POULOS, H. G., 1968. G&technique, 18, 449-471.

ANALYSIS OF THE SETTLEMENT OF PILE GROUPS

H. G. POULOS, B.E., Ph.D., A.M.I.E.Aust.*

SYNOPSIS
An analysis is made of the settlement interaction On fait d’abord une analyse de l’interaction de
between two identical piles in an elastic mass and the tassement entre deux pieux identiques dans une
masse Blastique et l’accroissement du tassement de
increase in settlement of each pile due to interaction
chaque pieu dii a l’interaction est exprime en fonc-
is expressed in terms of an interaction factor a. tion d’un facteur d’interaction a. On montre en-
It is then shown that, for symmetrical pile groups suite que pour des groupes de pieux symetriques
(those in which the piles settle equally and are also (ceux dans lesquels les pieux se tassent d’une faGon
&gale et supportent aussi des charges Bgales), l’ac-
equally loaded), the increase in settlement due to
croissement de tassement dfi a l’interaction peut
interaction may be obtained by superposition of the &tre obtenu par une superposition des valeurs de a
values of a for the individual piles in the group. pour les pieux individuels du groupe.
On the assumption that superposition also holds En supposant que la superposition vaut aussi pour
n’importe quel groupe general de pieux se tenant
for any general free standing pile group, the be-
librement, le comportement des groupes de pieux
haviour of pile groups is analysed for the case of a est analyse dans le cas d’une longrine rigide (tasse-
rigid pile cap (equal settlement of all piles) and a ment Bgal de tous les pieux) et d’une longrine flexible
flexible pile cap (equal load in all piles). For the (charge Bgale supporte par tous les pieux). Dans le
cas d’une longrine rigide, des valeurs sont obtenues
case of a rigid pile cap, values are obtained for the
pour le rapport du tassement du groupe au tasse-
ratio of the settlement of the group to the settlement ment d’un seul pieu supportant la m&me charge
of a single pile carrying the same total load (the totale (le facteur de reduction du groupe RG), et la
group reduction factor Rc), and the load distribu- repartition de la charge a l’interieur du groupe.
Dans le cas d’une longrine flexible, on donne des
tion within the group. For the case of a flexible
valeurs du tassement maximum et du tassement
pile cap, values of the maximum settlement and differentiel maximum.
maximum differential settlement are given. L’influence de l’espacement des pieux, de la
The influence of pile spacing, pile length, type of longueur des pieux, le type du groupe, la profon-
deur de la couche et le coefficient de Poisson pour
group, depth of layer and Poisson’s ratio of the
la couche sur le comportement de tassement des
layer on the settlement behaviour of pile groups is groupes de pieux est examine. Pour un groupe
examined. For a pile group in an ideal elastic two- de pieux dans un sol ideal elastique a deux stades,
phase soil, it is shown that, as with a single pile, the on montre que, comme dans le cas d’un seul pieu,
major proportion of the total final settlement la plus grande proportion du tassement final total
a lieu generalement sous forme d’un tassement im-
generally occurs as immediate settlement.
mtdiat.
Comparisons are made between reported observa- Des comparaisons sont faites entre les observa-
tions on the behaviour of pile groups from model and tions ayant fait l’objet de rapports sur le comporte-
field tests and the behaviour predicted by the theory. ment de groupes de pieux Q partir d’essais sur
It is found that, as well as predicting the correct maquette et sur les lieux et le comportement prevu
par la theorie. On s’aperqoit que, outre une pre-
trends, the theory gives quantitative values which
vision des tendances exactes, la theorie donne des
are in reasonable agreement with the observed valeurs quantitatives qui s’accordent de faGon
values. raisonnable avec les valeurs observees.

INTRODUCTION
Despite the wide use of pile groups in current foundation practice, only approximate
methods of estimating the settlement of pile groups are at present available. An examination
of such methods by Bjerrum et al. (1957) has indicated that they are far from satisfactory and
may lead to an underestimate of settlement.
The advent of computers has enabled the behaviour of a single axially loaded pile or pier
in an elastic mass to be examined theoretically. Such analyses have been carried out by
several investigators, e.g. Thurman and D’Appolonia (1965), Salas and Belzunce (1965),
D’Appolonia and Romualdi (1963), Poulos and Davis (1968), and in the latter paper solutions

* Lecturer in Civil Engineering, University of Sydney.

449

Downloaded by [ Columbia University] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 450 H. G. POULOS

have been presented for the settlement of an incompressible floating pile or pier in a finite
elastic layer. It has been found that general trends in the behaviour of single piles and piers
indicated by the theoretical analyses are in agreement with observed behaviour and that the
theory is also capable of giving satisfactory quantitative predictions of pile displacements.
The successful use of elastic theory for analysing the behaviour of a single pile therefore
suggests that it may also be used to examine the behaviour of pile groups.
In this Paper the analysis of a single pile is extended to the consideration of the settlement
interaction between two identical floating piles, such a system representing the simplest case
of a pile group. The analysis is then extended to the case of a general floating pile group with
any configuration of piles within the group. For a general pile group two limiting cases are
considered: that of a perfectly rigid pile cap, where all piles settle equally, and that of a per-
fectly flexible pile cap, where all piles carry equal load. It is likely that the former case will
be more appropriate for most practical cases.
It is impossible to present detailed solutions for all cases which might be encountered in
practice because of the large number of variables in this problem. The solutions which are
given in this Paper are therefore generally confined to the case of square groups of piles con-
taining up to twenty-five piles.
In the analyses carried out the piles are assumed to be incompressible, so that the solutions
are relevant to the case of a pile group in a soft clay. No slip is assumed to occur between
the pile and the adjacent soil and thus the solutions presented herein are valid only while
purely elastic conditions prevail. However, an examination of load-settlement curves for
pile groups obtained by several investigators reveals that these curves are essentially linear for
loads up to one third or one half of the ultimate load. Therefore for normal working loads the
assumption of elastic conditions appears to be justified. Although an approximate analysis
of the effects of local slip on the settlement of pile groups similar to that carried out for a single
pile may be possible, it is outside the scope of this Paper. In all analyses, it is assumed that
the base of the pile cap lies above the surface of the soil (i.e. the groups are free-standing).
If the pile cap rests on the soil the analysis of the problem becomes more complex since raft
action of the pile cap will also occur. A solution to this problem would therefore require the
additional consideration of compatibility between the displacements of the pile cap and the
piles. This aspect of group behaviour will be considered in a subsequent paper.

j-

Pilo 1 PilQ 2

Fig. 1. Group of two piles

Downloaded by [ Columbia University] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 SETTLEMENT OF PILE GROUPS 451

ANALYSIS OF A GROUP OF TWO PILES

The first step in the analysis of the settlement of pile groups is to consider the effects of
interaction between two equally loaded, identical piles. As previously outlined by Poulos and
Davis (1968), the behaviour of each pile may be examined by dividing it into a number of
cylindrical elements each loaded by a uniform load acting around the periphery of the element
and a circular uniformly loaded base. Referring to Fig. 1, the displacement pi of the soil
adjacent to the centre of the periphery of an element i on pile 1 due to pile 1 itself and the
adjacent pile 2 is
j=n
Pi = j~l~,(l~ir+z1,1)+pb(llib+21ib) . . . . . . (I)

where lIi, is the displacement influence factor at element i due to a uniform ring load on
element i on pile 1
Ji, is the displacement influence factor at element i due to a uniform ring load on
element i on pile 2
,Ii, is the displacement influence factor at element i due to a uniform load on the base
of pile 1, and similarly for Jib

A similar expression may be obtained for the displacement p,, of the soil directly beneath the
base of the pile
j=n
Pb = j-‘, $$(11bj+21bf) +pb(llbb +2Ibb) . . . . . (2)

where ll,j is the displacement factor for the pile base due to a uniform ring load on element i
on pile 1, and similarly for Jbj
rI,, is the displacement factor for the pile base due to a uniform load on the base of
pile 1, and similarly for Jbb

For the case of a semi-infinite mass, the displacement factors rIij and 21,1 may be obtained
by integration of the Mindlin equation for vertical displacement, over the cylindrical ring
elements j on pile 1 and pile 2 respectively, for the appropriate points i on pile 1. The factors
lIib and Jib may similarly be obtained by integrations of the Mindlin equation over the circular
base of pile 1 and pile 2. Details of these integrations are given by Poulos and Davis (1968).
For the case of a finite layer it is at present necessary to employ the approximation devised
by Steinbrenner (1934) to obtain values for the influence factors in equations (1) and (2).
From this approximation the influence factor for the displacement of a point X in a layer of
depth h is given by
10-h =IO-+m_Ih-+m . . . . . . . . (3)
where IO’“” is the displacement factor for a point X in a semi-infinite mass
Ih’“O is the displacement of a point directly beneath X at a depth k below the surface
of a semi-infinite mass

The influence factors IO’* and I h+m due to a point load within the semi-infinite mass
may be obtained from the Mindlin equation and integration of these factors for the cylindrical
elements and the circular base will then yield the required influence factors rIij, Jr,, lIib and
2I fb’
For all the elements on pile 1 the vertical displacements of the soil adjacent to the pile may
be expressed in matrix form as
[PI = ([III + [2z1)[$1 +$b([llb! +[2Ibl) . . . . . . (‘)
Since the two piles are identical, this equation also applies to pile 2.

Downloaded by [ Columbia University] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 452 H. G. POULOS

The soil displacement at each element may now be equated to the displacement of each
element of the pile. If this displacement is assumed to be unity, then for the pile elements

[p] = [I] . . . . . . . . . . (5)

Equating (4) and (5) gives

([11l+[zIl)[pl+pb([lrbl+[2lbl)
= 111 . . . . . . (6)
Equation (6) may be solved to obtain the distribution of shear stress along each pile and the
stress on the base for unit displacement of the pile, whence the displacement of each pile for a
unit load may be calculated.
In this manner the influence of an adjacent pile on the displacement of a pile may be deter-
mined for various spacings between the piles.

Solutions for a groupof two piles

Solutions to equation (6) were obtained by the use of the KDF9 computer at the University
of Sydney. In all cases the piles were divided into ten elements. Since the factors lIij and
Jij were previously evaluated for the analysis of a single pile, it was only necessary to carry
out integrations for Jr, and Jib for the relevant spacing between the piles.
The additional displacement of a pile in a semi-infinite mass due to an equally loaded
identical adjacent pile is expressed in terms of an interaction factor 01where

additional settlement due to adjacent pile

CZ=
settlement of pile under its own load

For a group of two piles CIranges between 1 for a zero spacing between the piles (i.e. one
pile superimposed on the other) and zero for an infinite spacing between the piles.
In Fig. 2 CCis plotted against the centre to centre spacing between the piles for various
values of L/d for the case v =0*5. The decreasing effect of interaction with increasing spacing
between the piles is clearly shown in this figure. However, the values of CIat large spacings
are surprisingly large. For example, at a spacing of 40 diameters the value for L/d = 100 is
about 0.2, i.e. the adjacent pile causes a 20% increase in settlement. The relative length L/d

0.2 0.1 d 0
% /s
Fig. 2. Settlement interaction between two piles in a semi-infinite mass

Downloaded by [ Columbia University] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 SETTLEMENT OF PILE GROUPS 453

of the pile has an important influence on the value of cz at any given spacing, a! increasing as
L/d increases. This effect becomes more pronounced as the spacing between the piles in-
creases. The influence of Poisson’s ratio v of the mass is also shown in Fig. 2, where values of
cx are plotted for the case L/d = 25 and v = 0. At any given spacing the effects of interaction
will be seen to be slightly greater for v=O than for v=O*5, although the maximum difference
between the values of CLis only about 0%. Thus the value of v has only a relatively minor
effect on interaction and in the subsequent solutions in this Paper, attention has been con-
centrated on the case v =0.5.
The effect of the presence of an underlying rigid base on the interaction between two piles
is shown in Fig. 3 for the case L/d =25 and v =0*5. Values of CIare plotted against pile spacing
s/d for values of the dimensionless layer depth h/L ranging between co and l-2. As might be
expected the effect of the rigid base is to ‘damp’ interaction, particularly at large spacings
where the value of a: for a finite layer is considerably less than that for a semi-infinite mass.
For close spacings the value of CCis not significantly affected by the layer depth when h/L is
greater than 2.5, but for more shallow layers (y:is markedly reduced by the rigid base, even
for close spacings.

ANALYSIS OF SYMMETRICAL PILE GROUPS

The analysis for the interaction between two piles can be extended to any number of piles,
provided that the arrangement of the piles is such that all piles in the group behave identically.
Such a group will be termed here a symmetrical group. In a symmetrical pile group the piles
are spaced equally around the circumference of a circle and each pile displaces equally while
also carrying an equal amount of load. For a group of m piles, the analogous equation to
equation (6) for two piles is
(L~l+LJl+ ...LJl..
*+r,~l)[Pl+P,([,~,l+r2~,1+.
..LJbl..
.+LJLJ)= PI . (7)
where [,I] is the displacement factor matrix for all the elements of pile 1 due to the ring
elements of pile k
[J,,] is the displacement factor matrix for all the elements of pile 1 due to the base load
on pile k

Fig. 3. Settlement interaction between two piles in a finite layer

Downloaded by [ Columbia University] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 454 H. G. POULOS

L/d= 25
3 -0.5

VL * La

I I I I I f I I I
OO 10 20 30 40
‘/d
Fig. 4. Settlement interaction-symmetrical pile groups

--.-. ,,

Fig. 5. Effect of pile interaction on distribution of Fig. 6. Effect of pile interaction on load
shear stress along pile taken by base

Downloaded by [ Columbia University] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 SETTLEMENT OF PILE GROUPS 455
Equation (7) may be solved as before to obtain the distribution of shear stress and the
displacement of each pile.
For the case L/d =25 and v =0.5, solutions have been obtained for symmetrical groups of
three and four piles. In Fig. 4 the interaction factor CLis plotted against the spacing s/d
between the piles for the above groups and also for a group of two piles. This figure shows that
the additional settlement of each pile in the group due to the adjacent piles is almost exactly
equal to the sum of the displacement increases due to each of the adjacent piles in turn.
Thus, for a group of three piles, the value of 01is twice the value for a group of two piles at
the same spacing, while for a group of four piles at a spacing of s diameters, the value of cc is
given by
a=2a1+az . . . . . . . . . (8)
where CQis the value of CYfor two piles at a spacing of s diameters
CQ,is the value of CYfor two piles at a spacing of 4% diameters
The corresponding displacement of each pile in the group due to equal loads P, in each pile
is thus
p = P,p,(1+2a,+crz) . . . . . . . . (9)
where pr is the displacement of a single pile under unit load
A comparison between the distributions of side shear for a single pile and for groups of two,
three and four piles is shown in Fig. 5 for a spacing of 2.5 diameters between the piles. As the
number of piles increases the shear stress tends to decrease along the major part of the pile
shaft, but increase near the pile tip, although the difference between the four distributions is
generally small.
The percentage of load taken by the base for each pile group is shown in Fig. 6 for various
spacings between the piles. For close spacings interaction causes an increase in the load
transmitted to the base of each pile in the group as compared with the case of a single pile.
The increase in base load becomes more marked as the number of piles in the group increases.
As the pile spacing increases the base load decreases until, at a spacing of about 25 diameters,
it is identical with that for a single pile.
The remarkable applicability of the principle of superposition to symmetrical groups of
three and four piles suggests that the displacement of any symmetrical pile group may be
calculated from the relationship between CIand s/d for a group of two piles, without having to
solve equation (7) directly. Further, it appears reasonable to extend the use of the principle
of superposition to the analysis of pile groups which are not symmetrical.

ANALYSIS OF GENERAL PILE GROUPS

For a general pile group, two limiting conditions may be considered :

(i) equal loads in all piles-this case corresponds to a perfectly flexible pile cap
(ii) equal settlement of all piles-this case corresponds to a perfectly rigid pile cap
For a group consisting of M piles, the displacement of any pile k in the group is, by super-
position,
,=m
plc = p1 2 Pjakj . . . . . . . . (10)
j=l
where ak, is the value of a for two piles corresponding to the spacing between pile k and pile j
Pj is the load on pile j
p1 is the displacement of a single pile under unit load
If the total load on the pile group is P, then
f=t?l
PG = 1 Pj . . . . . . . . . (11)
j=l

Downloaded by [ Columbia University] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 456 H. G. POULOS

For the case of equal loads in all piles, Pi= P&z. Hence equation (10) may be used
directly to calculate the displacement of each pile in the group and hence the maximum
settlement and differential settlement between the piles.
For the case of equal settlement the displacements are equated, giving m simultaneous
equations which may be solved for the unknown loads Pj in the group, whence the settlement
of the group may be calculated.
Thus the analysis of any general pile group requires only a knowledge of the relationship
between the interaction factor 0: and pile spacing for a group of two piles (Figs 2 and 3).
The foregoing analysis relies on the assumption that the principle of superposition holds
for general pile groups as well as for symmetrical groups. In contrast to symmetrical groups,
in general pile groups there will generally be piles between the pile at which the settlement
is calculated and the pile whose influence is being considered. The presence of such inter-
vening piles will have a reinforcing effect on the mass and will tend to decrease the influence
of a pile on another pile. Thus the use of the value of a: obtained from an analysis of two
piles using Mindlin’s equation may be somewhat inaccurate for this problem since Mindlin’s
equation is strictly applicable only to a uniform mass rather than to the pile reinforced mass
which actually exists. At the present time an analysis of the reinforcing effect of piles on the
displacements of an elastic mass is not feasible, and even if such an analysis were available, its
application to this problem would result in an increase in complexity of the solution not
commensurate with any increase in accuracy which might be obtained. Therefore, while
some inaccuracy may result for large groups of piles, the application of the principle of super-
position to general pile groups appears justified in view of the resultant simplification to the
analysis. Further justification of the method of analysis is provided by the quantitative
comparisons between theoretical and observed behaviour of pile groups described later.
In obtaining solutions for the settlement behaviour of general pile groups many variables
must be considered, including the following:

(i) the type of pile group

(ii) the L/d ratio of the piles in the group
(iii) the relative depth h/L of the soil layer
(iv) Poisson’s ratio v of the layer
(v) the spacing s/d between the piles

An exhaustive consideration of these variables is not practicable, so that the effects of them
may only be readily examined in relation to typical cases. Attention in this Paper has been
confined primarily to square groups of 4, 9, 16 and 25 piles, which will be subsequently
denoted as Z2, 32, 42 and 52 groups respectively.
The behaviour of each group is examined for spacings up to 40 diameters between adjacent
piles, for the values of L/d and h/L considered for a group of two piles. Because the effect of
ron the value of C(for a group of two piles is relatively minor, solutions are only reported for
v=O*5. In discussing these solutions consideration is given to the influence of the type of
group, the L/d ratio and the value of h/L on the settlement and the load distribution in a
group with a rigid pile cap, and the maximum and differential settlements of a group with a
flexible pile cap.

SOLTJTIOSS FOR PILE GROUPS WITH A PERFECTLY RIGID PILE CAP

Settlements
The most commonly employed parameter in considering the settlement of pile groups is
the settlement ratio R, which is defined as the ratio of the settlement of the pile group to the
settlement of a single pile carrying the same average load as a pile in the group. In many
cases it is useful to express the settlement of the group in terms of a group reduction factor

Downloaded by [ Columbia University] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 SETTLEMENT OF PILE GROUPS 457
Table 1. Settlement influence factors Z, for single pile
(after Poulos and Davis, 1968)

v=o
h/L\\ L/d j 10 / 25 100 10 25
I_
I 1100
2.54 1.16 1.95
2.44 1.07 1.86
2.31 0.96 1.75
2.11 0.80 1.11 1.58
1.89 I 0.62 1.44
L -
o.g4 /

R, where R, is the ratio of the settlement of the group to the settlement of a single pile
carrying the same total load as the group. For a group of m piles the settlement ratio R, is
related to the group reduction factor R, as
R, = R,/m . . . . . . . . . (12)
The value of R, represents the reduction in settlement which results from using a pile group
instead of a single pile to carry a given load.
The actual settlement of the group is given by
So = RS.S, = RomS, . . . . . . . . (13)
where S, is the settlement of a single pile carrying the same average load as a pile in the group.
From the analysis of a single pile carried out by Poulos and Davis (1968) S, may be ex-
pressed as

s, =& . . . . . . . . (14)

where P is the load on a single pile

L is the length of the pile
E is the Young’s modulus of the surrounding soil mass
I, is the displacement influence factor for a single pile

Since the total load on the group P,=mP the settlement of the group is

S, =gR,I, . . . . . . . (15)

Values of the single pile influence factor I, obtained by Poulos and Davis (1968) are
tabulated in Table I for a range of values of h/L and L/d.
For 22, 32, 42 and 52 groups, values of the group reduction factor R, are tabulated in
Table 2 for the case L/d =25 and Y=0.5. These values of R,, together with the values of I, in
Table 1, provide a rational basis for calculating the settlement of a pile group with a rigid pile
cap in terms of the settlement of a single pile.
From Table 2 it will be seen that the effect of the value of h/L on the settlement of square
pile groups is similar to that on the group of two piles shown in Fig. 3. As h/L decreases the
value of RG decreases at any given spacing because of the damping effect of the rigid base.
The smaller the value of h/L, the closer the spacing at which the limiting value of R, for no
interaction of l/m (corresponding to a settlement ratio R, of 1) is reached.
The effect of the value of L/d is shown in Fig. 7 where R, is plotted against s/d for a 32
group in a semi-infinite mass. As L/d increases, the settlement for any given spacing in-
creases, the increase being of the same order as that for a group of two piles shown in Fig. 2.

Downloaded by [ Columbia University] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
Downloaded by [ Columbia University] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
I I I I I 1
5

o-2 “/, 0 I 0
‘h
Fig. 7. Effect of L/d on settlement-32 group with rigid Fig. 8. Influence of type of group on settlement-groups with rigid
cap cap
 SETTLEMENT OF PILE GROUPS 459
Table 2. Values of RG-pile group with rigid cap. L/d = 25. v = 0.5

Group 22 I 32

5 2.5

__.
1 o.s39 oa19 0+315 0.670 0.593 0.464
.____
2.5 0.672 0.638 0.629 / 0.550 / 0.443 0.541 / 0.495 0.479 0.387 0.283

5 0.547 0.519 0.501 0.422 0.348 0.415 0.363 0.339 0.256 0.195
___,
10 0.425 0.408 0.385 0.323 0.291 0.303 / 0.245 0.220 0.165 0.141

20 0.366 0.317
~.___
40 ! 0.307 I 0.260

Group

5 0.334 0.277 0.250 0.176 0.128 0.281 0.220 0.194 0.129 0.091
__.
10 0.227 0.166 0.143 0.100 0.083 0.180 0.119 0.100 0.067 0.055
~.
20 0.148 0.093 0.083 0.069 0.066 0.112 0.062 0.054 0.045 0.042
___
40 0.105 0.066 0.064 0.063 0.063 0.070 0.041 0.041 0.040 0.040

Fig. 7 therefore provides some basis for correcting the values of I?, in Table 2 to allow for dif-
ferent values of L/d.
The effect of the number of piles in a group on the value of R, is shown in Fig. 8 for the
case L/d = 25 and h/L = co. As well as square groups values of R, for groups of 2, 3, 5 and 6
piles are plotted. For the group of 5 piles, the fifth pile is situated centrally between the outer
four piles, the spacing s being that between the outer piles. The reduction in settlement with
an increasing number of piles in the group, for a given spacing between the individual piles,
is clearly shown in this figure. However, at relatively close spacings (s/d<5) the use of an
increased number of piles to reduce settlement becomes increasingly ineffective if the same
spacing between individual piles is retained. For example, the settlement of a 52 group of
piles at a spacing of 5d is only about 20% less than that of a 42 group of piles at the same
spacing. It is interesting to note that, for the group of 5 piles, the settlement is identical with
that of a 22 group for spacings up to about lOd, i.e. the additional centre pile has no effect in
reducing the settlement.
For a given size of pile cap the pile spacing will vary as the number of piles in the group is
varied. The variation of R, with foundation breadth (centre to centre spacing between the
outermost piles) for various square groups is shown in Fig. 9. For all but the group of 4 piles
the value of R, for all groups is identical up to a breadth of about 16d. Beyond this breadth,
R, tends to decrease as the number of piles in the group increases. Fig. 9 therefore suggests

Downloaded by [ Columbia University] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 460 H. G. POULOS

Fig. 9. Settlement against breadth of group-rigid pile cap

Fig. 10 (opposite, above). Relative importance of immediate settlement-pile groups with rigid
cap

Fig. 11 (opposite, below). Equivalent length of single pier for same settlement as pile group

that over a considerable range of breadths the settlement of a pile group is primarily dependent
on the breadth of the group rather than the number of piles in the group. The results of
full-scale field tests carried out by Berezantzev, Khristoforov and Golubkov (1961), and data
collected by Skempton (1953) confirm this conclusion.
Thus in the design of a pile group, the use of a small number of piles at a relatively large
spacing is generally to be preferred to a larger number of piles at a closer spacing on the grounds
of economy. However, in practical problems there may be a limit to the spacing between
piles beyond which it is difficult to construct economically a pile cap of sufficient rigidity.
The relative importance of immediate settlement for various pile groups is shown in Fig. 10
where the ratio of the immediate settlement S, to the total final settlement S,, of the group is
plotted against pile spacing. The case L/d =25 and h/L = 00 has been considered and Poisson’s
ratio V’of the soil skeleton has been taken as zero. The calculation of the ratio .5,/S,, has been
described by Poulos and Davis (1968) in connexion with a single pile. As with a single pile
the predominant part of the total final settlement occurs as immediate settlement although
the ratio S,/S,, is generally less than for a single pile. As the number of piles in the group
increases SJS,, decreases, but even for a group of 25 piles, between 60% and 70% of the total
settlement occurs immediately. For most real soils V’will be considerably greater than zero

Downloaded by [ Columbia University] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 SETTLEMENT OF PILE GROUPS 461

SO that the relative proportion of immediate settlement will be even greater than that shown
in Fig. 10. Hence it may be concluded that consolidation settlement is generally of second-
ary importance for pile foundations and that accurate predictions of rates of settlement are
therefore not essential.
Figure 10 suggests a possible reason for the generally poor agreement between measured
settlement of pile groups and that predicted by approximate methods which rely on the use
of one dimensional consolidation theory. In one dimensional theory, all the settlement is
assumed to derive from pore pressure dissipation, whereas with pile foundations the major
proportion of the settlement occurs at constant volume under undrained conditions and only
a small amount is due to pore pressure dissipation. Therefore the tendency of approximate
methods to underestimate the settlement may be attributed either to the total neglect of the
immediate settlement if a purely one dimensional approach is employed, or to the underesti-
mate of S, which will occur when the pile group is replaced by an equivalent pad footing in
an approximate three dimensional analysis.
In making preliminary estimates of settlement it may be useful to replace the pile group
by an equivalent single circular pier of the same gross area as the group. The length L, of
such an equivalent pier may be obtained by comparing the solutions for the settlement of a

Downloaded by [ Columbia University] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 462 H. G. POULOS

pile group, as given in this Paper, with the settlement of a single pier (Poulos and Davis, 1968)
and determining the required value of L, in terms of L to give equal settlement of the pier and
the group. For the case h/L = co and v=O*5, values of L,IL have been plotted against the
pile spacing in Fig. 11 for values of L/d of 10, 25 and 100. It is interesting to note that the
value of L,IL for any spacing is not significantly influenced by the number of piles in the group
so that the curves shown in Fig. 11 are mean curves which apply to all square groups between
22 and 52.
The value of L,IL is markedly dependent on the pile spacing and the value of L/d. As the
spacing increases L,IL decreases, indicating that the group tends increasingly to behave as a
shallow pier. At a spacing of about 10d for L/d = 10 and 20d for L/d =25, L,IL becomes zero
and the pile group may then be considered as a surface pad footing. As L/d increases the
equivalent length L,IL increases for any given spacing, and for very slender piles (L/d = 100)
the spacing at which L,.L becomes zero and the group can be considered as a surface footing is
in excess of 60d.
For a pile group in a finite layer, the limiting spacing beyond which the group effectively
acts as a surface footing is even less than that shown in Fig. 11 due to the damping effect of the
rigid base. However, at the relatively close spacings commonly employed in practice the
rigid base will have little effect on the value of L,/L, so that it may generally be stated, from
Fig. 11, that L,IL will lie between 0.9 and 0.6 in most practical cases.

Load distribution within the group

The ratio P/P,, of the pile load to the average pile load in the group for 32, 4= and 5a
groups is tabulated in Table 3 for the case L/d =25 and h/L = co. The key for the identifi-
cation of the piles in each group is shown in Fig. 12. For close spacings, the load in the
outer piles of each group is considerably more than the average load P,, while the load
in the centre piles is less than P,,, and may even theoretically be negative (although it
is doubtful whether negative loads would occur in practical cases). As the pile spacing
increases the load distribution becomes more uniform, the load on the outer piles decreasing
while the load on the centre piles increases. From Table 3 it will be seen that the load distri-
bution is considerably influenced by the number of piles in the group, the major influence
being that the load on the outer piles increases as the number of piles increases.
The influence of the layer depth h/L is shown in Fig. 13 for a 32 group. As the layer
becomes more shallow the load distribution becomes more uniform at any given spacing. For
the case h/L = 1.2, all piles in the group carry equal load at a spacing of 20 diameters, whereas
for a semi-infinite mass (h/L = 00) a uniform load distribution is not attained until a spacing of
about 60 diameters is reached.
The effect of the L/d ratio on the load distribution within a 32 group in a semi-infinite mass
is shown in Fig. 14. The distribution for L/d = 10 is more uniform than that for L/d =I00

Fig. 12. Identification of piles in square groups

Downloaded by [ Columbia University] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 OE
1
“4 02
I
01
I
01
N
I1

Downloaded by [ Columbia University] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
2.
001=p5 --- z.1 = ‘/u ------
ot=“/1 ~
m=
‘4
J -0
__T) I-, *
_----
__-_-
__--- #’

!
/’

9
-- ---_
‘\
I --h)
-w
 464 H. G. POULOS

although the difference between the two distributions is not significant for practical purposes.
In comparison with the type of group and the layer depth, the value of L/d therefore has a
relatively minor influence on the load distribution in a pile group.

SOLUTIONS FOR PILE GROUPS WITH A PERFECTLY FLEXIBLE PILE CAP

Maximum settlement
For a general square pile group in which all piles take equal load the maximum settlement
occurs at the centre pile or piles, while the minimum settlement occurs at the corner piles.
For the limiting cases of zero spacing and infinite spacing, the maximum settlement S,,,
of a group with a flexible pile cap is equal to the settlement of S, of a group with a rigid pile
cap. For intermediate spacings, S,,, is generally greater than Sn, and comparisons have re-
vealed that the ratio S max/SR is almost independent of pile spacing for a practical range of
spacings up to about 2Od. The estimation of the maximum settlement of a group with a flex-
ible pile cap may therefore be greatly facilitated by multiplying the settlement of a group with
a rigid pile cap S, by the appropriate value of S msx/SR, S, being calculated from the values
of R, given in Table 2.
Values of SmiLx/SR are tabulated in Table 4 for the case L/d =25 and v=O*5 for various
values of h/L and for 32, 4= and 5a groups. It will be seen from Table 4 that S,,,/S, increases
as the number of piles in the group increases. In the limit, this ratio would tend to the ratio

Table 3. Values of PIPav in pile group with rigid cap. L/d = 25. v = O-5. h/L = co

Group 32 4= 52

Pile Pile Pile

sld
1 2 3 1 2 3
-- -~ 1l23_%_J56
1 2.31 0.33 - 1.57 2.99 0.75 -0.49 4.12 0.75 1.87 -1.56 0.03 1.16
~-___-___~ -__ / I ~___--
2.5 1.52 0.74 -0.05 2.02 0.96 0.05 2.58 1.18 1.16 0.01 0.10 0.19
-__~ -- ~__-_____--
5 1.32 0.84 0.34 1.71 0.99 0.31 2.11 1.20 1.08 0.35 0.27 0.22
-~___ --_- ______~__
10 1.22 0.89 0.55 1.49 1.oo 0.50 1.74 1.16 1.07 0.56 0.47 0.37
--~--____ -~___
20 1.13 0.94 0.72 1.26 1.00 0.73 1.40 1.09 1.05 0.74 0.71 0.67
___-~ ~~ -/- __-
40 1.06 0.98 OM 1.03 0.99 0.98 1.27 I.07 1.02 0.84 0.80 0.75

Table 4. Values of S,,,/S,. L/d = 25. v = 0.5

I
Group 1 32 42 52

‘1
h/L
co 1.5 1.2 cc 1.5 1.2 co 1.5 1.2
sld
1.13 1.15 1.15 1.13 1.17 1.18 1.18 1.25 1.26
i.5 1.13 1.17 1.16 1.14 1.20 1.17 1.19 1.30 1.24
5 1.13 1.18 1.13 1.15 1.20 1.15 l-21 1.30 1.23
10 1.14 1.15 1.10 1.16 1.16 1.11 1a24 1.20 1.11
20 1.14 1.05 1.01 I.13 1*OS 1.Ol I.18 1.04 1.02
40 1.08 1*oo 1.00 1.06 1.00 1.oo 1.17 1.oo 1*oo

Downloaded by [ Columbia University] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 L ~25
Group /d
h zoo
/L
I)=05 v -0.5

I
,-
>

Downloaded by [ Columbia University] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
0 10 20 30 0 10 20 40
%

Fig. 15. Effect of group size on m-iximum differential settle- Fig. 16. Effect of layer depth on maximum differential settle-
ment-pile group with flexible cap ment-3a group with flexible cap
 466 H. G. POULOS

between the corresponding settlements for a square area. The value of h/L has a relatively
small influence on S ,,,JSR at close spacings, but for s> IOd, S,,,/S, decreases rapidly as h/L
decreases.
The ratio Smax/.SR is also slightly influenced by the value of L/d, decreasing as L/d increases.
For example, for a 32 group the average value of S ,,J.S, for L/d=10 is I.15 whereas the
corresponding value for L/d = 100 is l-10.
However, in summary it may generally be stated that the use of a flexible rather than a
rigid pile cap will result in an increase in maximum settlement of between 10% and 30%.

Maximum diff’erential settlement

For various groups of piles with a flexible pile cap, the ratio Sa/Smpx of the maximum
differential settlement of the group to the maximum settlement is plotted in Fig. 15 for the
case L/d =25 and h/L= a~. At the limiting spacings of zero and infinity the differential
settlement is zero, while at intermediate spacings the differential settlement increases to a
maximum value at a spacing of about 15d and then decreases for larger spacings. The ratio
Sd/Sm*x increases as the number of piles in the group increases and for a 52 group, the maximum
differential settlement may be as much as about O-3 times the maximum settlement of the
group.
The effect of the layer depth h/L on the ratio .5,/S,,, is shown in Fig. 16 for a 32 group.
At close spacings the presence of a rigid base tends to increase the relative differential settle-
ment, but as the spacing increases .S,jS,,, decreases more rapidly than is the case with an
infinite depth.
The effect of L/d on the differential settlement is similar to the effect of the layer depth Iz/L.
At close spacings Sd/Smax increases as L/d decreases, while at larger spacings the greater effects
of interaction with relatively slender piles results in S,jS,,, being larger for larger values of
L/d.

COMPARISONS BETWEEN THEORETICAL AND OBSERVED BEHAVIOUR OF PILE

GROUPS

An indication of the applicability of the theoretical solutions presented in the previous

sections to field problems may be obtained by comparisons between these solutions and pub-
lished records of observations on model and full-scale pile groups. The results of tests on
groups of model piles have been reported by Whitaker (1957), Sowers et al. (1961), Hanna
(1963) and Saffery and Tate (1961), while tests on full-scale groups have been carried out by
Berezantzev, Khristoforov and Golubkov (1961).
Comparisons between the theoretical and observed settlements of 22, 32, 42, and 52 groups
with rigid pile caps are shown in Figs 17-20 in which the settlement ratio R, is plotted against
pile spacing. A value of L/d=25 has been assumed while Y has been taken as 0.5. Because
the depth of the soil layer in which the groups are founded differs in each case and in some cases
is not stated, theoretical curves have been drawn for two cases, h/L = co and h/L = l-5, which
represent the limits between which the experimental results probably lie. For all groups it
will be seen that the theoretical curves generally agree with the observations and predict the
decrease in settlement ratio with increasing spacing reasonably accurately. It is also interest-
ing to note that there is some measure of agreement between the theoretical curves and the
results of the model tests in sand carried out by Hanna (1963).
Skempton (1953) has also collected data from a number of field tests on groups of driven
piles in sand, and on the basis of these data has suggested a single relationship between the
breadth of a group and the settlement ratio R, for a given average applied pressure. For a
given total load on the group, Skempton’s suggested curve, in terms of the group reduction
factor R,, is shown in Fig. 9, together with the theoretical curves for L/d =25 previously

Downloaded by [ Columbia University] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 SETTLEMENT OF PILE GROUPS 467

Measured values-

0 Whltaker (1957)
Eerezontzev et 01 (1961)
Sowers et al (1961)
Hanno (1963) loose sand
x ., .,
dense sand
Saffery and Tate (1961)

L,d=25

Fig. 17. Comparison between theoretical and Fig. 18. Comparison between theoretical and
measured settlement ratios-2z group measured settlement ratios-3’ group

+
=

-1, M005WOd

0
a
+
*
Sowers
Homa
,,
“oI”es-

Berezantzev
et
(1963)
),
01
et al

loose
dense
(1961)
(1961)
5rJ”d
sand

+
MeOwred values -

0 Whltaker(l957) $, = 24

. ,,
” ““48
61 Berezantzev et 01 (1961)
c Henna (1963) loose sand
* .I I.
dense sand
L/d = 25

Fig. 19. Comparison between theoretical and Fig. 20. Comparison between theoretical and
measured settlement ratios-42 group measured settlement ratios-S’ group

Downloaded by [ Columbia University] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 468

E f

I
I
\

\
a ‘1
Q
a ‘\ ‘.
0 a t---_
---

Downloaded by [ Columbia University] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 SETTLEMENT OF PILE GROUPS 469
described. It will be seen that Skempton’s curve lies somewhat above the theoretical curves,
but follows the theoretical trend very closely. The actual points on which Skempton’s curve
is based are also plotted in Fig. 9 and these points, within the range of values of group breadth
to diameter shown, agree even more closely with the theoretical curves than does Skempton’s
curve.
Comparisons between measured and theoretical distributions of load within a pile group
with a rigid pile cap are shown in Figs 21 and 22. In Fig. 21 the results of tests on a 32 group
by Whitaker (1957) and Sowers et al. (1961) are plotted in terms of the ratio of the pile load to
the average pile load. For comparison the theoretical curves for L/d = 25 and v =0*5 and for an
average value of h/L of 25 are shown. The theoretical magnitudes of load agree well with the
results obtained by Sowers et al. except for the centre pile 3, where, at close spacings, the
measured load is small but not negative as the theory suggests. The load distribution ob-
tained by Whitaker is considerably more uniform than that obtained by Sowers et al. or that
given by the theory. Despite this difference in magnitude, for which no reason can be as-
certained, Whitaker’s results nevertheless confirm that the outer piles carry the most load and
the centre pile the least, and that the load distribution within the group becomes more uniform
with increasing pile spacing.
The theoretical load distribution for a 52 group is compared with the observations of
Whitaker (1957) in Fig. 22. As well as giving the correct trend the theory predicts magnitudes
of loads which agree very well with the measured values, for all piles except piles 4, where
the measured loads are greater than those predicted.
From the foregoing comparisons between theoretical and observed settlements and load
distributions, it may generally be concluded that the theoretical approach described in this
Paper not only enables broad trends in pile group behaviour to be studied, but is also capable
of predicting the magnitude of settlement and loads within the group with reasonable
accuracy.
In applying the theoretical solutions to field problems, it must be borne in mind that the
theory at present takes no account of various aspects which may influence the behaviour of a
group, such as the order of driving of the piles and layering of the soil profile. Also the theory
does not take into account any raft action between the pile cap and the soil. Any such action
in a field problem may well have a significant effect in reducing the settlement of the group.
Because of the difficulty in determining appropriate values of the soil parameters E,, E ’and
v’ for predictions of the immediate and total final settlement, it may be more satisfactory in
many practical problems to carry out a field loading test on a single pile and then to apply the
theoretical values of settlement ratio R, to the results of this test in order to predict settle-
ments. With such a procedure, uncertainties in the experimental determination of soil
parameters are obviated and variations of soil type or soil properties with depth are largely
taken into account. It must, however, be borne in mind that such non-homogeneity may
modify the interaction between piles as compared with the case of a homogeneous mass, and
thus limit the validity of the theoretical curves in this Paper. A study of the modifying
effects of non-homogeneity of the soil on pile interaction may well be possible in the future
by means of the finite element technique.

CONCLUSIONS

The method of analysis presented in this Paper enables the settlement behaviour of all
configurations of pile groups to be determined by considering only the effects of interaction
between two piles and superimposing the individual effects of adjacent piles in the group.
It is therefore possible to relate the settlement of a pile group to the settlement of a single
pile.
Solutions have been obtained for the settlement and the load distribution in a group with a

Downloaded by [ Columbia University] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 470 H. G. POULOS

perfectly rigid pile cap and for the maximum settlement and the maximum differential settle-
ment of a group with a perfectly flexible pile cap.
The influence of the type of group, the relative length L/d of the piles in the group and the
relative depth h/L of the soil layer on the settlement behaviour of a pile group have been
examined. The main conclusions may be summarized as follows.

1. For a group of two piles the interaction between the piles at a given spacing increases
as the values of L/d and layer depth h/L increase. For slender piles in a deep layer of soil the
increase in settlement of a pile due to an identical pile may be very significant, even for
spacings as large as 50 diameters. However, as the depth of the soil layer decreases the effect
of interaction tends to be reduced considerably as compared with the case of a semi-infinite
mass. The value of Poisson’s ratio Y of the soil also slightly influences interaction, the inter-
action factor CLtending to increase as v decreases.
2. Because the behaviour of a general pile group is derived from the behaviour of a group
of two piles, the effects of the values of L/d, h/L and v on the settlement of the group will be
similar to those for the group of two piles.
3. For any given spacing between the piles the settlement of a pile group is largely in-
fluenced by the number of piles in the group. The larger the number of piles, the smaller the
settlement under a given load.
4. For a pile group of given dimensions the settlement of the group under a given load is
almost independent of the number of piles in the group for a practical range of spacings.
From a settlement consideration it is therefore more economical to employ a smaller number of
piles at a larger spacing.
5. As with a single pile, the major proportion of the settlement of a pile group occurs as
immediate settlement, although this proportion is generally smaller for a group than for a single
pile.
6. The load distribution within a pile group with a rigid pile cap is generally non-uniform
at closer spacings, the outer piles taking the most load and the centre piles taking the least
load. At any given spacing the load distribution tends to become more non-uniform as the
relative pile length L/d and the depth of layer h/L increase.
7. The ratio of the maximum settlement of a group with a perfectly flexible pile cap to the
settlement of an identical group with a rigid pile cap generally varies between 1.1 and 1.3, the
higher values being associated with larger groups.
8. The maximum differential settlement in a group with a flexible pile cap increases as the
number of piles in the group increases, and for any group is a maximum at a spacing of about
15 diameters.
9. Comparisons between theoretical settlements and load distributions and those observed
in field and model tests show that, as well as predicting the correct trends in behaviour of pile
groups, the theory is capable of predicting the magnitude of settlements and loads within the
group with reasonable accuracy.

ACKNOWLEDGEMENTS

The major part of the work described in this Paper was made possible by a grant from the
Australian Research Grants Committee. Valuable assistance has also been received from
Frankipile Australia. The computational work was carried out on the KDF9 computer in
the Basser Computing Department of the School of Physics, University of Sydney. The
Author gratefully acknowledges the comments and advice of Professor E. H. Davis, Associate
Professor of Soil Mechanics, and Dr I. K. Lee, Senior Lecturer in Civil Engineering, University
of Sydney.

Downloaded by [ Columbia University] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 SETTLEMENT OF PILE GROUPS 471
REFERENCES
BEREZANTZEV,V. G., KHRISTOFOROV,V. & GOLUBKOV,V. (1961). Load bearing capacity and deformation
of piled foundations. Proc. 5th Id. Conf. Soil Mech. 2, 11-15.
BJERRUM, L., JONSON, W. & OSTENFELD, C. (1957). The settlement of a bridge abutment on friction piles.
Proc. 4th Int. Conf. Soil Mech. 2, 14.
D’APPOLONIA, E. & ROMUALDI, J. P. (1963). Load transfer in end-bearing steel H-piles. Am. Sot. civ.
Engrs. 89, SMZ, l-25.
HANNA, T. H. (1963). Model studies of foundation groups in sand. G&otechnique 13, No. 4, 334-351.
POULOS, H. G. & DAVIS, E. H. (1968). The settlement behaviour of single axially-loaded incompressible
piles and piers. Gkotechnique 18, No. 3, 351-371.
SAFFERY, M. R. &TATE, A. P. K. (1961). Model tests on pile groups in a clay soil with particular reference
to the behaviour of the group when it is loaded eccentrically. Proc. 5th Znt. Conf. Soil Mech. 2, 129-134.
SALAS, J. A. J. & BELZUNCE, J. A. (1965). Resolution theorique de la distribution des forces dans les pieux.
Proc. 6th Int. Conf. Soil Mech. 2, 309-313.
SKEMPTON, A. W. (1953). Discussion: piles and pile foundations, settlement of pile foundations. Proc.
3rd Int. Conf. Soil Mech. 3, 172.
SOWERS, G. F., MARTIN, C. B., WILSON, L. L. & FAUSOLD, M. (1961). The bearing capacity of friction pile
groups in homogeneous clay from model studies. Proc. 5th Int. Conf. Soil Mech. 2. 155-159.
STEINBRENNER,W. (1934). Tafeln znr Setzungberechnung. Strasse 1 121.
THURMAN, A. G. & D’APPOLONIA, E. (1965). Computed movement of friction and end-bearing piles em-
bedded in uniform and stratified soils. Proc. 6th Int. Conf. Soil Mech. 2, 323-327.
WHITAKER, T. (1957). Experiments with model piles in groups. Gdotechnique 7, No. 4, 147.

3*

Downloaded by [ Columbia University] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.