You are on page 1of 13

RANDOLPH,M. F. (1981). G&technique 31, No.

2, 247-259

The response of flexible piles to lateral loading

M. F. RANDOLPH*

The majority of piles encountered in practice may be NOTATION


regarded as ‘flexible’ under lateral loading. That is, the
undrained shear strength
induced deformations and bending moments are confined
Young’s modulus
to the upper part of the pile and the overall length of the
pile does not significantly affect the response of the pile. effective Young’s modulus of pile
The results of a parametric study, conducted using the Young’s modulus of soil
finite element method and treating the soil as an elastic bending rigidity of pile
continuum with a linearly varying soil modulus, are shear modulus of soil
presented as simple algebraic expressions. These expres- product of G( 1 + 3v/4)
sions are similar in form to those which arise from a average value of G* over active length of
Winkler idealization of the soil, enabling immediate pile
estimates to be made of: the active length of the pile; the lateral load on pile
ground level deformations; and the maximum bending
stiffness ratio of pile
moment down the pile. In addition, expressions are
presented quantifying the effect of interaction between
coefficient of subgrade reaction, also
neighbouring piles. from which the behaviour ofgroups of stiffness H/u of pile
piles subjected to lateral loading may be deduced. length of pile
Application of the solutions is demonstrated by the critical length of pile
analysis of lateral loading tests on single piles and on moment applied to pile at ground level
groups of piles. rate ofincrease of soil shear modulus with
depth
On peut considkrer la majorit des pieux rencontrts dans
la pratique comme ‘flexibles’ dans des conditions de
product of m( 1 + 3v/4)
charge latkrale. Les dkformations induites et les moments radius of pile
de flexion sont restreints g la partie supirieure du pieu et spacing between pile centrelines
la longueur globale du pieu n’a pas d’effet significatif lateral deflexion of pile at ground level
sur la rkponse de celui-ci. Les rtsultats d’une itude depth
paramktrique faisant appel g la m6thodedesirltments finis interaction coefficient
et dans laquelle le sol esf reprtsentt par un continuum general deformation
klastique dont le module est $ variation linkaire, sont rotation of pile at ground level
prt%ent&s sous forme d’expressions algibriques simples.
Poisson’s ratio
Les expressions sont semblables du point de vue de la
mathematical constant
forme B celles obtenues i partir d’une idbalisation Winkler
du sol, et permettent l’estimation immtdiate: de la factor giving relative homogeneity of soil
longueur active du pieu; des deformations au niveau du angle from the direction of loading of the
sol; et du moment de flexion maximum en bas du pieu. En pile
outre, I’article prksente des expressions permettant de
quantifier Peffet de l’interaction entre pieux adjacents, et coefficient of subgrade reaction for the soil (from
de dkduire le comportement de groupes de pieux soumis B the analogous problem of a strip foundation).
une charge la&ale. L’application des solutions est Closed form solutions for this idealization are
dbmontrke par I’analyse d’essais de charge 1atCrale sur des available from Hetenyi (1946) for a variety of
pieux uniques et sur des groupes de pieux. loading conditions and end restraints on the pile.
The solutions for the deflexion u and rotation 8 of
INTRODUCTION the loaded end of the pile reduce to a particularly
The simplest idealization of the problem of a simple form for the case of long piles. For a pile of
laterally loaded pile is that of an elastic beam, bending rigidity (EZ),, embedded in soil with co-
loaded transversely and restrained by uniform, efficient of subgrade reaction k,’ there is a critical
linear springs acting along the length of the beam. length of pile beyond which the pile behaves as if it
The spring stiffness is generally referred to as the is infinitely long (and thus the length of the pile does

Discussion on this Paper closes 1 September, 1981. For ’ k is defined here as the ratio of the load per unit length of
further details see inside back cover. pile to the local deflexion, thus haviqs the same units as
*Cambridge University. modulus.
247
248 M. F. RANDOLPH

not influence the deformation at the loaded end). method since the pile is modelled more accurately.
This critical length may be expressed as Also, heterogeneous soil conditions may be readily,
and correctly, modelled.
l,-4[(El),/k]1’4 (1) To date, the pile designer must choose between
For piles that are longer than their critical length, the succinct solutions available using the simple
the solutions of Hetenyi (1946), for a pile loaded by idealization of subgrade reaction (for example
a lateral load H and bending moment M, yield equations (2)) and the more cumbersome solutions
expressions for the deflexion u and rotation 8 at the in chart form provided by the integral equation
loaded end given by approach. This Paper describes the results of a
parametric study of the response of laterally loaded
piles embedded in an elastic soil continuum. The
study was conducted using the finite element
method and the results are fitted by algebraic
expressions, similar in form to equations (2), from
which the lateral response of single piles may be
From this simple idealization, analytical readily calculated. In addition, the patterns of soil
approaches for laterally loaded piles have movement around a laterally loaded pile, obtained
developed in two separate directions. The first of from the finite element analysis, are used to develop
these retains the conceptual model of treating the expressions giving interaction factors between
soil restraint as discrete springs. The model is neighbouring piles, by which means the solutions
improved by allowing the spring stiffness to vary for single piles may be extended to deal with pile
along the length of the pile (Reese & Matlock, 1956; groups.
Matlock & Reese, 1960) and, subsequently, by
replacing the linear springs by non-linear p-y SINGLE FLEXIBLE PILES
curves (Matlock & Ripperger, 1958; Gill & Demars, In practical applications, few piles deform over
1970; Matlock, 1970; Reese, Cox & Coop, 1975). their whole length under lateral loading. Rather,
The limitations of this approach are twofold. the deformations and induced bending moments
Firstly, difficulties exist in choosing appropriate py reduce to negligible proportions within a few
curves for a given combination of pile size and soil (typically less than 10) diameters of the ground
type. Secondly, replacement of the soil continuum surface. As such, the length of the pile is rarely a
by discrete springs precludes the extension of the relevant parameter when developing solutions for
analysis to pile groups since interaction between laterally loaded piles.
neighbouring piles may not be taken into account. The traditional form of presentation of influence
The second development in solutions for factors for laterally loaded piles (e.g. Poulos, 1971a)
laterally loaded piles has made use of the integral is unnecessarily complicated for the majority of
equation (often referred to as the boundary piles encountered in practice. Dimensional analysis
element) method of analysis, modelling the soil as a shows that, if the pile length is immaterial, any
homogeneous elastic continuum (Poulos, 1971a, b, particular influence factor will be a function solely
1972). The pile is idealized as an infinitely thin strip of the stiffness ratio E,/G, where E, is the effective
of the same width and bending rigidity as the Young’s modulus of the pile, defined as
prototype pile. This idealization and the
subsequent extension of the method to deal with E, = (E0/(n&4) (3)
layered soil profiles (Poulos, 1973; Banerjee & and of Poisson’s ratio for the soil. Thus the
Davies, 1978) introduce a degree of approximation influence factor may be depicted by a single set of
into the solutions. The accuracy of the solutions is curves showing the variation with stiffness ratio
also affected by the manner in which the pile is (one curve for each value of Poisson’s ratio) (e.g.
discretized. Evangelista & Viggiani (1976) have Kuhlemeyer, 1979a). In Poulos’s formulation
pointed out that the use of elements of varying size (followed also by Banerjee & Davies, 1978) the pile
down the pile gives improved accuracy over the length is introduced into the stiffness ratio K,,
solutions given by Poulos (1971a). defined as
Analysis of single laterally loaded piles is also
K, = (EI),/E, i4 (4)
economically viable using the finite element
method in conjunction with Fourier techniques where E, is the Young’s modulus ofthe soil. Hence a
(Wilson, 1965). Such a technique has been used, set ofcurves is needed with a different curve for each
for example, by Carayannacou-Trezos (1977), value of pile length. This set of curves must then be
Randolph (1977) and Kuhlemeyer (1979a, b). In repeated if the effect of Poisson’s ratio is to be
principle, the method enables a more rigorous allowed for. For any given pile length, interpolation
solution to be achieved than the integral equation is generally required between the published curves
RESPONSE OF FLEXIBLE PILES TO LATERAL LOADING 249

of Poulos (1971a). The expressions presented below 0.16- Finite element x Randolph, 1977
avoid such interpolation and are valid for all piles results I + Kuhlemeyer, 1979
which are longer than the critical length 1, beyond ~ 0.12. %\
which the pile length no longer affects the response ;_ \
0
under lateral loading. Such piles will be termed ,o 0.08 -
5
flexible.
The finite element formulation used in the
development of the expressions has been described
by Randolph (1977). The method is similar to that 0
employed by Kuhlemeyer (1979a) except that linear (a)
strain triangles were used, thus avoiding the need 0,06-
for special reduced integration techniques, as L
described by Kuhlemeyer (1979b) for rectangular
elements. The formulation was tested by analysing
0.05- \
a free-standing cantilever and also a laterally
0.04 -
loaded rigid punch on the surface of an elastic half-
$
space. Computed deformations were accurate to & 0.03- \
better than 1% for the cantilever and about 10% for N0 *
the rigid punch (Randolph, 1977). For the problem ;
c-02 - \
of a laterally loaded pile, which is to some extent a Equations (7)
combination of the above two problems, accuracies
0.01 - \<
of better than 10% may therefore be expected.
A parametric study was performed for piles in
0 *
homogeneous soil, characterized by a shear 10’ 102 103 104 10s 10s
modulus G and Poisson’s ratio v, and also in soil EpIG*
with stiffness proportional to depth. The latter type W
of soil may be characterized by a parameter m, Fig. 1. Variation of pile deflexion with stiffness ratio
giving the rate of increase of shear modulus with (homogeneous soil): (a) lateral force H; (b) lateral moment
depth M

01 I I I I I

10 102 10s 104 1 o5 10s


ED/G*

Fig. 2. Pile deflexion due to lateral force for homogeneous soil (equivalent to I,,; Poulos, 1971a)
250 M. F. RANDOLPH

G = mz = mr, z/r0 (5)


For a particular pile radius r,, it is often more
convenient to use the product mr, (the rate of
increase of shear modulus with each pile radius)
which has the same dimensions as shear modulus.
The range of stiffness ratios covered in the
parametric study was lOO<E,/G< lo6 and
400 < E,/mr, ~4 x 106. In addition, the pile length
0 I
was varied in order to correlate the critical pile
(4 length (or, more correctly, the critical slenderness
ratio, (I/r,),) with stiffness ratio.
It was found that the effect of variations in
Finite
element x Randolph, 1977 Poisson’s ratio on the deformation of a laterally
loaded pile could be adequately represented by
considering a parameter G* given by
G* = G(l + 3v/4) (6)
(or, for soil with stiffness proportional to depth,
m* = m(1 + 3v/4)). In effect, this reflects a greater
dependence of the pile deformations on the (hori-
Equations(8) zontal) Young’s modulus of the soil, ES = 2G( 1+ v),
than on the shear modulus. This is in contrast to the
0 case of an axially loaded pile, where the vertical
102 103 104 105 106 10' settlement depends primarily on the shear modulus
Ep* r. G and is relatively unaffected by variations in
03 Poisson’s ratio (with G held constant). In the
Fig. 3. Variation of pile detlexion with stiffness ratio (soil expressions presented below, the parameters G*
stiffness proportional to depth): (a) lateral force H; (h) and m* will be used in order to avoid the necessity
lateral moment A4 of including separate solutions for different values
of Poisson’s ratio.

0.07-

0.06-

O-05-

<
'F 0-04-
No
;
o Banerjee& Davies,1976
003-

0.02-

O.Ol- (8)

0.
10' 103 104 105 106 10'
Ep* r.

Fig. 4. Pile deflexion due to lateral force for soil stiffness proportional to depth (equivalent to I,,‘; Poulos, 1973)
RESPONSE OF FLEXIBLE PILES TO LATERAL LOADING 251

For homogeneous soil conditions it was found


,Pile
that the pile deformations at ground level obtained
from the finite element analyses could be modelled
by

(7)

Figure 1 shows the deflexions computed from the


finite element analyses, together with the curves
obtained from equations (7). Pile deflexions
tabulated by Kuhlemeyer (1979a) have also been
included and show good agreement with the Fig. 5. Definition of p, and G,
Author’s own results. The agreement between the
finite element results and the above expressions is
generally better than 3% except for the case of very equations (8). The agreement is generally better
stiff piles (E,/G > 5 x 104) under lateral load H (Fig. than 2% except for the case of E,/mr, = 400 where
l(a)) where the discrepancy rises to above 10%. the discrepancy is 7%. Figure 4 shows a comparison
Kuhlemeyer (1979a) also fits his finite element of the pile deflexion due to a load H, with integral
results by simple power law relationships. In place equation results published by Poulos (1973) and by
of the exponents - l/7, - 317 and - 517 above, he Banerjee & Davies (1978). The agreement is reason-
reports values of -0.170, -0.407 and -0.663 ably good, particularly with the results from
respectively. Over the range of pile stiffness ratios Banerjee & Davies (1978) who used a more rigorous
commonly encountered (500 < E,/G* ,< 104), the numerical approach than that of Poulos (1973).
discrepancies between the two sets of exponents By noting the effect on the ground level
would only amount to about 10% difference in the deformations of varying the pile length, the critical
calculated deformation of the pile. slenderness ratio may be correlated with the stiff-
The relationships given by equations (7) are ness ratio. Expressions for the critical slenderness
generally in good agreement with results obtained ratio, similar in form to equation (1) for the
using an integral equation approach (Poulos, subgrade reaction approach, are
1971a; Banerjee 8~ Davies, 1978). As an example, (I/r,), = 2(E,/G*)2’7 (9)
Fig. 2 shows a comparison of the computed
deflexion of the pile due to an applied load H. The for homogeneous soil and
integral equation results show some dependence of
(I/r,), = 2(&/m* rJ219 (10)
the predicted pile deformations on the pile length,
even for relatively flexible piles where the pile for soil with stiffness proportional to depth. In both
length should not affect the ground level these expressions, Ep is the effective Young’s
deformations. This effect is particularly noticeable modulus of a solid circular pile of radius r,, (see
in the results taken from Poulos (1971a) and equation (3)).
probably reflects numerical inaccuracies (see In many circumstances, particularly in cohesive
Evangelista & Viggiani, 1976). or layered soil, the variation of soil stiffness with
For the case of soil with stiffness proportional to depth may fall between the two extremes of homo-
depth, the corresponding expressions are geneity and proportionality. It is helpful to
-319
combine equations (7) and (8) and also equations (9)
and (10) so that they are applicable to more general
variations of soil stiffness with depth. The first step
in accomplishing this is to define a characteristic
modulus G, which is the average value of G* over
the active length of the pile (i.e. over depths less than
1,). As a first approximation, G, may be taken as the
e = 0.60 value of G* at a depth of I,/2 (see Fig. 5). Following
equation (9) the critical slenderness ratio may then
be defined as

Figure 3 compares deflexions u obtained from the (11)


finite element analyses with the expressions given in For the extreme case of soil with stiffness propor-
252 M. F. RANDOLPH

tional to depth, the above expression reduces to of equations (2); the deformation of a free-headed
equation (10) since G, is then given by G, = m* 1,/2. pile at ground level may be readily calculated for
For general variations of shear modulus with any combination of lateral force and moment at the
depth, some iteration is necessary. For example, pile head. For fixed-headed piles, the fixing moment
suppose that the estimated variation of shear (for 0 = 0) is given by
modulus with depth is
M, = - [0.375/(~,)“~] H&/2),
G = 10+4zMN/m’ (12) leading to a deflexion u which is approximately half
where z is in metres. For a pile of radius 0.3 m and that for a free-headed pile under the same lateral
effective modulus E, = 3 x lo4 MN/m’, a first force H.
guess for the critical length might be 5 m. Taking a Generalized profiles of deformation and bending
value for Poisson’s ratio of 0.5, the characteristic moments may be drawn for piles subjected to
modulus G, may be calculated as lateral force H or moment M. These are shown, for
G, = (lo+4 x 5/2)(1+0.375) = 27.5 MN/m’ different values of pc, in Figs 6 and 7 respectively.
(13)
The maximum bending moment induced in a free-
giving a revised estimate for I, of headed pile subjected to a lateral force H may be
2 x 0.3 x (3 x 104/27.5)“’ = 4.43 m estimated from Fig. 6 as

Further iteration yields final values of W,,,, = (O.~/P,) HI, (16)


The maximum moment occurs at a depth of
I, = 4.49m, G, = 26.1 MN/m’
between 1,/4 (homogeneous soil) and 1,/3 (soil with
The concept of a characteristic modulus G, may stiffness proportional to depth).
be used to combine equations (7) and (8) to give Some comment is apposite concerning the
general expressions for the ground level defor- applicability of expressions based on elastic soil
mations of a laterally loaded pile of response for estimating the response of piles under
lateral loading. Clearly, the large strains which
u = (F$y’ [,-27H(;)- ’ +0.3+)-‘1 (14) occur in the soil close to the head of a laterally
loaded pile will reduce the relevant secant shear
modulus to a low value. It is probable that the

1
-2 -3
idealization of the soil as a material with stiffness
P, G,
0 = (F,/G,)“’ [ 0.3H 0 i +0.8(p,)“’ M 0+ proportional to depth (and thus zero stiffness at
ground level) is a better idealization than that of a
A parameter pc has been introduced, reflecting the
homogeneous soil, when considering lateral
relative homogeneity of the soil deposit. It is
loading. At large pile deformations a limiting
conveniently defined (see Fig. 5) as
pressure will be reached over the upper part of the
G*z=r,4
A__& _ G*z=l,4 pile shaft (for example, Matlock, 1970; Reese et al.,
(15)
PC = G*r=r,,z
G, 1975) and the deformation of the pile in this region
Thus pc varies from unity for homogeneous soil, may be calculated treating the pile as a simple
cantilever. Below the failed region, the expressions
down to 0.5 for soil with stiffness proportional to
given by equations (14) may be used to calculate the
depth. In the example calculation above, pc would
deformations at the transition point between failed
equal
and unfailed regions of soil. In this way, if required,
(10 +4 x 4.49/4) x 1.375/26.1 = 0.76
the complete load-deformation response of the pile
For the case of soil with stiffness proportional to may be calculated (without the aid of a computer).
depth, the expressions in equations (14) reduce to In practice, sufficiently accurate estimates of pile
those in equations (8) on substitution of the deformation and induced bending moments, under
expression for the critical length I,. For homo- working load conditions, may be made by choosing
geneous soil, the expressions in equations (14) give a suitable profile of shear modulus for the soil and
similar expressions to those in equations (7) but using equations (14) and Figs 6 and 7 directly.
with the predicted deflexions being some 8%
greater. This brings the results more in line with APPLICATION TO FIELD PILE TESTS
those calculated using integral equation analyses Tests reported by McClelland & Focht, 1956
(see Fig. 2). The power ofthe normalized curves in Figs 6 and
From equations (14) the deformation of a 7 lies in the ability to be able to estimate the
laterally loaded pile is governed by the product complete profile of bending moment or the
p, G,. This product represents the value of G* at a deflected shape of a laterally loaded pile, with
depth of one quarter of the active length of the pile. minimal computation. To illustrate application of
The general form of equations (14) is similar to that these curves, a lateral loading test on a 0.61 m dia.
RESPONSE OF FLEXIBLE PILES TO LATERAL LOADING 253

l/7
“roG, Ed
7-q MIHIc
( )

0 0.1 0.2 03 o-4 o-5 0.6 0 005 o-1 0.1 5 o-2 0.25
1 1

(a) (b)
Fig. 6. Generalized curves giving (a) deflected pile shape and (b) bending moment profile for lateral force loading.

317

-0 2 0 0.2 04 0.6 f?.A 0 0.2 04 06 0.8 1

= o-75

“‘+

s”
0.6 -

0.8 -

l-
(4 (b)
Fig. 7. Generalized curves giving (a) deflected pile shape and (b) bending moment profile for moment loading

pipe pile, reported by McClelland & Focht (1956), The test configuration was such that a negative
will be studied. bending moment of 814kNm was also applied at
The bending rigidity of the test pile was the point of application of the lateral load, and
465 MN/m* giving an effective Young’s modulus of some of the jack load was taken by the part of the
E, = 68 420 MN/m’. The pile was driven into soft, test pile extending above this level. The reported
normally consolidated deltaic clay in the Gulf of profile of bending moment down the upper part of
Mexico, to a penetration of 23m. McClelland & the pile is consistent with a lateral load of about
Focht report shear strengths of the clay increasing 300 kN being taken by the embedded length of pile
linearly with depth at a rate of approximately (this figure is slightly greater than that of 267 kN
5 kN/m’ per m. Profiles of measured bending quoted by McClelland & Focht).
moment and deflected shape of the pile (the latter The rapid increase in the strength of the soil with
obtained by integrating the bending moments depth leads logically to the assumption that the
twice) are given for a jacking load of 356 kN applied stiffness of the soil may be taken as proportional to
at a distance of 1%3m above the ground surface. depth. This assumption is consistent with the
254 M. F. RANDOLPH

=++++ -2- M: kN m
-800 -600 3j3200 200 400 600 40
I 1 ,

Theoret!cal E
x-%-x Experimental ij

(4 UN
Fig. 8. Comparison with pile test results (McClelland & Focht, 1956): (a) bending moment distribution;
(b) deflected pile shape

normally consolidated state of the clay--the stiff- good, although the maximum measured bending
ness being proportional to the effective stress level. moment of 470 kNm is underestimated by the
Back-analysis of the measured pile deflexion of theoretical curve by 20%. This appears to be due to
30mm at the jacking point leads to a variation of overestimating the stiffness of the top metre or so of
shear modulus in the clay of G = 0.8~ MN/m’, with soil, since the measured bending moment distri-
z in metres, taking a Poisson’s ratio of 0.3. This bution remains almost linear (indicating little
value of shear modulus gives a critical length for the shedding of load to the soil) to a depth of 2m.
pile of 9.5 m and a characteristic soil stiffness of
G, = 464 MN/m*. Tests reported by Gill & Demars, 1970
The profile of bending moment down the pile, A particularly useful set of lateral load tests on
and the deflected shape of the pile may be ob- piles has been described by Gill & Demars (1970).
tained from Figs 6 and 7, superimposing the Four piles of varying size and bending rigidity (see
effect of a lateral load of 300 kN with that of an Table 1) were tested at El Centro Naval Air Station,
applied bending moment (at ground level) of California, in a deposit of silty clay with a
- 814 + 300 x 1.83 = - 265 kNm. Figure 8 shows a desiccated crust extending to a depth of about 1 m.
comparison of the calculated profiles with those Tests were conducted on all four piles in a dry area
given by McClelland & Focht (1956). The agree- of the site, and also in an area which had been
ment between the two sets of curves is reasonably flooded and left for several days to allow the

Table 1. Details of laterally loaded pile tests (Gill & Demars, 1970)

Pile stiffness H/u at ground level: kN/mm


Lateral loading
Pile details parameter Dry area Flooded area

I
E,: Calculated Calculated
MN/m’ Measured* (P, = 1) Measured* (P, = 0.5)

0,057 78280 1.68 1.60 0.75 0.80


0,109 44830 3.42 3.60 1.58 1.80
0.162 44550 5.49 6.51 3.31 3.29
0.203 36290 8.09 8.73 4.38 4.37

* Measured pile stiffnesses are calculated at a ground-line deflexion of O.lr, except for the case of the largest pile in the dry
part of the site where the value at u = 11mm (when the test was stopped) has been taken.
RESPONSE OF FLEXIBLE PILES TO LATERAL LOADING 255

c,,: kN/m’

3-

3.5 - Measured H/u: kN/m

Fig. 10. Measured and predicted pile stiffnesses (Gill &


Demars, 1970)
4-

the critical length of the pile in question. For


4.5 - example, the second pile (radius 0.109 m, effective
Young’s modulus 44830 MN/m’) may be esti-
mated to have a critical length of, say,
5- 20r, = 2.18 m. At lJ2, c, = 60 kNm/m2 (from Fig.
9) whence G = 50 x 60 = 3000 kN/m’. Taking
v = 0.2 for the soil, G, = 3 x 1.15 = 3.45 MN/m’.
5,5-
This value of G, would indicate a critical length of
Fig. 9. Interpreted profiles of shear strength c, at El 3.26 m (from equation (11)). Further iteration leads
Centro (Gill & Demars, 1970) to values of I, = 3,22m, G, = 3.63 MN/m2 (see
Table 1). Corresponding values for the other three
piles are also shown in Table 1.
effective stresses to come to equilibrium. Thus a From the reported loaddeflexion curves, values
total of eight pile tests are reported for the site. of the stiffness H/u at ground level have been
Interpreted profiles of shear strength for the silty calculated at a ground level deflexion of O.lr,
clay are shown in Fig. 9. The effect of flooding an (except for the test on the largest pile in the dry part
area of the site has been to reduce the strength of the of the site, where the value at the end of the test-
soil in the desiccated crust. In order to analyse the u = 11 mm = O.O54r,-has been taken). The
pile tests, values of the characteristic modulus G, calculated and measured values of stiffness are
and of the parameter p, must be chosen. The latter generally in good agreement (see Fig. lo), with the
parameter reflects the relative homogeneity of the maximum discrepancy being about 20%
soil over the upper half of the active length of pile. It This example of the application of the expres-
seems reasonable to adopt a value of p, = 0.5 sions given in the section on Single Flexible Piles
(corresponding to soil with stiffness proportional to demonstrates the ease with which adequate esti-
depth) for the flooded part of the site, and a value of mates of the stiffness of laterally loaded piles may
p, = 1 (constant soil stiffness with depth) for the dry be made, even for relatively non-uniform soil
area. Estimating, in the first instance, a stiffness for conditions. The ability to allow for particular site
the soil of G - 5Oc,, values of G, may be chosen, conditions, such as flooding, merely by altering one
after some iteration, which correspond to the parameter (reducing p, from 1 to 05) illustrates the
measured shear strength at a depth of 1,/2 where I, is versatility of the analytical approach.
256 M. F. RANDOLPH

INTERACTION BETWEEN PILES direction of loading (see inset to Fig. 11).


The extension of solutions for the lateral A number of features of the expression above are
response of single piles to deal with groups of worthy of note. Firstly, the interaction for piles on a
closely spaced piles is possible by the use of line normal to the direction of loading (II/ = 90”) at
interaction factors (Poulos, 1971b). The interaction a given spacing is half that for piles along the line of
factor is defined as the fractional increase in loading (+ = 0”). Secondly, the interaction for piles
deformation of one pile due to the presence of a in a soil with stiffness proportional to depth
similarly loaded neighbouring pile. If k represents (p, = 0.5) is half that for piles in homogeneous soil
the stiffness (load divided by deformation) of a (p, = 1). Thirdly, the interaction factors are
single isolated pile, then the deformation of the ith inversely proportional to the pile spacing. While, in
pile in a group of n piles may be written (Poulos, practice, pile stiffnesses and spacings are rarely such
1971b) that s/r,, < 2(E,/G,)‘/‘, occasionally, for very closely
1 n spaced piles, the expression in equation (18) may
6i = ijF,%j3 (17) grossly overestimate the interaction. In applying
the expression (e.g. Randolph, 1980), it is found
where aij is the interaction factor between the ith more realistic to replace tlpf by l-(401,,)- ’ when
and jth pile (aii being equal to unity), and Pj is the the value of c(pF given by equation (17) is greater
load on the jth pile. than 0.5. This modification ensures that, as the pile
For laterally loaded piles, there are a number of spacing s tends to zero, the calculated interaction
different types of interaction factors. Using factor tends to unity.
Poulos’s (1971b) notation these are Figure 11 shows a comparison of the expression
in equation (18) with interaction factors for piles in
(4 C+ giving the increase in deflexion of free-
headed piles subjected to lateral load homogeneous soil published by Poulos (197 lb) (no
corresponding values for soil with stiffness
(b) apM: giving the increase in deflexion of free-
headed piles subjected to moment loading proportional to depth have been found in the
literature). The agreement is generally good,
(4 tlgH: giving the increase in rotation of free-
headed piles subjected to lateral load although interaction factors for piles in line with the
applied load ($ = 0”) at close spacings tend to be
(Qf = %H from the reciprocal theorem)
overpredicted compared with Poulos’s values. One
(4 crenr: giving the increase in rotation of free- possible explanation of this is the idealization of the
headed piles subjected to moment loading
pile as a thin strip for the integral equation analysis
(e) c+: giving the increase in deflexion of fixed- of Poulos (1971). This has the effect of increasing the
headed piles (where the pile heads are
restrained against rotation) subjected to lateral amount of soil between piles compared with, say,
load circular piles, thus leading to lower interaction
factors at close spacings. In practice, the discre-
In practice, the last of these interaction factors is the pancy between the two sets of results will rarely be
most relevant, since the majority of pile groups are significant since few foundations have values of
capped with a pile cap of sufficient stiffness to s/[r,(E,/G,)“‘] less than about 2.
prevent significant rotation of the pile heads. Poulos (1971b) notes that interaction factors for
From finite element analyses of laterally loaded, fixed-headed piles are greater than for free-headed
fixed-headed, piles, Randolph (1977) has shown piles. Reasonably accurate estimates of the factor
that the pattern of lateral movement at the soil CQ, may be obtained by replacing the multiplier 0.6
surface, around a laterally loaded pile, can be in equation (18) by 0.5 to give
related directly to values ofa,, published by Poulos
(197lb); that is, interaction factors may be deduced
a,,10.5p,(E,/G,)‘!‘~(l +cos’ I,+) (19)
by noting the magnitude of soil movement, as a
proportion of the pile movement, at any particular
Again, where the value of Q, given by this expres-
location. This finding enables interaction factors to
sion exceeds 0.5, a corrected value of l -(4a,,)- ’
be computed for a variety of pile stiffness ratios and
should be taken.
for non-homogeneous (i.e. pC< 1) as well as homo-
Values of the other three interaction factors (a,,,
geneous soil. It is found that normalization by the
aBH and cl& are considerably smaller than apH at
stiffness ratio E,/G, allows the interaction factors
realistic pile spacings. Good estimates of them may
ctpF.to be expressed approximately in the form
be made by
2
apF N 0.6pc(Ep/GJ1/'? ( 1+ cos’ I//) (18) %M = %f= fx pH (20)

and
where s is the pile spacing and $ is the angle
between the line joining the pile centres and the (21)
RESPONSE OF FLEXIBLE PILES TO LATERAL LOADING 257

I I 1 I I I I 1 J
2 4 6 a 10 12 14 16 ia 20 22

Fig. 11. Interaction factors for fixed-headed piles (aoF: Poulos, 1971b)

In summary, expressions have been presented in giving G, = 0.289 MN/m2. The value of I, is slightly
this section which enable interaction factors for greater than the embedded length of the pile,
laterally loaded piles to be calculated. By although by insufficient a margin to cause signifi-
normalizing the pile spacing s by the product cant error in the use of equation (8) (see Randolph,
r,(E,IG,)“‘, a single expression serves to give 1977). The low value of m* r,corresponding to a
interaction factors for a wide range of pile stiffness value of m of 2.2 MN/m3 for v = 0.2-is due to the
ratios (not forgetting the restriction to flexible large displacement of the pile at the load level of
piles). A particular advantage of the equations 0.03 kN, leading to significant plastic straining in
presented is the ease with which they may be the soil.
incorporated into computer programs for the Tests were conducted on groups of 2 piles and of
analysis of general pile groups (e.g. Randolph, 3 piles in a line along the direction of loading. The
1980). pile groups were loaded at the same level of 125 mm
As an example of how such interaction factors above the sand surface as the single pile, by means
may be applied in practice, lateral load tests of a pile cap designed to allow free rotation of the
conducted on model pile groups, reported by pile heads (Fig. 12(a)). The interaction factors for a
Williams (1979), will be analysed. Tests were con- given pile spacing s may be calculated from
ducted on tubular aluminium piles of external equations (19) and (20) as
radius 3.97mm and wall thickness 0.71 mm
c+, = 0.5 x 0.5 x (37 500/0.289)“‘~ 2$ = 2.69?
(whence E, = 37 500 MN/m’) embedded to a depth
of 200mm in dense medium-grained to coarse-
(22)
grained sand. The piles were loaded at a height of
125mm above the sand surface. From tests on a for s > 5.38r, (i.e. tlpH< 0.5), or
single pile, the deflexion at the sand surface for a
lateral load of 0.03 kN was 5.7 mm. Assuming a soil %H = 1-0.093s (23)
r.
stiffness increasing proportionally with depth,
equation (8) may be used to back-analyse a value for s< 5.38r,. Similarly, apM is given by
for m*r, of 0.01 MN/m’. The critical length may
then be calculated as apM = apH 2 = 7.24f:r or (l-0.093;)

I, = 2ro(Ep/m* r,,)2’9 = 230 mm ii4)


258 M. F. RANDOLPH

Free rotation efficiency with pile spacing for the 2-pile and 3-pile
groups. The theoretical curves are compared with
data from Williams (1979). There is reasonably
good agreement for pile spacings of up to s = 12r,;
thereafter, the theoretical group efficiencies are
lower than those measured, which rapidly
approach unity. One of the reasons for this under-
prediction is that, in practice, the pile cap through
which the piles were loaded provided some
restraint against rotation of the pile heads.
Williams (1979) estimates that the restraint may
have amounted to 20% of the moment at the sand
(a) surface. This would account for measured values of
efficiency greater than unity (Fig. 12(b)). If the
12 calculation shown in equation (25) is modified to
r allow for this moment restraint (by reducing the
second term on the right-hand side by 20%) then a
l-
group efficiency of 0.90 is calculated. In general, the
two curves in Fig. 12(b) would be raised by
0,6- approximately lo%, giving improved agreement
3 piles (theoretical) with the measured data.
7\ Two tests are also reported by Williams (1979) on
0,6 - groups of 2 piles and of 3 piles in line, at a spacing of
s = 8r,, loaded perpendicular to the line of the
Experimental data +2 piles group. Measured efficiencies of 0.85 (3 piles) and of
0.4 Williams (1979) x3 piles
0.95 (2 piles) compare with efficiencies calculated by
t the method above of 0.86 and 0.92 respectively.
0.2
CONCLUSIONS
t Algebraic expressions have been presented which
I
0 4 a 12 16 207 allow the behaviour of flexible piles under lateral
S/r0 loading to be calculated, in terms of fundamental
(4
soil properties. The expressions are based on the
results of finite element studies of the response of a
Fig. 12. Model pile tests from Williams (1979): (a) test
laterally loaded cylindrical pile embedded in elastic
arrangement (3 piles); (b) group efficiencies
soil with stiffness varying linearly with depth.
The concept of a characteristic soil stiffness has
been introduced, representing the average soil
Thus for a pile spacing of s = 8r, the interaction stiffness over the active length of the pile. This has
factors are apH = 0.336, Q,, = 0.113, and the
enabled the ground level deformations of the pile to
deflexion of the piles at a load per pile of 0.03 kN is be calculated from a single pair of expressions,
0.03 covering the range of soil conditions from constant
u = 0.54 (1+0.336) soil stiffness to soil stiffness proportional to depth.
0.01 x 000397
The critical length of the pile, beyond which it
+060 x 0.125 x 0.03 37 500 -5’g behaves as infinitely long, may be calculated
(1+0.113)
0.01 x 000397~ ( 0.01 > directly from the ratio of the equivalent modulus of
the pile E, to the characteristic soil stiffness G,.
= 7.05 mm (25)
Charts have been presented showing the deformed
Thus the deflexion of the pile group is a factor of shape of the pile, and bending moment distribution
1.23 greater than that of a single pile at the same down the pile, for ground level loading by a lateral
load per pile. Expressed as an efficiency, the force H or moment M.
efficiency of the 2-pile group is q = l/1.23 = 0.81. The theoretical solutions have been used to
A similar calculation may be performed for three analyse field load tests on single piles, reported in
piles in line, except in that case each pile interacts the literature. The versatility of the solutions pre-
with two others and, since the deflexion of the sented, in catering with piles of different size and
central pile will be different from that of the end stiffness, embedded in soil either with or without a
piles, an average efficiency must be calculated. stiff, desiccated crust, has been demonstrated.
Figure 12(b) shows the variation of the group The approach of estimating interaction factors
RESPONSE OF FLEXIBLE PILES TO LATERAL LOADING 259

between neighbouring piles, from the deformations Kuhlemeyer, R. L. (1979a). Static and dynamic laterally
in the soil around a single pile, has led to simple loaded floating piles. J. Geotech. Engng Div., Am. Sot.
expressions for interaction factors for fixed-headed Civ. Engrs 105, No. GT2, 289-304.
Kuhlemeyer, R. L. (1979b). Bending element for circular
and for free-headed piles. These expressions com-
beams and piles. J. Geotech. Engng Div., Am. Sot. Ciu.
pare well with interaction factors deduced from
Engrs 105, No. GT2, 325-330.
integral equation analyses by Poulos (1971b). McClelland, B. & Focht, J. A. (1956). Soil modulus for
Values of group efficiency calculated using the laterally loaded piles. J. Soil Mech. Fdn Div., Am. Sot.
interaction factors give reasonable agreement with Cio. Engrs 82, No. SM4.
values measured in laboratory model tests on pile Matlock, H. (1970). Correlations for design of laterally
groups containing 2 and 3 piles. loaded piles in soft clay. Proc. 2nd Ofihore Tech. Cotrf,
In conclusion, the aim of this Paper has been to Houston, Texas 1, 5777594.
present solutions which are accurate, within the Matlock, H. and Reese, L. C. (1960). Generalised solutions
for laterally loaded piles. J. Soil Mech. Fdn Div., Am.
framework of elastic soil response, and yet simple
Sot. Ciu. Engrs 86, No. SM5, 63391.
enough to be of practical use in estimating the
Matlock, H. & Ripperger, E. A. (1958). Measurements of
response of laterally loaded piles and pile groups. soil pressure on a laterally loaded pile. Proc. Am. Sot.
The soil stiffness has been characterized by an Test. Mater. 58, 1245-1259.
elastic modulus, rather than by a coefficient of Poulos, H. G. (1971a). Behaviour of laterally loaded piles:
subgrade reaction, in order to avoid the problems isingle piles. J. Soil Mech. Fdn Div., Am. Sot. Cio.
which stem from the effect of pile size and stiffness Engrs 97, No. SM5, 711-731.
on appropriate choice of the latter. Finally, as Poulos, H. G. (1971b). Behaviour of laterally loaded piles:
Kuhlemeyer (1979a) has pointed out, the II-pile groups. J. Soil Mech. Fdn Dir., Am. Sot. Ciu.
tremendous simplification which follows from cast- Engrs 97, SM5, 733--751.
Poulos, H. G. (1972). Behaviour of laterally loaded piles:
ing the solutions in a form which is independent of
III&socketed piles. J. Soil Mech. Fdn Div., Am. Sot.
the embedded length of the pile, must be empha- Ciu Engrs 98, No. SM4, 341-360.
sized. Only in comparatively rare cases will the Poulos. H. G. (1973). Load-deflection prediction for
length of the pile be a relevant parameter in laterally loaded piles. Aust. Geomech. J.G3,No. 1, l-8.
calculating its response under lateral loading. Randolph, M. F. (1977). A theoretical study of the
performance ofpiles. PhD thesis, University of Cam-
REFERENCES bridge.
Banejee, P. K. & Davies, T. G. (1978). The behaviour of Randolph, M. F. (1980). PIGLET a computer program for
axially and laterally loaded single piles embedded in the analysis and design of pile groups under general
loading conditions. Cambridge University Research
nonhomogeneous soils. Geotechnique 28, No. 3,
309-326. Report, CUED/D-Soils TR91.
Carayannacou-Trezos, S. (1977). Comportement des pieux Reese, L. C., Cox, W. R. & Koop, R. D. (1975). Field
sollicites horizontalemenf. Dr Ing thesis, Universite testing and analysis of laterally loaded piles in stiff
clay. Proc. 7th Qffshore Tech. Co@, Houston, Texas 2,
Paris VI.
Evangelista, A. & Viggiani, C. (1976). Accuracy of nu- 473483.
Reese, L. C. & Matlock, H. (1956). Non-dimensional
merical solutions for laterally loaded piles in elastic
solutions for laterally loaded piles with soil modulus
half-space. Proc. 2nd Int. Conf: Num. Meth. Geomech.,
proportional to depth. Proc. 8th Texas Conf Soil
Blacksburg, Virginia 3, 1367-I 370.
Gill, H. L. & Demars, K. R. (1970). Displacement of Mech. Fdn Engng, 141.
Williams. D. J. (1979). The behaviour of modelpiles in
laterally loaded structures in nonlinearly responsive
dense’sand. PhD thesis, University of Cambridge.
soil. Technical report R670. Port Hueneme,
Wilson, E. L. (1965). Structural analysis of axisymmetric
California: Naval Civil Engineering Laboratory.
solids. J. Am. Inst. Aer. Astr. 3, 226992274.
Hetenyi, M. (1946). Beams on elastic foundations. Ann
Arbor: University of Michigan Press.

You might also like