaction factor
piles. Each pile is discretized into elements, soil method can be employed in which different
and pile displacement equations are assembled values of soil Young’s modulus are used to
for each element, compatibility of soil and pile determine the single pile behaviour and the
displacements is imposed at elements in the non- interaction factors. Higher values are gener-
failure state, and the vertical equilibrium equation ally used for the latter to reflect the lower
for each pile is written. In addition, the pile head level of strain (and hence the greater stiffness)
conditions must be specified. Usually, a rigid cap of the soil between the piles as compared with
connects the piles, so that all piles will undergo the soil immediately adjacent to each pile.
an equal head displacement. By specifying this This method is termed the modified inter-
condition, a further set of equations is obtained, action factor method, and has been described
which enables the load increment on each pile in detail by Poulos (1988a).
head to be computed, in addition to the distribu-
tion of incremental pile-soil stress and displace- Comparisons between these approaches will be
ment (Hewitt, 1988). discussed later in the Paper.
In the evaluation of the displacement influence
factors lij, use may again be made of the elastic
solution of Mindlin (1936) to determine the dis- Loading conditions
placement of an element of a pile due to all ele- Residual stress effects. Most analyses of pile
ments of that pile and the other piles in the response assume an initially stress-free pile,
group. A convenient approximation for this although it is well-recognized that residual
evaluation has been developed by El-Sharnouby stresses exist in piles, due to installation effects,
and Novak (1985) which avoids the need for particularly in driven piles. Ideally, the driving of
double integration of the Mindlin equation and a pile should be modelled using a dynamic
hence reduces the computation time substantially. analysis such as that used by Holloway et al.
A number of methods of implementing the above (1978). However, a simpler first approximation
analysis have been developed, including the fol- can be employed in which the pile (at final
lowing. penetration) is loaded to failure in compression
and then unloaded back to zero load. Poulos
(a) A complete analysis of the group is performed (1987) shows that this procedure gives residual
using a continuum analysis-this has been stresses in the pile which appear to be reasonably
done by Banerjee & Driscoll (1976) and realistic. Leonards & Darrag (1989) point out
Poulos & Hewitt (1986). that assumptions made regarding the soil
modulus near the pile tip can significantly influ- interest to compare the solutions with an inde-
ence the computed residual stresses in the lower pendent analysis method such as the finite
part of the pile, particularly for piles in sand. element method. Therefore, this section will make
Cyclic loading. At least two methods can be such comparisons for three different problems: (a)
adopted to simulate cyclic loading of a pile a single axially loaded pile; (b) an axially loaded
pile group; and (c) a pile subjected to downdrag
(a) a single-step analysis, in which a single iter-
forces by external vertical soil movements.
ative analysis is carried out to determine the
behaviour of a pile after a specified number of
cycles, N
Single pile
(b) a cycle-by-cycle analysis, in which the applica-
For a single pile having a typical relative com-
tion of each of the N cycles is modelled in
pressibility K = EJEs of 1000, (where E, is the
turn.
Young’s modulus of pile and E, is Young’s
Details of these methods are given by Poulos modulus of soil) in a deep homogeneous elastic
(1983). The cycle-by-cycle analysis is considerably soil mass, Fig. 3 compares two sets of solutions
more time-consuming, but much more versatile, for the settlement as a function of the length-to-
in that it can accommodate strain-softening inter- diameter ratio L/d, the boundary element
face behaviour and sequences of irregular cyclic approach, and the approximate analytical
loading. approach developed by Randolph & Wroth
External soil movements. The effects of external (1978). For L/d > 15, the agreement is close, the
soil movements, such as those arising from soil difference being less than 10%. For smaller values
consolidation due to external loading or of Lfd, the Randolph 8z Wroth solution gives
dewatering, or from soil heave due to wetting of smaller settlements, possibly due to the effects of
expansive clay layers, can be analysed directly, the assumption of the soil surrounding the pile
provided that the external free-field soil move- being a series of concentric cylinders. A similar
ments at each element (i.e. values of ASei in equa- measure of agreement is found between the
tion (1)) can be specified. In the case of Author’s solution and the finite element solutions
one-dimensional loading of a clay layer, these of Valliappan et al. (1974).
movements may be determined from Terzaghi’s Figure 4 shows various solutions for the settle-
one-dimensional consolidation theory, and the ment of a single pile in a Gibson soil layer. There
time-dependency of the pile behaviour can readily is reasonable agreement among the three solu-
be determined (Poulos & Davis, 1980). The free- tions compared but as in Fig. 3, the greatest
field movements in an expansive soil layer are less potential for differences appears to be for rela-
easily computed, although a number of empirical tively short piles (L/d < 15).
or semi-empirical methods exist (e.g. Blight, 1965; In comparing non-linear solutions for single
Van Der Merwe, 1964; Cameron & Walsh, 1984). pile response, the problem considered is that
analysed by Jardine et al. (1986). They have
employed a finite element analysis involving the
Computer codes use of a non-linear soil model, the LPC2 model,
The preceding analyses almost invariably
require computer evaluation, and a number of
04
computer programs have been written
examining various aspects of axial pile behaviour.
for I
’ IP
I I
I
A selection of these is given in Table 3. This list is
by no means exhaustive and there is no doubt a
great number of other codes in existence. Also
excluded from this list are codes based on finite
element analysis.
0 20
0 15
\, dj2LL L
element analyses are capable of predicting a very
similar load-settlement response to that from a
category 3C non-linear finite element analysis.
Method Values of I,
K = 30 K=30000 K = 30 K=30000
,--- 0.15 I I I
h/L==
f lo 4,
:5
I I I
0
1 2 5
(No. of ‘Piles,“’
0 2 4 6 0 10
Settlement of Pile Head (mm) IbI Infinite Soil Layer
Fig. 5. Comparison between various non-linear analyses of single pile Fig. 6. Comparison between solutions for group settlement in Gibson soil (a) finite soil
load+&tlement behaviour: (a) EP = 3 x 10’ MN/m’; (b) Ep = 3 layer; (b) infinite soil layer
x 10’ MN/m2
PILE BEHAVIOUR-THEORY AND APPLICATION 311
interaction factors thus obtained are in close head displacement with time as the soil layer con-
agreement with the values of Poulos & Davis solidates. Also shown is the solution obtained by
(1980) determined from the boundary element Poulos & Davis (1980) from a boundary element
analysis. analysis, and it can be seen that, for a value of
O’Neill & Ha (1982) have compared the behav- drained Poisson’s ratio of the soil of 0.3, the
iour of pile groups predicted by the program agreement between the two solutions is close.
DEFPIG with that predicted by a kind of hybrid There are, however, some differences in the detail
analysis implemented by means of the program of the developed shear stresses along the pile
PILGPI. It is found that the two programs give shaft, especially near the soil surface, possibly
comparable predicted behaviour, although differ- because of discretization inaccuracies in both
ent values of soil Young’s modulus are required analyses. Other independent analyses of the same
to give exact numerical agreement. These com- general problem by means of continuum-based
parisons reinforce the fact that the value of soil boundary element analyses (e.g. Chin, 1988; Kog
modulus is not unique but must be selected care- et al., 1986) are in general agreement with the
fully for use with the method of analysis Poulos & Davis solutions.
employed.
In summary, while the foregoing comparisons
demonstrate some differences among various
CHARACTERISTICS OF PILE BEHAVIOUR
methods, they also indicate generally satisfactory
In this section, some of the more significant
agreement between the boundary element solu-
characteristics of pile behaviour will be itemized.
tions using interaction factors and solutions from
Most have been derived from theoretical analyses
the other approaches. Finite element analyses can
but in many cases are also supported by measure-
be illuminating in that they reveal detailed behav-
ments made from laboratory and field tests.
ioural characteristics, but it would appear that
Unless otherwise stated, the solutions described
adequate practical predictions of group settle-
will have been obtained from boundary element
ment can be obtained from simpler approaches
analyses based on elastic continuum theory, with
based on boundary element analysis.
an elastic, or an elastic-plastic interface model,
using simplified distributions of soil Young’s
modulus and shaft resistance with depth (either
Single pile subjected to vertical soil movements
constant or linearly increasing).
Small (1988) has used the finite element method
Four aspects of behaviour will be considered
to analyse a single end-bearing pile resting on a
rigid base, in a soil layer which is subjected to (4 the load-settlement behaviour of a single pile
external loading so that vertical soil movements under static axial loading
occur and downdrag (or negative friction) forces (b) the settlement of pile groups under static axial
are developed in the pile. The problem is illus- loading
trated in Fig. 7 which shows the finite element piles subjected to external soil movements
solution obtained for the development of pile ii piles subjected to cyclic axial loading.
L/d=25
K=lOOO -
yL/q=O.5
.-e 0.6
k
d P
.- (a)End-Bearing
-A 5
5
vr
x u4
(blNoatmg
0 20 40 60 80 100
L/d
Fig. 9. Relative settlement of end bearing and floating pile. Homogeneous soil
prrdL/P
L/d=25
E,/E,=lOOO
v,=o.s
P/P P
P
Modulus
Distribution
The settlement and load transfer are influenced is in excess of 085, and is almost independent of
by the distribution of soil Young’s modulus along L/d. This theoretical conclusion is supported by
the pile shaft. An example is given in Fig. 10 for a the results of model tests on brass piles in kaolin,
typical friction pile in two different soils: a homo- reported by Mattes & Poulos (1971) and field
geneous soil, and a Gibson soil in which the soil maintained loading tests, e.g. Whitaker & Cooke,
modulus increases linearly with depth, from zero 1966. A corollary to the above observations is
at the soil surface. For the homogeneous soil, the that the rate of consolidation settlement of piles
distribution of shear stress 7 is relatively uniform in clay is not likely to be an important consider-
with depth, whereas for the Gibson soil, r ation in design. Time effects stemming from creep
increases with depth. The similarity between the at higher load levels are likely to be more impor-
stress distribution and the distribution of soil tant than consolidation time effects (Edil &
Young’s modulus may explain why load transfer Mochtar, 1988).
approaches can give reasonable predictions of At normal working loads (of the order of
pile behaviour. For the same average value of 4&50% of the ultimate load), non-linear behav-
Young’s modulus along the pile shaft, the pile iour of the soil generally does not have a substan-
head settlements in this case are reasonably tial influence on pile settlement. Careful mode1
similar, and differ by only about 7%. However, tests reported by Butterfield & Abdrabbo (1983)
the tip settlement for the pile in the Gibson soil is support this theoretical conclusion.
about 18% less than that of the pile in the homo- For a typical pile in a homogeneous soil, Fig.
geneous soil. 11 gives some indication of the potential influence
The major part of the final settlement of a of non-linear soil behaviour on settlement. The
single pile is immediate settlement, and occurs on ratio of settlement of the pile in a purely elastic
application of the load because the load is trans- soil Selas to the settlement of the pile in a soil with
ferred to the soil essentially by shear, with rela- hyperbolic response S, is plotted as a function of
tively little change in mean stress. For practical the load level P/P,, where P, is the ultimate load
values of vS’(of the order of 0.3-0.4) and L/d, the capacity. The initial tangent Young’s modulus of
theory suggests that the ratio SJS,, of immediate the hyperbolic soil model is assumed equal to the
to final settlement of a pile in a homogeneous soil modulus of the purely elastic soil. S,,,,j3 is gener-
0.8
L/d=25
K=lOOO Note: Elastic settlement
_ Homogeneous Soil Selas determined
0.2
Esi/fs=SOO from initialtangent
modulus E,i
OL I 1 I I I
0 0.2 0.1, 0.6 0.8 1.0
P/P"
ally less than unity (indicating that non-linearity oped at the tip. Resistance to applied tensile
results in an increase in the settlement), the extent loading comes primarily from the pile tip, the
of this increase depending on the hyperbolic response of which is less stiff than the pile shaft;
parameters adopted. For the parameters con- therefore, the pile head movements will be greater
sidered to be most realistic (curves c and d), non- than if compression is applied.
linearity causes an increase of between 11 and If the pile-soil interface can strain-soften, both
25% in the settlement, as compared with the set- the load-settlement behaviour and the ultimate
tlement determined from a purely elastic analysis load capacity will be affected. The ultimate load is
using the initial tangent Young’s modulus of the no longer statically determinate, but will depend
soil. It should be noted that, in this case, the non- on the relative stiffness of the pile, the ratio of
linearity arising from pile-soil slip is only signifi- peak to ultimate resistance, and the post-peak
cant at load levels close to failure and therefore behaviour. Randolph (1983) presents solutions for
that the use of elastic theory with an appropri- a reduction factor to be applied to pile shaft
ately reduced secant modulus would give a rea- capacities based on peak values of shaft resist-
sonable prediction of settlement. However, for ance, in order to allow for the effects of progres-
slender compressible piles, pile-soil slip may have sive failure along the pile due to strain-softening.
a more dominant influence on the non-linearity of
the load-settlement response.
No initial residual stresses
Residual stresses that remain in the pile after ---- With inihal residual stresses
installation may influence the pile head stiffness
and hence the calculated pile head movements.
The theoretical analysis of Poulos (1987) has
demonstrated that, for cases in which significant
residual stresses remain (e.g. for a driven pile in
dense sand), the stiffness of the pile head in
tension may be smaller than in compression. Fig.
12 illustrates the point for an elastic-plastic inter-
face. If no account is taken of residual stresses,
the pile head stiffness in tension and compression 0 so 100 0 20 40 60
is the same, but if residual stresses are allowed Settlement Imm) Movement (mm1
for, substantially larger movements can occur in
la) Compression Loading lb) Tension Loadmg
tension than in compression. Under zero net load,
the residual stresses are such that tensile pile-soil Fig. 12. Effect of residual stresses on load-deflection
slip occurs over a significant amount of the pile behaviour pile in sand (Poulos, 1987): (a) compression;
shaft, and a residual compressive load is devel- (b) tension
These solutions reveal that reductions in the peak pile characteristics, and those related to the
load capacity are only likely to be significant for geometry of the piles and the pile group. The
relatively long compressible piles. important soil and pile characteristics are the pile
stiffness factor K, the ratio of EJE, of Young’s
moduli of bearing stratum to soil, and the dis-
Pile group under static loading tribution of the soil Young’s modulus E, with
Before the development of modern analytical depth. Fig. 13 illustrates the influence of these
techniques, it was commonly believed that no factors on the two-pile interaction factor c(. It
rational relationship existed between the behav- may be seen that cxdecreases as K decreases, or as
iours of single piles and pile group. In his James E,,/E, increases, or as the distribution of soil
Forrest Lecture, Terzaghi (1939) stated Young’s modulus becomes less uniform with
Both theoretical considerations and experience leave depth. Consequently, it should be expected that
no doubt that there is no relation whatever between early published solutions for interaction factors,
the settlement of an individual pile at a given load which are for a rigid pile (K = co) in a homoge-
and that of a large group of piles having the same neous mass (E,/E, = 1 and E, constant with
load per pile.
stiffness of group
R, = (9)
sum of individual pile stiffness
settlement ratio R, and a more uniform distribu- by Poulos & Mattes (1971), in which the group is
tion among the piles, than is predicted by the replaced by an equivalent block. This solution
conventional analysis. agrees well with the solution computed from the
The relative proportion of immediate settle- interaction factor method.
ment Si to final settlement S,, decreases as the The effect on group settlement of the pile cap
size of the group increases. For groups of piles in being in contact with the soil is relatively small
either a homogeneous soil, with a typical value of unless the pile spacing is large and the group is
v,’ of 0.3, the ratio SJS,, decreases from 0.93 for relatively small. Even for piles at an unusually
a single pile to 0.85 for a 25-pile group. Similar large centre-to-centre spacing of 10 diameters, the
values are found for groups in a Gibson soil. The reduction in settlement due to cap contact is only
pile stiffness factor K has little influence on about 5%. Therefore, for most practical purposes,
si/sTF. For pile groups, the consolidation settle- the influence of pile cap contact on settlement at
ment (and hence the rate of consolidation working loads can be ignored.
settlement) is more important than for a single
pile, but it is still likely to be the minor com-
ponent of settlement unless soft compressible Piles subjected to external soil movements
layers exist beneath the pile tips. There are several circumstances under which
The stiffness of underlying soil layers may have loading may be induced in piles by external soil
a significant influence on pile group interaction movements, but attention will be confined here to
and settlement. Fig. 16 shows an example of the two problems
influence of the relative stiffness of the underlying (a) end bearing piles subjected to negative friction
soil on the settlement of a pile group. Clearly, the by settlement of the surrounding soil; the soil
presence of a softer layer (I&,/& < 1) may sub- movement is assumed to occur over the entire
stantially increase the settlement, as compared depth of the soil layer, and to vary linearly
with the case of a homogeneous soil mass with depth
(E,/E, = 1). Also shown is the settlement com- (b) floating piles in a relatively stiff expansive
puted from the approximate approach described clay, subjected to swelling or shrinking move-
0.20
0 15
L/d=25
h/L=l.S
s/d=3
U
? v~=vS=o.5
w” 010 E/E,=1000
=I
G2 Group
0 05
0
01 1 10 100
Et/Es
Fig. 16. Effect of modulus of underlying stratum on group settlement
merits; these movements are generally In contrast to the case of conventional axial
assumed to vary linearly with depth, but to loading, the development of pile-soil slip is
extend to only a limited distance below the advantageous. It leads to reduced pile movement
surface. and downdrag forces as compared with the
purely elastic interface condition. A typical
example is shown in Fig. 17. Pile-soil slip starts
Negative friction on end-bearing piles when the soil surface movement is about lOmm,
Solutions from purely elastic theory show that and is almost complete at about 25 mm. Elastic
the movement and downdrag force induced in the theory would seriously overestimate the pile force
pile depend on the dimensionless parameters L/d, and movement for soil movements in excess of
K and EJE, (Poulos & Davis, 1980; Chin, 1988). about 20 mm.
However, slip at the pile-soil interface generally Because of the dominant influence of pile-soil
plays a dominant role in pile behaviour. Poulos slip, the soil model used in the analysis of nega-
& Davis (1980) present solutions that indicate the tive friction generally does not have a major effect
circumstances under which slip occurs along vir- on the solution. Fig. 18 shows a typical example
tually the entire pile shaft. For typical normally in which three models, the elastic-plastic contin-
consolidated or lightly-overconsolidated soils, if a uum, the elastic-plastic t-z model, and the hyper-
surface pressure q is applied to the soil, full slip is bolic t-z model, have been used, with the same
likely to occur if values of initial tangent Young’s modulus and
shaft resistance. Before the development of full
q > (0.3 - 0.5)y’L (12) slip, the continuum model tends to give values of
where y’ is the submerged unit weight of soil and downdrag force and pile movement which are
L is the embedded length of pile. about 30% smaller than the other two methods,
The soil surface movement required to develop but as pile-soil slip develops, the differences
full slip depends on the relative stiffness of the become less. It will also be observed that the non-
pile and the distribution of soil modulus and shaft
resistance with depth. Values typically range I I
25
.=
al
-0 10 20 30 40 50
0.
Soil Surface Settlement 5, (mm) Ib) Downdrag Force Distribution
Fig. 17. Influence of pile-soil slip on downdrag force and Fig. lg. Influence of soil model on computed downdrng
pile movement force in end-bearing pile
L
For a given length of pile, the diameter has
relatively little effect on the movement of the pile.
Fig. 20 shows an example of the variation of pile
head movement and maximum tensile stress in
the pile as a function of diameter. Solutions for
both a purely elastic interface, and with consider-
ation of pile-soil slip, are shown. The elastic solu-
tions are conservative and give larger movements
and stresses than the solutions incorporating slip.
For pile diameters between about 0.2 and 2.0 m,
the pile head movement is almost constant when
2.0
pile-soil slip is accounted for. The maximum
tensile stress in the pile increases with decreasing
Fig. 19. Elastic solutions for pile movement in expansive diameter, but does not reach a significant level
soil-uniform pile diameter until the diameter is less than 0.2 m. The impor-
e+
g
D
:
J
0.8
0.6 hIi\Unstable
L
7nn 200
MPa
ES max
1=100cycles
100
I Stable
I I I
Pile Length L (ml
OL I \
Fig. 23. Effect of pile length on stability diagram zone
0 0.2 0.4 0.6 0.8 1.0 boundaries. Driven pile in clay (Poulos, 1988b)
Normalised Mean Load P,/Cl,
Fig. 22. Typical cyclic stability diagram for offshore ductile cyclic behaviour, while short stiff piles
pile; N = 100 cycles (lee, 1988) show a brittle cyclic behaviour.
Group effects may influence the cyclic behav-
iour of the piles, but to a relatively minor extent.
It has been observed that two-way cyclic Hewitt (1988) has performed analyses of a single
loading (about zero mean load) has a more severe pile, and four- and eight-pile groups, using the
effect on piles than one-way cyclic loading (for Matlock & Foo model for degradation of shaft
which the minimum load is zer+i.e. the cyclic resistance. For two-way loading of relatively stiff
load equals the mean load). This observation can piles, these results are shown in Fig. 24 together
readily be verified from the cyclic stability with laboratory data on model piles in kaolin.
diagram. Referring to Fig. 22, for two-way cyclic The theory shows that the maximum cyclic load
loading, the unstable zone for N = 100 cycles
begins at a normalized cyclic load of about 0.57;
this is also the maximum load that can be sus-
tained under two-way cyclic loading for 100
cycles. For one-way cycling (P, = P,), the
unstable zone is reached when PJQ, = 0.46, so
that the maximum load that can be sustained
under one-way loading conditions is 0.92 Q,,
substantially greater than the value of 0.57 Q, for
one-way loading.
As piles become shorter or stiffer, the positions
Model Test Data (Hewitt, 1988)
of the boundaries between the various zones
alter; both the stable and unstable zone bound-
at‘ies increase, but the metastable zone shrinks. l Failure within 10 cycles
Fig. 23 shows the effect of pile length on the zone 0 No failure at 10 cycles
boundaries for a 1.5 m dia. driven steel tube pile
in clay (Poulos, 1988b). For long compressible 25mm dia. aluminium piles
piles, there is a large metastable zone. Such piles in overconsolidated kaolin
exhibit a gradual decrease in load capacity as the
cyclic load level is increased. In contrast, short
stiff piles have a very limited metastable zone, and Number of Piles in Group
may fail abruptly after small increases in cyclic
load above the stable zone. It may thus be con- Fig. 24. Influence of number of piles on failure under
sidered that long compressible piles exhibit a two-way cyclic loading
that can be sustained decreases as the number of stress path which the soil adjacent to the pile
piles increases. The experimental results for 10 follows. Laboratory model pile tests may over-
cycles are in good agreement with the theoretical come this deficiency to some extent, but may not
curve. accurately reflect the behaviour of prototype piles
because of the presence of scale effects, particu-
larly for piles in sand. Some potential exists for
ESTIMATION OF GEOTECHNICAL more sophisticated tests such as the constant
PARAMETERS normal stiffness (CNS) direct shear test (Johnston
The most significant parameters required for & Lam, 1984; Ooi & Carter, 1987), and this type
many of the category 2A, 2B, 3A and 3B analyses of test has been used in the design of grouted piles
of pile behaviour under static loading are as in offshore carbonate sediments (Johnston et al.,
follows 1988). However, the direct utilization of labor-
atory tests for pile design is infrequent in practice,
(a) the shaft resistancef,
and still requires further research before it can be
(b) the end bearing resistance&
applied with confidence.
(c) Young’s modulus of the soil E,
The most reliable means of determiningf, and
(d) Poisson’s ratio of the soil v,
E, is by backfiguring from the results of pile load
(e) the hyperbolic curve-fitting constant R,
tests. Methods for interpreting the pile load test
For prediction of cyclic axial response, a number data have been detailed by Poulos & Davis (1980)
of other parameters are required, and some and Stewart & Kulhawy (1981), among others.
limited data on these is presented by Poulos Such methods are particularly effective if the pile
(1988~). Attention will be concentrated here on is instrumented so that details of the load transfer
the above parameters, and in particular on f,, fb along the pile shaft are available; it is then pos-
and E,. sible to determine detailed distributions of soil
For calculation of axial pile load capacity, f, modulus and limiting pile-soil friction along the
and fb must be estimated as accurately as pos- pile shaft.
sible. For the calculation of settlement resulting A variant of the pile loading test is the pile
from direct axial loading, the theoretical solutions section test. This involves the testing of a series of
reveal that the choice of an appropriate value of relatively short rigid sections at different depths,
E, is generally crucial, unless the piles are long in order to determine the distribution of limiting
and compressible. For piles in soil subjected to pile-soil friction and soil modulus with depth. In
external movement, the pile behaviour is gener- such tests, the sections must be installed in a
ally much less dependent on E, and, provided similar manner to the prototype pile in order to
that the soil movement is known,* an approx- obtain appropriate data. Examples of section
imate estimate of E, may be adequate, although tests have been reported by Hyden et al. (1988),
reasonable estimates of shaft and end-bearing and Williams & van der Zwaag (1988); in both
resistance are desirable. cases, tests were carried out on grouted pile sec-
tions in marine calcareous sediments.
In most practical situations, it is not possible to
Methods of determining parameters carry out such testing, at least in the early stages
For evaluating the parameters for static pile of design. Resort is frequently made to correla-
response, a number of methods can be contem- tions between the pile design parameters and
plated, including laboratory or field test data.
(a) laboratory testing
(b) appropriate interpretation of field pile load
Shaft resistancef,
tests
Tables 5 and 6 summarize available methods
(c) empirical correlations with laboratory-
for determining the shaft resistancef, from labor-
determined parameters
atory strength data, for both driven and bored
(d) empirical correlations with the results of
piles. Effective stress approaches can be used for
in-situ test data.
all soil types, whereas a total stress approach is
Conventional laboratory tests, such as triaxial or still adopted commonly for piles in clay. The
oedometer tests, are generally not suitable for parameters tl and /i’ (or K and 6) are usually
direct measurement of the soil Young’s modulus obtained from empirical correlations, despite the
as they do not follow, even approximately, the fact that the effective stress /3 approach is funda-
mentally sound and falls into category 2.
* The soil movement is treated here as an independent A summary of some suggested correlations
variable, although it will be influenced to some extent between f, and the standard penetration resist-
by the soil modulus. ance N are given in Table 7. Considerable varia-
Table 5. Shaft resistancef, for driven piles, determination from laboratory strength data
tions occur in these correlations, particularly for sleeve resistance of the penetrometer. Corrections
bored and cast-in-place piles. are applied, depending on soil type, pile type,
Figures 25 and 26 show values off, correlated relative pile length, and depth below the surface.
with static cone resistance q,. These relationships Robertson et nl. (1985) have found the method
have been developed by the Author from the cor- proposed by Schmertmann to provide a more
relations suggested by Bustamante & Gianeselli reliable prediction than the direct correlation to
(1982) and cover a wide range of pile types in q, , when applied to piles in a clayey silt.
both clay and silica sand. The classification of Extensive correlations have been developed in
these pile types is shown in Table 8. It should be France between shaft resistance and the limit
emphasized that several other correlations have pressure p, deduced from pressuremeter measure-
been proposed and that wide variations exist ments. Based on the results of over 300 pile load
between some of these. Fig. 27 shows an example tests at more than 100 sites, Bustamante et al.
of this variability, for driven piles in silica sand. (1987) have proposed correlations, similar in
The potential inaccuracy of shaft capacity predic- nature to those with the CPT in Figs 25 and 26.
tion using category 1 correlations, especially for Correlations such as those outlined must
loose sands, is clearly demonstrated. always be employed with caution, as a number of
Schmertmann (1975; 1978) proposes a different other factors may also influence shaft resistance
approach to the utilization of cone data, whereby e.g. the presence of overlying layers (Tomlinson,
the pile shaft resistance is related to the measured 1977).
Table 6. Shaft resistnncef, for bored piles, determination from laboratory strength data
K/K, = 213 to 1; K, is
function of OCR; 6 depends
on interface materials
“0 2 L 6 8 10 12 14 16
Cone Reststance qc IMPal Cone Resistance qc (MN/m21
Fig. 25. Design values of shaft resistance for piles in clay Fig. 26. Design v&es of shaft resistance for piles in
(based on Bustamante & Giaoeselli, 1982) sand (based on Bustamante & Gianeselli, 1982)
Table 7. Correlations between shaft resistaacef, and SPT value, withf, = (I + BN kN/m*
small changes in 4’ can theoretically lead to These correlations contrast with procedures such
large changes in N,, although the effects of as those proposed by Belcotec (1985) and De
soil compressibility are then more important Ruiter & Beringen (1979), in which a factor of
and may reduce the dependence of N, on 4’. unity is applied to the average value (computed
differently than in the Bustamante & Gianeselli
Table 10 shows some empirical correlations approach). However, the latter approaches are
between f, and the standard penetration resist- confined to driven piles, whereas the Bustamante
ance in the vicinity of the pile tip. These correla- & Gianeselli approach is more general, simpler to
tions indicate that bored or cast-in-place piles apply, and probably more conservative.
develop a significantly smaller end-bearing resist- Baguelin et al. (1986) and Bustamante et al.
ance than do driven piles. (1987) have related f, to the pressuremeter limit
Bustamante & Gianeselli (1982) suggested cor- pressure p, by way of a factor k, which depends
relations between fb and the average cone pen- on pile and soil types. For non-displacement
etration resistance value near the pile tip. The piles, k, is 1.2 for clays and silts and 1.1 for sands,
correlation factor to the latter value is between whereas for driven piles, k, is 1.8 for clays and
0.3 and 0.55, depending on soil and pile type. silts, and ranges between 3.2 and 4.2 for sands.
Soil Young’s modulus E, However, for many clays, the difference between
Ideally, for piles in clay, a distinction should be the drained and undrained modulus values is not
made between the undrained Young’s modulus, great and the approximate nature of most correl-
used for calculations of immediate or undrained ations makes such a distinction impractical. It is
settlement, and the drained Young’s modulus, therefore suggested that the correlations present-
used for calculations of total settlement of a pile. ed in this Paper should be considered to apply to
Table 9. End bearing capacity of pile tip,f,, determination from laboratory data
100
“E
2 80I-
E6c
5
l-
g 4()-
%
6
2()- Llndramed Shear Strength cU lkN/m2)
Cl Bustamante b, Gianeselli 11982) Fig. 28. Correlations for soil modulus for piles in clay
(after Callanan & Kulhawy, 1985)
tests, and differences between the clay types. Cal- Fig. 29. Comparison between correlations for soil
lanan & Kulhawy (1985) find that values of EJc, modulus driven piles in sand
Table 10. Correlations between end bearing resistancef, and SPT value;fb = KN MN/m’
Table 11 shows some suggested correlations tions with pressuremeter data have been made
between E, and cone penetration resistance q,, with respect to the axial load transfer curves for a
and, as with most of the other correlations, the pile. A summary of a number of these correla-
range is large. Two correlations for initial tangent tions, for both pile shaft and pile tip responses, is
modulus E,, are shown, both being derived from given by Frank (1985).
dynamic triaxial tests but believed to be relevant
for piles.
Correlations between E, and pressuremeter Poisson’s ratio v,
data have not been extensively developed, Poisson’s ratio of the soil is a necessary input
although Frank (1985) suggests that the initial parameter into analyses that involve elastic con-
tangent modulus can be taken as the initial tinuum theory, but its effect is generally quite
tangent modulus of the expansion curve from the minor when the solutions are expressed in terms
self-boring pressuremeter. Most of the correla- of Young’s modulus of the soil. For saturated
Table 11. Correlations between soil Young’s modulus E, and CPT value-driven piles
E c-1
t = 53q”-=
E E,, and q, in MN/m* Imai & Tonouchi
Dynamic modulus value (1982)
clays under undrained conditions, v, can be taken predictions of settlement are often achieved with
as 0.5. For clays under drained conditions, v, gen- elastic or elastic-plastic theory (Fig. 11).
erally lies within the range 0.35 &- 0.05, whereas
for silica sands, v, is usually within the range
0.3 + 0.1. Lower values, within the range DESIGN CHARTS FOR PILES AND PILE
0.15 &- 0.1, are applicable for many marine cal- GROUPS
careous sediments. Dimensionless category 2 solutions for pile set-
tlement, such as those presented by Butterfield &
Banerjee (1971), Randolph & Wroth (1978),
Hyperbolic curvejtting constant R, Banerjee (1978), Poulos & Davis (1980) and
If a hyperbolic interface model is used, R, Butterfield & Douglas (1981) are useful for pre-
defines the degree of non-linearity and can range liminary design purposes and can often form the
between 0 (an elastic-perfectly plastic response) basis for final design calculations. In applying
and 1.0 (an asymptotic hyperbolic response in such solutions, it is necessary for the user to
which the limiting pile-soil stress is never determine the most appropriate simplified soil
reached). Limited experience suggests that differ- profile and the values of Young’s modulus for the
ent values of R, should be used for shaft and pile soil in this profile. These determinations may be
tip elements. For the shaft, there is a relatively difficult in the early stages of a project when little
small amount of non-linearity, and values of R, in quantitative data is available, and it is therefore
the range CO.5 may be appropriate. In contrast, of value to develop design charts based on hypo-
the pile tip response is often highly non-linear, thetical but realistic soil data. A limited series of
and it is suggested that a value of R, of about 0.9 such charts is presented in this section for driven
may give a reasonable fit with observed behav- and bored piles in both silica sand and clay pro-
iour. files. The geometrical and soil parameters are
In much of the published theoretical work to defined in Fig. 30. The assumed parameters for
date, elastic-plastic response has been assumed the sand and clay profiles are summarized in
(i.e. R, = 0 for both shaft and tip). As pointed out Table 12 and are based on the Author’s judge-
by Poulos & Davis (1980) and Frank (1985), dis- ment and the correlations presented in the pre-
placements at load levels approaching failure are ceding section.
therefore often seriously underestimated. Fortu- For all the charts presented, Young’s modulus
nately however, at normal working loads, the of the pile, E, , has been taken as 30 000 MN/m’.
effects of non-linearity are not great and adequate This value is representative of concrete piles and
I
2
16-
0
n
*
0
Published
Soft clay
Medium ;Lay
Stiff clay
Unspecified
Load Tests
or uncertam
Soft Clay
14 - \
- 12-
Esm,‘cd b-/ e
3
?
IO-
I-~ 12
Medwm
i
= 10
4
8
5
.
0 o b
1 1 1 I
0 5 10 15 20 25 30
Pile Length (ml
I I I I I 1
1 5 10 15 20 25 30
Fig. 32. Design chart for settlement of driven piles in
Pile Length (ml
sand (diameter = @5 f O-1 m)
Fig. 34. Design chart for settlement of bored piles in
are generally consistent with the theoretical sand (diameter 4: 0.5 f O-l m)
values, although in some cases, they are less than
the computed value for the stiffest or densest soil
condition. Thus, it may be expected that the
design charts will occasionally give a conservative
estimate of settlement.
12
Medwm Clay
\
e 10
-? 8
Settlement of pile groups
wl Following the approach adopted for single
piles, charts have been developed to enable a
6
rapid estimate to be made of group settlements
for both driven and bored piles in various soil
L types. The most convenient form of presentation
is in terms of the group settlement ratio R, and,
2 more specifically, the exponent w of the settle-
ment ratio in equation (11). An elastic boundary
I I ,. , I I element analysis, by means of the computer
0 5 10 15 20 25 program DEFPIG, has been used to obtain the sol-
Pde Length Iml
utions, with allowance being made for the higher
Fig. 33. Design chart for settlement of bored piles in stiffness of the soil between the piles than at the
clay (diameter = @6 f @l m) pile-soil interface (Poulos, 1988a).
wlfy’ywIl’-l
0-
0 10 20 30 “0 10 20 30 0 10 20 30 0 10 20 30
Pile Length LmI Pile Length lm) Pile Length Im) Pile Length (ml
ta) Soft Clay (b) Stiff Clay la) Medium Clay tb) Stiff Clay
0 10 20 30 0 10 20 30
Pile Length (m) Pile Length (m) Pile Length (ml Pile Length (m]
icj Loose Sand Id1 Very Dense Sand tc) Medium Sand Id) Very Dense Sand
Fig. 35. Settlement ratio expownt o, driven piles, s/d = 3: (a) soft Fig. 36. Settlement ratio exponent o, bored piles, s/d = 3: (a) medium
clay; (b) stiff clay; (c) loose sand; (d) very dease sand clay; (b) stiff clay; (c) medium sand; (d) very dense sand
PILE BEHAVIOUR-THEORY AND APPLICATION 399
The design charts are shown in Figs 35 and 36 required of the settlement of the group under a
as plots of the settlement ratio exponent w total load of 3.2 MN, which is equivalent to a
against pile length for four soil types and for safety factor of about 2.5 on the ultimate capac-
driven and bored friction piles of lengths between ity.
10 and 30 m. The centre-to-centre spacing From Fig. 31, for 20 m piles in soft clay, the
between the piles is assumed to be 3 diameters. value of S/P is about 14.0 mm/MN. Thus, at the
The average value of w for square groups of average working load of 3.218 = 0.4 MN, the set-
between 4 and 25 piles, and for diameters of 0.3 tlement of a single pile would be 14.0 x 0.4 = 5.6
m and 0.6 m, is plotted, together with the range mm. From Fig. 35, the average settlement ratio
of theoretical values of UI for each pile length. For exponent w for 20 m long driven piles in soft clay
piles in clay, w is almost independent of pile is about 0.45. Thus, for an 8-pile group, R, =
length, whereas for the piles in silica sand, w 8°.45 = 2.55. Therefore, the group settlement is
decreases as the pile length increases. The range estimated to be 2.55 x 5.6 = 14.3 mm.
of values of o is narrow for shorter pile lengths,
but increases as the pile length increases. Over the
entire range of pile and soil types, w varies Ultimate load capacity of single piles
between 0.41 and 0.51 for 10 m long piles, and Although this section focuses on pile settle-
between 0.22 and 0.47 for 30 m long piles. There ment, it is relevant to remark that design charts
is therefore a clear implication that the use of may also be developed for pile load capacity.
these design charts is likely to be less accurate for Such charts are shown for driven piles in Figs
long piles than for short piles. 37-39 and are classified as category 1 charts
To illustrate the use of the design charts for because they are based on the empirical values of
both single piles and pile groups, consider the shaft and end bearing resistance shown in Table
case of a group of eight driven piles (in a 4 x 2 12. Fig. 37 plots the ultimate shaft resistance
configuration) each 20 m long and 0.6 m dia., in a against pile length for four pile diameters and
deep layer of soft clay. The centre-to-centre three classifications of clay condition, soft, firm
spacing between the piles is 1.8 m. An estimate is and hard. Figs 38 and 39 plot the pile tip cap-
z 3 6
E
L ld=0.3ml
z 2
%
Hard
%
=I 1 Medium
QJ
f
.E Soft
s
Y#zI 0
0 10 20 30
Pile Length lml Pile Length (m1
z
5
x 6 pxq
C
Y Hard
Et 4
U
Medium
CE
m
si 2
al Soft
%
.E 0U
5 0 10 20 30 0 10 20 30
Pile Length (ml Pile Length (m)
Fig. 37. Design charts for ultimate shaft capacity of driven piles in
clay:(a)d=~3m;(b)d=O-6m;(c)d=@9m;(d)d=1~2m
I I
1 Driven Pile 1
0
0 0.5 1.0 1.5
Base Diameter (ml Pile Tip Diameter (m)
Fig. 38. Design chart for ultimate tip capacity of driven Fig. 39. Design chart for ultimate tip capacity of driven
piles in clay piles in sand
acity, as a function of pile tip diameter, for tips Case study I-sensitivity of single pile performance
bearing in either clay or sand. Charts such as calculations
these are perhaps over-simple and must be used The case analysed is a pile test described by
with due caution. Nevertheless, they can give a Gurtowski & Wu (1984). The test pile was located
useful preliminary appreciation of the pile length at the site of the West Seattle Freeway in the
and diameter requirements to develop a specified USA. For the site under consideration (site A) the
ultimate load capacity. geotechnical profile is shown in Fig. 40 together
with SPT data, which was the only quantitative
geotechnical data available. The water table was
located about 3 m below the surface.
CASE STUDIES
The test pile was a O-61 m wide octagonal pre-
In predicting the behaviour of pile foundations,
stressed concrete hollow pile with a plug at the
the geotechnical engineer is faced with a number
tip, and was driven to a depth of about 30 m.
of decisions, including
(a) the method of analysis, and the associated soil Blows per 300mm
model, to be used Soil Log
0 40 80 120
(b) the way in which the soil profile can be sim- Fill
plified and idealized for the analysis
(c) the geotechnical parameters to be used. Loose Sand
Table 13. Analysis conditions and parameters for performance sensitivity evalu-
ation
A ‘standard’ set of conditions and parameters initial tangent modulus. The pile head settlement
was chosen, and the effects of deviations from and pile load at mid-depth are relatively insensi-
these standards were examined. The standard tive to the soil model, whereas the tip settlement
values chosen, and the deviations considered, are is influenced more. The two models involving the
summarized in Table 13. The soil parameters are use of the initial tangent modulus give smaller tip
derived from the correlations presented in Tables movements than those using the secant modulus;
I, 10 and 12, and Fig. 29. all, however, predict a tip settlement which is
Figure 41 compares the measured load- much smaller than the measured value.
settlement behaviour and load distributions with The effect of the soil profile idealization is
those predicted from the standard analysis. The shown in Fig. 43. The use of a homogeneous
standard analysis predicts the pile head load- profile leads to a significantly smaller head settle-
settlement behaviour quite well, and also gives a ment and mid-depth load than the other three
good prediction of the pile load distribution at profiles. The use of a detailed profile in which E,
three different applied load levels. However, the and f, vary as N varies gives a predicted behav-
head load against tip settlement relationship is iour which is little different from that for the two-
not well predicted, indicating that perhaps too layer profile. The indiscriminate assumption of a
large a stiffness has been assigned to the soil at homogeneous soil profile can lead to inaccuracy
and beneath the pile tip. in the predicted pile behaviour, whereas more
In order to enable ready appreciation of the realistic (but nevertheless simplified) modelling of
effects of deviations from the standard conditions, the soil profile variation gives quite adequate
attention has been concentrated on the pile head results in this case.
settlement, the tip settlement, and the load at Figure 44 shows the effect of the soil modulus
mid-depth of the pile, for an applied load of 1.8 correlation used for the analysis. The larger the
MN, which would be a normal working load for soil modulus, the smaller are the settlements and
this pile. the smaller is the load in the pile at mid-depth.
The influence of the soil model is shown in Fig. The tip settlement is particularly affected by E,.
42. In some cases, E, is taken as the secant However, it is interesting to note that an increase
modulus, whereas in others, it is taken as the in E, from 2.5 N to 7 N MN/m’ reduces the head
L n
Observed:
a
l
Tell tale rods
Strain gauges
settlement by less than 30%. This occurs because desirable to attempt to determine the pile
the pile is relatively compressible. In this case, modulus as accurately as possible.
because the piles are concrete, some uncertainty
will be present in the choice of the pile modulus
E, The importance of E, is demonstrated in Fig. Case study 2-sensitivity of pile group performance
45. Of all the factors considered, it has the great- calculations
est influence on the pile head settlement, although The sensitivity of theoretical predictions of pile
it has little influence on the tip settlement. The group settlement and load distribution will be
load distribution is also affected by the pile discussed with respect to the well-documented
modulus, especially in the lower stiff soil layer. case study described by O’Neill et al. (1981;
The foregoing study therefore reveals that the 1982).
head settlement of a relatively long compressible Figure 46 summarizes the geotechnical data at
pile may be influenced as much by the pile the test site which was located at the University
modulus as by the soil model, the soil profile of Houston. The site consists of various layers of
idealization, or the soil modulus correlation stiff to very stiff clay, and geotechnical data is
adopted. The soil modulus may have a substan- available from standard penetration tests, cone
tial influence on the pile tip settlement, while the penetration tests, pressuremeter tests, unconsoli-
soil profile idealization can influence both load dated undrained triaxial tests, laboratory consoli-
distribution and pile head settlement significantly. dation tests, and seismic cross-hole tests.
The effect of the soil model is not great, except for Vertical load tests were performed on full-scale
the tip settlement. Thus, in predicting the settle- pile groups and single piles, with measurements
ment of long compressible piles, it would seem being made of settlement and load distribution
1,,‘.‘,,,.+ mo,,u,us) l I
nsfer I I
for groups of nine, five and four piles, as well as assumed to be the same as that adjacent to the
two single piles. The piles were about 13 m long, piles, and the second a modified analysis in which
0.273 m dia. steel tubes with a 9.3 mm wall thick- the soil between the piles was assumed to be
ness. stiffer than near the piles. The seismic cross-hole
In examining the influence of the method of data were used to estimate the small-strain
analysis, only a limited number of methods were Young’s modulus of the soil between the piles.
considered. The programs DEFPIG, PIGLET and For all analyses, the soil Young’s modulus E,
GAP~X were employed, with two DEFPIG analyses (near the piles) was assumed to vary linearly with
being carried out, the first a conventional analysis depth, according to the relationship E, = 40
in which the soil modulus between the piles was + 5.382 MN/m*, this relationship being based on
Applied Load:l.BMN
+I
Elastic-Plastic Continuum
Analysis
ES&N MN/m2
Fig. 43. Influence of soil profile idealization on predicted behaviour of single pile
Pile A
Gurtowski and Wu 1198Ll
Applied Load=l.tlMN
E,=7.5N-94.5 MN/m’ l
I
Two-Layer Soil Profile I
Fig. 44. Influence of soil modulus correlation on predicted behaviour of single pile
I 3
Pile Modulus E, (GN/m*I Pile Modulus E, (GN/m’l
(al Head Settlement (b) Tip Settlement
Load (MN)
Pile A
Gurtowski and Wu (198L)
Applied Load=l.BMN
Elastic-Plastic Continuum
E,=4N MN/m2
Settlement (mm/MN1
Program
=;-I
I PIGLET II
Measured __( .
I
I GAPFIX
Fig. 47. Effect of analysis method on single pile and group settlements
the correlation E, = 750 c, . The values of c, were appears to be reasonably well predicted by the
obtained from the pressuremeter test data, the analysis.
profile being simplified to vary linearly with The foregoing comparisons suggest that, given
depth from 53 kN/m’ at the surface to 147 a set of common parameters, most of the methods
kN/m’ at a depth of 13 m. considered give a similar prediction of group
The calculated settlements of both the single behaviour. However, for group settlement, better
pile and the nine-pile group are shown in Fig. 47. agreement is obtained with the modified DEFPIG
Also shown are the estimates of settlement analysis in which the greater stiffness of the soil
derived from the category 2 design charts present- between the piles is allowed for.
ed in this Paper for driven piles in stiff clay. There To examine the influence of the idealization of
are some differences between the settlements the soil profile on the group settlement, four dif-
derived from the different analyses, but these are ferent distributions of soil modulus with depth
not great. The single pile settlements given by the have been used, each being based on the
DEFPIG and GAPFIX analyses are close to the mea- undrained shear strength distribution obtained
sured value, but the group settlements are over- from the pressuremeter tests
estimated by all approaches. The modified DEFPIG
(a) the linearly varying distribution of E, con-
analysis gives the most satisfactory prediction of
sidered above
group settlement. The category 2 design charts
(b) a homogeneous profile in which E, = 75
overestimate both the single pile and group settle-
MN/m2
ments, although this is to be expected as these
charts are meant for design rather than prediction
and therefore tend to be conservative. ---- Measured
For all the group configurations tested by Theoretical, using conventional
x
O’Neill et al., Fig. 48 compares the measured set- interactjon factors (p=ll
tlement ratios with those determined from the * Theoretical, using modified
conventional and the modified DEFPIG analyses interaction factors lp-4.8)
(Poulos, 1988a). The conventional analysis con- 3
siderably overestimates R,,whereas the modified ar” x
analysis gives significantly closer agreement with 2
_G’[~“~~
(c) a two layer profile in which the first layer of the soil Young’s modulus has been investigated
extends to a depth of 8 m and has E, = 55 by using six different procedures and employing
MN/m’, and the second extends to great the conventional DEFPIG analysis. Fig. 51 shows
depth and has E, = 100 MN/m’ the predicted and measured settlements of the
(d) a profile which follows the detailed distribu- single pile and the nine-pile group. It is imme-
tion of c, with depth, and in which E, = 750 diately apparent that the predicted settlements
c, are more sensitive to E, than to either the method
of analysis or the soil profile idealization. The
The program DEFPIG was used for all calculations;
closest predictions are given by the correlations
no account was taken of increased soil stiffness
E, = 750 c, and E, = 4N MN/m*, whereas E, =
between the piles.
200 c, and E, = 20 q, lead to gross overestimates
Figure 50 shows the predicted settlements of
of both single pile and group settlements. The
the single pile and the nine-pile group. All four
correlations leading to large settlements also lead
profiles give similar settlements, which agree well
to more non-uniform distributions of pile head
with the measured settlement for the single pile,
load.
but are about 50% too large for the group. There
For this case study, therefore, in which the piles
is also little variation among the solutions for the
were significantly shorter and less compressible
pile head load distribution, but in all cases the
than in the preceding case, the two factors which
predicted load distribution is more non-uniform
appear to be crucial in accurately predicting the
than that measured. It therefore appears that, in
group behaviour are
this particular case, the idealization of the soil
profile is not a crucial factor in the prediction of (a) the magnitude of the soil Young’s modulus E,
pile group behaviour. (b) the greater stiffness of the soil between the
The influence of the method of determination piles than locally near the piles.
Fig. 50. Effect of soil profile idealization on single pile and group settlements
0 5 10 15 0 5 10
I, 1 I
ES= 200~” 0 I
l
I
I
I.
ES= 750~” 1
E,= 20% 1 a I
0
I
E,: 4N MN/m* 1. M easured l Measured
L---
I
Poulos (1972) a
I l
I
E, from PMT l
I
_I l
Single Pile 9-Pile Group
Tests of O’Neill et al (1982)
Calculated from conventional
DEFPIG Program
Fig. 51. Effect of modulus correlation on single pile and group settlements
The general tendency with all conventional shows the soil profile and summarizes the avail-
methods of group analysis is to overestimate the able geotechnical data at the Hemmant site. The
interaction and to predict a settlement which is upper 26 m was essentially clay, underlain by
too large and a load distribution which is too very dense sand. Data were obtained from in situ
non-uniform. Such tendencies may not always be static cone and dilatometer tests, and from labor-
as significant as they appear to be for the case atory triaxial tests. Values of undrained shear
considered here. strength in the clays, derived from these three
sources, agreed well.
In making his predictions, the Author derived
the required parameters as follows.
Case study 3-a class A prediction of pile
behaviour (a) The shaft resistance f, in the clay soils was
In conjunction with the Fifth Australia-New correlated with the undrained shear strength
Zealand Conference on Geomechanics, held in c, , by way of the adhesion factor a which was
Sydney in 1988, geotechnical engineers were determined from the suggestions of Semple &
invited to predict the load-settlement per- Rigden (1984). In the sandy soils,f, was taken
formance of two driven precast concrete piles at a to be 0.25 times the vertical effective over-
site in Hemmant, Queensland, in Australia. After burden pressure. The values off, thus com-
driving, each pile was left for several weeks and puted were generally about 20% larger than
then tested to failure by static load testing pro- those determined from the Schmertmann-
cedures in accordance with the Australian Stan- Nottingham correlations with cone sleeve
dard Piling Code. Sixteen engineers, including the resistance.
Author, submitted ‘class A’ predictions based on (4 The end bearing resistance fb was determined
static analyses, while four engineers undertook from the average cone penetration data within
dynamic analyses using data from restriking tests the vicinity of the pile tip. Values of 14
performed four weeks after the static failure. A MN/m2 and 2.45 MN/m’ were determined
detailed description of the prediction exercise is for piles 1 and 2 respectively.
given by Douglas (1989). (4 The initial tangent value of Young’s modulus
The piles were Balken piles, of high strength E, of the clays was estimated to be 25 times
precast concrete, 275 mm square, cast in lengths the cone resistance. In the underlying dense
of up to 12.2 m, with mechanical jointing during sands, a value of E, of 350 MN/m2 was
driving. The piles were driven using a Banut 600 adopted. Poisson’s ratio was taken as 0.5. The
piling rig with a 5 t hydraulic hammer. Each pile hyperbolic curve fitting factor R, was taken as
was instrumented so that dynamic measurements 0.9 for both shaft and base elements.
could be taken during driving. Pile 1 was driven (d) The pile modulus E, was taken as 35000
to a depth of 30.5 m and pile 2 to 25 m. Fig. 52 MN/m’.
b
h
Clay - high plasticity
Pile 2
- 25
0 Dilatometer
0 UU biaxial
Pile 1 30
A category 3B boundary element analysis was underestimated the capacity, presumably because
performed to predict the load-settlement behav- of underestimation of the shaft resistance f,.
iour, using the computer program PIES (Table 3). Table 14 also summarizes the predicted and mea-
Table 14 summarizes the Author’s predictions sured pile head stiffness values for pile 1. It is
of ultimate load, the measured values, and also noticeable that the range of predictions is sub-
the range of predictions made by the other par- stantially greater than for the ultimate load pre-
ticipants. For both piles, the Author’s prediction dictions. Twelve of the 16 predictors, including
lay between the extremes, and was within f20% the Author, underpredicted the pile head stiffness.
of the measured load. For pile 1, all 16 predictors For pile 2, the measured stiffness was similar to
CONCLUSIONS
Looor----l
3000 - Maximum
This Paper has attempted to review the
analysis of piles and pile groups under axial
loading and to classify and place in perspective a
number of existing methods of analysis. Several of
these can be placed within the framework of the
boundary element method, and can be used, at
least in principle, to model many of the practical
features of real piling problems.
A detailed examination has been made of some
of the more significant aspects of the behaviour of
single piles and pile groups, as revealed by the
Prediction
theoretical solutions. The problems considered
include piles subjected to conventional static
0 10 20 30 40 50 loading, to cyclic loading, and to loadings which
Settlement Imm)
arise from externally-imposed soil movements.
Fig. 53. Predicted and measured load-settlement hehav- Despite the simplifications inherent in the theory,
iour pile 1 there is compelling evidence to demonstrate that
the behavioural characteristics revealed by the
theory are consistent with observations made
that of pile 1, and the Author and most of the from field and laboratory tests on piles.
predictors correctly predicted this characteristic. The importance of selecting appropriate geo-
Figure 53 shows the measured load-settlement technical parameters has been emphasized, espe-
curve for pile 1, the extreme predictions, and the cially for problems that involve conventional
Author’s prediction. The most notable features of loading. For practical application of most of the
this figure are theories, it is generally necessary to make use of
empirical correlations in order to obtain values of
(a) the wide spread of predicted performance
shaft resistance, end-bearing resistance and soil
(b) the considerable difference between the
Young’s modulus. A number of such correlations
Author’s prediction and the measurements,
are summarized in the Paper.
despite the use of a site-specific category 3B
By means of two case studies, an examination
analysis.
has been made of the sensitivity of pile per-
Following the revelation of the field measure- formance predictions to factors over which the
ments, the Author re-computed the performance geotechnical analyst has some control, in particu-
of pile 1 using simple c tegory 2 methods of cal- lar, the method of analysis, the idealization of the
culation based on the parameters shown in Table soil profile and the values of the soil parameters
12. For calculation of ultimate load, the soil used in the analyses. It is concluded that, while all
profile along the shaft was idealized as a 1 m zone these factors may have an influence, the latter two
of zero resistance, 3 m of soft clay, 4 m of loose are generally of greater importance than the
sand, 13 m of soft clay, 5 m of stiff clay and 5.5 m method of analysis.
of dense sand. The tip was assumed to also be in A further case study has been described to
dense sand. The computed capacity in this case demonstrate the difficulty of accurate prior pre-
was 2420 kN, close to the initial predicted value. diction of pile performance, even in relatively
For settlement, the pile head stiffness was com- simple geotechnical conditions in which a con-
puted to be between 70 and 180 kN/mm, depend- siderable amount of geotechnical data is avail-
ing on whether the soil profile was taken broadly able. In this case, much depends on the
as a soft clay or a stiff clay. The average of these experience of the individual making the predic-
two values is 125 kN/mm, which corresponds to tion, and the way in which he or she interprets
the Author’s initial predicted value. Thus, for this the available data to obtain the required geotech-
case, rapid category 2 calculations, performed nical parameters for the analysis employed.
within about 10 min, give predictions which, The Author firmly believes that theoretical
although not in close agreement with the mea- analyses have led to a significantly increased
surements, were no more inaccurate than the understanding of the mechanics of pile-soil inter-
original category 3B predictions. This case study action and an improved appreciation of the
emphasizes once more that the method of calcu- factors which influence pile behaviour. A most
lation may play a far less significant role in pile important feature of a soundly-based theoretical
performance prediction than does the idealization analysis is its ability to answer ‘what-if questions,
of the soil profile and the selection of the geotech- for example, what if the pile length is increased,
nical parameters. what if the number of piles is reduced, what if the
Berezantzev, V. G., Khristoforov, V. & Golubkov, V. Works Dept. of Greece 3, July-Sept, 221-226.
(1961). Load bearing capacity and deformation of Coyle, H. M. & Reese, L. C. (1966). Load transfer for
piled foundations. Proc. 5th Int. Conf Soil Mech. axially loaded piles in clay. J. Soil Mechs Fdn
Fdn Engng, Paris, 2, 11-15. Engng, Am. Sot. Cio. Engrs 92, No. SM2, l-26.
Blight, G. E. (1965). The time-rate of heave of structures D’Appolonia, D. J., D’Appolonia, E. & Brissette, R. F.
on expansive clays. In Moisture equilibria and mois- (1970). Closure to settlement of spread footings on
ture changes in soils beneath covered areas. u. 78, sand. J. Soil Mechs Fdn Enana, Am. Sot. Civ. Enars
Sydney: Butterworths. %, No. SM2,754761. _ _
Blieht.
_ G. E. (1984). Power station foundation in deen
,
Datta, M., Gulhati, S. K. & Rao, G. V. (1980). An
expansive soils. Proc. 5th Int. Conf Expansive Soil;, appraisal of the existing practise of determining the
Adelaide, 1, 22. axial load capacity of deep penetration piles in cal-
Booker, J. R. & Poulos, H. G. (1976). Analysis of creep careous sands. Proc. I2th Annual OTC, Houston
settlement of pile foundations. J. Geotech. Engng, Paper OTC 3867,119-130.
Am. Sot. Cio. Engrs 102, No. GTl, 1-14. Decourt, L. (1982). Prediction of the bearing capacity of
Briaud, J.-L. & Tucker, L. M. (1988). Measured and piles based exclusively on N values of the SPT. Proc.
predicted axial response of 98 piles. J. Geotech. ESOPTZ, Amsterdam 1, 29-34.
Engng, Am. Sot. Ciu. Engrs 114, No. 9,9841001. Denver, H. (1982). Modulus of elasticity determined by
Burland. J. B. (1973). Shaft friction of piles in clav-a SPT and CPT. Proc. ESOPT2, Amsterdam 1,35-40.
simple fundamental approach. Ground Engng 6;No. De Ruiter, J. & Beringen, F. L. (1979). Pile foundations
3, 3&42. for large North Sea structures’. Marine Geotechnol.,
Bustamante, M., Frank, R. & Gianeselli, L. (1987). Le 3, No. 3, 267-314.
dimensionnement des fondations profondes. Bull. Desai, C. S. (1974). Numerical design-analysis for piles
Liaison Labs P. et Ch 149, May-June, 13-22. in sands. J. Geotech. Engng, Am. Sot. Cio. Engrs 100,
Bustamante, M. & Gianeselli, L. (1982). Pile bearing No. GT6,613635.
capacity prediction by means of static penetrometer Donaldson, G. W. (1967). The use of small-diameter
CPT’. Proc. ESOPT II, Amsterdam 2,49>500. piles in expansive soil. Proc. 4th Regional Co& for
Butterfield, R. SC Abdrabbo, F. M. (1983). Static Africa on Soil Mechs Fdn Enana 1.253.
working load tests on small scale pile groups in clay. Douglas, D. J. (1989). Prediction &thy for driven piles.
In Developments in soil mechanics and foundation 5th Australia-New Zealand Geomechs Conf, Aust.
engineering I, model studies, pp. 111-168. London: Geomechs, special issue.
Applied Science Publishers. __ Dutt, R. N. & Ingram, W. B. (1984). Jackup rig siting in
Butterfield. R. & Baneriee. _ P. K. (1971). , I
The elastic calcareous soils. Proc. 16th Annual OTC, Houston
analysis of compressible piles and pile groups. G&o- Paper OTC 4840,541-548.
technique 21, No. 1,4%60. Edil, T. B. & Mochtar, I. B. (1988). Creep response of
Butterfield, R. & Douglas, R. A. (1981). Flexibility coef model pile in clay. J. Geotech. Engng, Am. Sot. Ciu.
Jicients for the design of piles and pile groups. Tech. Engrs 114, No. 11,1245-1260.
Note 108. London: CIRIA. El Shamouby, B. & Novak, M. (1985). Static and low-
Butterfield, R. & Ghosh, N. (1977). The response of frequency response of pile groups. Can. Geotech. J.
single piles in clay to axial load. Proc. 9th Int. Conf 22, No. 1,79-94.
Soil Mechs Fdn Engng, Tokyo 1,451457. Findlay, J. D. (1984). Discussion, Piling and ground
Callanan, J. F. & Kulhawy, F. H. (1985). Evaluation of treatment, pp. 189-190, London: Instn. Civ. Engrs.
procedures for predicting foundation uplift movements. Thomas Telford.
Report to Electric Power Research Inst., No. EPRI Fleming, W. G. K. & Thorburn, S. (1984). Recent piling
EL-4107, Cornell University. advances. Piling and Ground Treatment, pp. t-16.
Cameron, D. A. & Walsh, P. F. (1984). The prediction London : Instn of Civil Engrs. Thomas Telford.
of moisture induced foundation movements using Fleming, W. G. K., Weltman, A. J., Randolph, M. F. &
the instability index. Aust. Geomechs News, No. 8, Elson, W. K. (1985). Pilina enaineerina. New York:
511. Surrey University Press, Halsted Press:
Chan, K. S., Karashudi, P. & Lee, S. L. (1974). Force at Fletcher. M. S. & Mizon. D. H. (1984). Piles in chalk for
a point in the interior of a layered elastic half-space. Orwell bridge. Piling and Ground Treatment, pp.
Ini. J. Solids Struct. 10, 11761199. 203-209, London: Instn. Civil Engrs. Thomas
Cheuna. Y. K.. Tham. L. G. & Guo. D. J. (1988). Telford.
Analysis of pile group by infinite layer method. Ged- Focht, J. A. (1967). Discussion to paper by Coyle and
technique 38, No. 3,415431. Reese, J. Soil Me&. Fdn. Engng, Am. Sot. Civ.
Chin, J. T. (1988). Analysis of piles and pile groups Engrs 93, No. SMl, 133-138.
embedded in a layered half-space. ME thesis, Frank, R. (1985). Recent developments in the prediction
National University of Singapore. of pile behaviour from pressuremeter results. Proc.
Chow, Y. K. (1986). Analysis of vertically loaded pile Symp.from Theory to Practice on Deep Foundations,
groups. Int. J. Numer. Anal. Meth. Geomechs 10, Porto Alegre, Brazil 1, 69-99.
59-72. Giroud, J. P., Tran-Vo-Nhiem & Obin, J. P. (1973).
Christoulas, S. G. (1988). Dimensionnement de pieux. Tables pour le calcul des fondations 3, Paris: Dunod.
Quelques experiences et recherches en G&e. Bull. Gurtowski, T. M. & Wu, M.-J. (1984). Compression
Liaison Labs. P. et. Ch. 154, I-10. load tests on concrete piles in alluvium. Anal. and
Christoulas, S. & Pachakis, M. (1987). Pile settlement Design of Pile Foundations, Am. Sot. Ciu. Engrs, 138-
based on SPT results. Bull. Res. Centre of Public 153.
Ha, H. B. & O’Neill, M. W. (1983). Discussion on Theo- in a group of piles, to be published in J. Geotech.
retical t-z curves. J. Geotech. Engng, Am. Sot. Ciu. Engng, A&. Sot. Cio. Engrs.
Engrs 109, No. GTlO, 1353-1355. Lambe. T. W. (1973). Predictions in soil engineering.
Hewitt, C. M. (1988). Cyclic response of offshore pile Gtdtechniquh 23, No. 2, 149-202.
groups, PhD thesis, University of Sydney, Australia. Lee, C. Y. (1988). The behaviour of grouted piles in ON-
Hobbs, N. B. (1977). Behaviour and design of piles in shore foundations. PhD thesis, University of Sydney,
chalk-an introduction to the discussion of the Australia.
papers on chalk. Piles in Weak Rock, pp. 149-175, Leonards, G. A. & Darrag, A. A. (1989). Discussion to
London: Instn. Civil Engrs. Analysis of residual stress effects in piles, H. G.
Holeyman, A. (1985). Dynamic non-linear skin friction Poulos. J. Geotech. Engng, Am. Sot. Civ. Engrs 115,
of piles. Proc. Symp. on Penetrability, Driuability of No. 4, 589-596.
Piles, 11th Int. Con& Soil Mechs Fdn Engng, San Martin, R. E., Seli, J. J., Powell, G. W. & Bertoulin, M.
Francisco. (1987). Concrete pile design in Tidewater, Virginia.
Holloway, D. M., Clough, G. W. & Vesic, A. S. (1978). J. Geotech. Engng Am. Sot. Civ. Engrs 113, No. 6,
The effects of residual driving stresses on pile per- 568-585.
formance under axial loads. Proc. Z&h OTC, Matlock, H. & Foo, S. C. (1980). Axial analysis of piles
Houston, 2225-2236. using a hysteretic and degrading soil model. Proc.
Hull. T. S. (1987). The static behaviour of laterallv Co& Numer. Methods in Offshore Piling _ pp.
__ 16s
loaded piies, PhD thesis, University of Sydney, Au;- 185; London, Inst. Civil Eng;;.
tralia. Mattes. N. S. & Poulos, H. G. (1971). Model tests on
Hyden, A. M., Hulett, J. M., Murff, J. D. & Abbs, A. F. piles in clay. Proc. ‘1st Aus;.-New Zealand Conf:
(1988). Design practice for grouted piles in Bass Geomechs, Melbourne, 254-259.
Strait calcareous soils. In Engineering for calcareous McClelland. B. (1974). Design of deco nenetration niles
sediments 1, pp. 297-304. Rotterdam: A. A. for ocean &uct&es. J.-Geotech: Engng, Am. ‘Sec.
Balkema. Civ. Engrs 100,No. GT7,705-747.
Imai, T. & Tonouchi, K. (1982). Correlation of N value Meyerhof, G. G. (1956). Penetration tests and bearing
with S-wave velocity and shear modulus. Proc. capacity of cohesionless soils. J. Soil Mechs. Fdn
ESOPT2, Amsterdam 1, 67-72. Engng, Am. Sot. Ciu. Engrs 82, No. SMl, 1-19.
Jardine, R. J., Potts, D. M., Fourie, A. B. & Burland, J. Meyerhof, G. G. (1959). Compaction of sands and
B. (1986). Studies of the influence of non-linear bearing capacity of piles. J. Soil Mechs Fdn Engng,
stress-strain characteristics in soil-structure inter- Am. Sic. Cio. Engrs k, No. SM6, l-29.
action’. GCotechnique 36, No. 3,377-396. Meverhof, G. G. (1963). Some recent research on the
Johnston, I. W. & Lam, T. S. K. (1984). Frictional char- -karing capacity of foundations. Can. Geotech. J. 1,
acteristics of planar concrete-rock interfaces under No. 1, 1626.
constant normal stiffness conditions. Proc. 4th Meyerhof, G. G. (1976). Bearing capacity and settlement
Aust.-New Zealand Co& on Geomechs, Perth, 2, of pile foundations’. J. Geotech. Engng, Am. Sot. Ciu.
397401. Engrs 102, No. GT3,19%228.
Johnston, I. W., Carter, J. P., Novello, E. A. & Ooi, L. Milovic, D. & Stevanovic, S. (1982). Some soil param-
H. (1988). Constant normal stiffness direct shear eters determined by cone penetration tests. Proc.
testing of calcarenite. In Engineering for calcareous ESOPT2, Amsterdam, 2,70%714.
sediments 2, pp. 541-553, Rotterdam: A. A. Mindlin, R. D. (1936). Force at a point in the interior of
Balkema. a semi-infinite solid. Physics 7;195-202.
Keg, Y. C., Karunaratne, G. P. & Lee, S. L. (1986). Naurov. J.-F.. Brucv. F.. Le Tirant. P. & Kervadec. J.-P.
Effects of negative friction on piles in layered soils. (19&). Design aid ihstallation’of piles in calcaieous
Geotech. Engnq 17,21 l-234. formations. Proc. 3rd Int. Conf Numer. Methods in
Komomik, A.- (i974). Penetration testing in Israel. Ofihore Piling, Nantes, 461480.
State-of-the-art renort. Proc. ESOPT, Stockholm 1. Nystrom, G. A. (1984). Finite-strain axial analysis of
183-192. - piles in clay. Anal. Design Pile Fdns, Am. Sot. Cio.
Kraft, L. M. & Lyons, C. G. (1974). State-of-the art: Engrs, l-20.
ultimate axial capacity of grouted piles. Proc. 6th Ohya, S., Imai, T. & Matsubara, M. (1982). Relation-
Annual OTC, Houston Paper OTC 2081,487-503. ships between N-value by SPT and LLT measure-
Kraft, L. M., Ray, R. P. & Kagawa, T. (1981). Theoreti- ment results. Proc. ESOPTZ, Amsterdam 1, 125-130.
cal t-z curves. J. Geotech. Engng, Am. Sot. Civ. O’Neill, M. W. (1983). Group action in offshore piles.
Engrs 107,No. GTll, 1543-1561. Proc. Am. Sot. Ciu. Engrs Co& Geotech. Practice in
Kulhawy, F. H., O’Rourke, T. D., Stewart, J. P. 8c Offshore Engng, Austin, 2564.
Beech, J. F. (1982). Transmission line structure foun- O’Neill, M. W., Ghazzaly, 0. I. 8t Ha, H. B. (1977).
dations for uplift/compression loading. Supplemental Analysis of three-dimensional pile groups with non-
Volume: Load Test Summaries, Res. Proj. 1493-1, linear soil response and pile-soil-pile interaction.
final report to EPRI, Cornell University. Proc. 9th Annual OTC, Houston Paper OTC 2838,
Kulhawy, F. H., Trautmann, C. H., Beech, J. F., 245-256.
O’Rourke, T. D., McGuire, W., Wood, W. A. & O’Neill, M. W. & Ha, H. B. (1982). Comparative model-
Capano, C. (1983). Transmission line structure foun- ling of vertical pile groups. Proc. 2nd Znt. Conf:
dations for uplift-compression loading. Report EL- Numer. Methods in Offshore Piling, Austin, 399-418.
2870, Electric Power Research Inst., Palo Alto. O’Neill, M. W., Hawkins, R. A. & Mahar, L. J. (1981).
Kuwabara, F. & Poulos, H. G. (1989). Downdrag forces Field study of pile group action: final report. Report
of design methods for foundations under axial uplift the unusual opportunity, with respect to a distant
and compression loading. Report for EPRI, No. EL- overseas lecturer, of briefing him at first hand
3771, Cornell University. when I attended the Australia-New Zealand
Steinbrenner, W. (1934). Tafeln zur Setzungberechnung. Geomechanics Conference in Sydney in August
Die Strasse 1, 221.
1988.
Stewart, J. F. & Kulhawy, F. H. (1981). Interpretation
of uplift load distribution data. Proc. 10th Int. Conf.
In my invitation to Harry, I said that we
Soil Mechs Fdn Engng, Stockholm 2,277-280. assumed the subject of piling could well feature in
Terzaghi, K. (1939). Soil mechanics-a new chapter in his presentation. Within two months, I had a
engineering science. J. Znstn Ciu. Engrs 12, 106141. letter outlining the lecture much as we heard it
Terzaghi, K. & Peck, R. B. (1967). Soil mechanics in this evening. By the time we had lunch together
engineering practice. 2nd Ed., New York: Wiley. in Sydney eight months later, all I had to do was
Thorbum, S. & MacVicar, S. L. (1971). Pile load tests to make encouraging noises, since it was clear to me
failure in the Clyde alluvium. Behaviour of Piles, that a balanced, well planned and extremely
London: Institution of Civil Engineers. 1-7, 53-54.
appropriate lecture was in prospect. I took the
Tomlinson, M. J. (1957). The adhesion of piles driven
precaution of telling Harry how even greater men
into clay soils. Proc. 4th Znt. Conf Soil Mechs Fdn
Engng, London, 2,66-71. had found the Rankine Lecture occasion a difli-
Tomlinson, M. J. (1977). Pile design and construction cult one, that the audience varied from the least
practice. London: Cement and Concrete Associ- to the greatest in British Geotechnics, and that all
ation. would wish to go away having learned something.
Valliappan, S., Lee, I. K. & Boonlualohr, P. (1974). Set- I mentioned visual aids en passant remembering
tlement of piles in layered soils. Proc. 7th Biennial some past examples of slides that left something
Conf., Aust. Road Res. Board, Adelaide I, Pt 7, 144- to be desired.
153. Towards the end of the Sydney Conference, I
Van Der Merwe, D. H. (1964). The prediction of heave
had the further opportunity of listening to Harry
from the plasticity index and the percentage clay
fraction. Civil Engr in South Africa 6, No. 6, 103. Poulos eloquently deliver the John Jaeger Me-
Van Impe, W. F. (1986). Evaluation of deformation and morial Address on the subject of calcareous soils.
bearing capacity parameters of foundations, from By then no doubts remained.
static CPT-results. 4th Znt. Geotech. Seminar Field Against this background, therefore, and in
Znstrumentn and In-Situ Measurements Nanyang common with many others who had been privi-
Tech. Inst., Singapore, 51-70. leged to hear Harry Poulos speak on other
Verbrugge, J. C. (1982). The evaluation of pile cap load- occasions, I arrived this evening, not merely with
settlement diagram from the CPT results. Proc. the usual anticipation of the big event, but with
ESOPT2, Amsterdam 2,923-926.
the confidence that we were to hear an excellent
Vesic, A. S. (1972). Expansion of cavities in infinite soil
mass. J. Soil Mechs Fdn Engng, Am. Sot. Civ. Engrs
lecture presented with clarity and authority.
98, No. SM3,265-290. Ladies and Gentlemen, we have not been disap-
Williams, A. F. & van der Zwaag, G. L. (1988). Analysis pointed.
and evaluation of grouted section tests, Engng for Harry, you have demonstrated vividly the use
calcareous sediments 2, pp. 493-502, Rotterdam: which can be made of what Terzaghi somewhat
Balkema. scornfully referred to as ‘Science’ and you have
Withiam, J. L. & Kulhawy, F. H. (1979). Analytical put in clear perspective the other less prudent
model for drilled shaft foundations. Proc. 3rd Znt. statement he made with Ralph Peck that theoreti-
Con& Numer. Methods in Geomechs, Aachen 3, 1115-
cal refinements for piles are completely out of
1122.
Wright, S. J. & Reese, L. C. (1979). Design of large dia-
place. Your balanced approach giving a sense of
meter bored piles. Ground Engng Nov., 17-50. perspective on the input parameters for design
Yamashita. K.. Tomono. M. & Kakurai, M. (1987). A through your sensitivity analyses and presenting
method for estimating immediate settlement of piles applications of theory through case studies and a
and pile groups. Soils and Fdns 27, No. 1,61-76. class A prediction have clearly demonstrated how
theory should be used, not as a master but as an
essential tool in geotechnical engineering.
Professor Poulos, the British Geotechnical
VOTE OF THANKS Society is enormously grateful to you for pre-
S. F. BROWN paring such an excellent and stimulating Rankine
Ladies and Gentlemen, Lecture, which was full of useful practical infor-
About 18 months ago, I wrote to Harry Poulos in mation and will, I predict, become a standard ref-
my then capacity as BGS chairman inviting him erence when it appears in Gtotechnique in due
to deliver the 1989 Rankine Lecture. As a conse- course.
quence, I have carried a certain responsibility We thank you and admire you for delivering
over this period, which is now happily discharged. this lecture with such EASE+loquence, aplomb,
My responsibility was enhanced by having had stamina and enthusiasm.