Acta Geotechnica (2014) 9:469–484

DOI 10.1007/s11440-013-0240-9

RESEARCH PAPER

A finite element–based approach for predictions of rigid pile

group stiffness efficiency in clays
Brian B. Sheil • Bryan A. McCabe

Received: 17 February 2012 / Accepted: 4 March 2013 / Published online: 17 August 2013
Ó Springer-Verlag Berlin Heidelberg 2013

Abstract This paper presents a finite element parametric pile design. When considering the serviceability of groups
study of several variables that affect the stiffness efficiency of closely spaced piles, the stiffness efficiency (gg) term is
of rigidly capped pile groups with a view to developing a a convenient gauge of the extent of pile-to-pile interaction
solution for preliminary design purposes. Previous empir- within the group under working conditions. The value of gg
ical solutions from linear elastic work had identified a is defined as follows by Butterfield and Douglas [1]:
significant dependence of stiffness efficiency on pile group 
gg ¼ k g ½ N k s  ð1Þ
size and group spacing, and in this study, the effect of the
pile length-to-diameter ratio, the compressibility of a stiff where kg is the stiffness of the pile group, ks is the stiffness
bearing stratum beneath the pile group and the depth below of an equivalent single pile, and N is the number of piles in
ground level to the stiff bearing stratum are also consid- the group.
ered. Pile groups in a soft clay/silt are modelled using Numerous rigorous methods have been employed in the
PLAXIS 3D Foundation in conjunction with a soil model literature that consider the pile–soil system as one com-
that captures the stress dependency of soil stiffness. The posite continuum including the boundary element method
trends from the soft soil study have been formulated into a [2], the finite element method [3] and methods based on
set of equations which can be used to predict the stiffness mobilisable strength design principles [4]. These methods
efficiency of pile groups. This new approach captures more can become both computationally intensive and time-con-
variables than previous simpler empirical prediction suming if nonlinear pile–soil–pile interaction and large
methods and performs better when applied to a database of groups with non-standard geometries are also to be mod-
29 published pile group case histories. elled. In contrast, appropriate empirical approaches can
allow the designer to assess the effects of various variables
Keywords Numerical modelling  Pile group  and also to obtain a preliminary estimate of gg quite quickly.
Pile interaction  Stiffness efficiency Early approaches developed for the prediction of gg,
however, incorporated simplifying assumptions such as a
linear elastic (LE) soil medium [1, 5, 6], which leads to an
1 Introduction overestimate of the extent of pile interaction and can
considerably underestimate the value of gg [7]. Fleming
The escalating demand for large-scale structures has placed et al. [5] proposed the following approach based on linear
increased demands upon geotechnical practitioners in that elasticity:
there is now as much design focus on the serviceability gg  N e ð2Þ
limit state of foundations as on the ultimate limit state in
where the exponent, e, typically ranged between 0.5 and
0.6.
B. B. Sheil  B. A. McCabe (&) Castelli and Maugeri [8] later improved upon the sim-
College of Engineering and Informatics,
National University of Ireland, Galway, Ireland plified approach proposed by Fleming et al. [5] by rec-
e-mail: [Link]@[Link] ognising the importance of considering soil nonlinearity

123
470 Acta Geotechnica (2014) 9:469–484

and thus used hyperbolic load transfer functions to model a Table 1 Typical properties of the Belfast sleech
more realistic nonlinear pile–soil–pile interaction accord- Clay fraction (%) 20 ± 10
ing to: Fines content (%) 90 ± 5
h ie
Water content (%) 60 ± 10
gg ¼ D=Dg ð3Þ
Plasticity index (%) 35 ± 5
where Dg = equivalent diameter of the plan area of the pile Organic content (%) 11 ± 1
group; D = diameter of the pile (or equivalent diameter for Peak vane strength (kPa) 22 ± 2
a square pile), and an exponent of e = 0.15 was derived Overconsolidation ratio (OCR) 1.1–2.0
from a limited database of case histories. Friction angle (°) 33 ± 1
More recently, McCabe and Lehane [7] developed Eq.
(4) as the best fit to gg values derived from nine published
case histories with single pile and pile group data:
  0:66
Dg D
gg ¼ ð4Þ
N
Further field data are required to validate the form of
Eq. (4).
However, there are important variables other than pile
group size and pile spacing that influence stiffness effi-
ciency, and there remains a need for a simple approach
capable of incorporating additional variables without sac-
rificing the stress dependency of soil stiffness from the
process. In this study, additional normalised variables Fig. 1 Load test arrangement
considered include the effect of the pile length-to-diameter
ratio (L/D), the relative depth below ground level of a 2.2 Pile load test details
stiffer layer beneath the pile base (h/L) and the relative
stiffnesses of the lower and upper layers (E2/E1). Expres- Precast concrete piles, B = 250 mm wide, were driven to a
sions for gg are derived from curve fitting to the output depth of 6.0 m at the Belfast site. The piles were installed
from FE analyses of pile groups in soft soil. PLAXIS 3-D using a 5-tonne hydraulic hammer with a drop height of
Foundation version 2.2 [9], was used in conjunction with 0.45 m. The group piles were arranged with a centre pile
the commercially available hardening soil (HS) model, and surrounded by four corner piles at a spacing-to-width (s/
data from McCabe and Lehane [7] were used to provide B) ratio of 3.0. Purpose-built load cells were positioned at
initial calibration for the model. The validity of this new the heads of all piles. Maintained-load compression testing
approach is appraised against a database of published on both the single and 5-pile group (with a threshold creep
friction and end-bearing pile group case histories in clays. rate of 0.024 mm/h) was carried out using Kentledge dead
weight as shown in Fig. 1. The rigid steel pile cap con-
sisted of two orthogonal layers of six steel I-sections
2 Belfast pile test programme positioned between 1.8-m square steel plates. The piles
were loaded to failure almost 3 months after installation in
2.1 General site description increments of 5–15 % of the estimated ultimate load for a
single pile, by which time over 90 % of the excess pore
Single pile and pile group load test data from the pro- pressures generated during installation had dissipated [7].
gramme reported by McCabe and Lehane [7] at Belfast,
Northern Ireland, are modelled in this paper. The Belfast
soil profile consists of a layer of made ground which 3 Soil model for parametric study
extends to a depth of *1.0 m, a layer of silty sand from 1.0
to 1.7 m and a lightly overconsolidated soft estuarine silt 3.1 Belfast general soil parameters
(locally known as sleech) to a depth of 8.5 m. The sum-
mary properties of the sleech are given in Table 1. A The basic soil properties used for the constitutive model
stratum of medium dense sand exists at 8.5 m below were derived from laboratory tests reported by Lehane [10]
ground level, and the water table was found at approx. and McCabe and Lehane [7] and are included in Table 2.
1.4 m below ground level (with a small tidal variation). Overconsolidation ratios (OCR) of 1.2 and 2 were selected

123
Acta Geotechnica (2014) 9:469–484 471

Table 2 Belfast soil parameters used in HS model 140

Fill Silty sand Sleech Lower sand McCabe [11]; T16

120
PLAXIS HS
Depth (m) 0.0–1.0 1.0–1.7 1.7–8.5 8.5–12.0
csat (kN/m3) 18.5 17.5 16.5 20 100
cunsat (kN/m3) 18.5 17.5 16 20
OCR 1.0 2 1.2 2 80

Gsec/p'0
POP (kPa) 15 0 0 0
u0 (°) 33 33 33 40 60

W (°) 0 0 0 0
K0 0.46 0.66 0.5 0.46 40

m 0.5 0.5 1.0 0.5

20
0ref
Eoed (MPa)a 25 12 6.5 75
0ref a
E50 (MPa) 25 12 6.5 75
0
0ref 75 36 19.5 225 0.1 1 10
Eur (MPa)
pref (kPa) 30 30 30 30 Axial Strain (%)
c0 (kPa) 1.0 1.0 1.0 1.0
Fig. 2 HS predictions of triaxial stiffness
ky (m/day) 1.0 2.6 9 10-5 2.6 9 10-5 8.64
kx (m/day) 1.0 2.6 9 10-5 5.2 9 10-5 8.64
Rinter 0.9 0.9 0.55 1.0 where E = Young’s modulus, pref = reference stress for
a ref ref
stiffness Eref and m = an exponent for stress-level depen-
E50 set equal to Eoed in the absence of an appropriate reference
dency of stiffness. According to Schanz et al. [12], a value
of m = 1 should be selected to simulate a stress depen-
dency of soil stiffness typical in soft clays and has thus
as the average values encountered in the sleech and silty been selected for the sleech. A value of m = 0.5 for the fill
sand, respectively [11]. A pre-overburden pressure (POP) material has been documented as being typical for hard
of 15 kPa was chosen for the fill material to represent a soils [12], while an intermediate value of 0.75 was chosen
total drop in vertical effective stress due to fluctuating for the silty sand.
ground levels and erosion of material. A high friction A value of zero was chosen as a realistic value for the
angle, u0 , of 33° was selected for the fill material, the silty dilatancy angle of the soil, w [9]. The value of K0 was
0
sand and the sleech based on the triaxial tests documented obtained using the relationship K0 ¼ ð1  sin u0 ÞOCRsin u
by McCabe [11]. [13]. The Young’s modulus at half the maximum deviator
ref
stress, E50 , for the sleech was selected by calibrating pre-
3.2 Hardening soil model and parameters dicted stress–strain curves in triaxial compression using the
‘Soil Test’ facility in PLAXIS to the measured curves
The hardening soil (HS) model was chosen to model the documented by McCabe [11] where pref was set equal to
behaviour of the fill and sleech in PLAXIS 3-D Founda- the initial mean effective stress at the beginning of
tion. The main advantage of the HS model over an elastic undrained shearing, p00 = 30 kPa (Fig. 2). The secant shear
perfectly plastic model is that the yield surface is not fixed stiffness values, Gsec, in Fig. 2 (calculated from Esec with
in principal state but instead can expand due to plastic m = 0.5 for undrained conditions) have been normalised by
straining [9]. The HS model is an improvement on the p00 . The stress history of the sleech was modelled as normal
Duncan–Chang hyperbolic model in that the theory of consolidation (K0 = 0.5) with subsequent swelling to the
plasticity is used as opposed to the theory of elasticity. It current in situ overconsolidation ratio of 1.2 (which was
also includes soil dilatancy and a yield cap although creep representative of the average value in the sleech encoun-
behaviour is not considered. tered by the piles); a reasonable match can be seen in
One of the main features of the HS model is its ability to Fig. 3.
capture the stress dependency of soil stiffness. The ref
A value of E50 for the fill material, corresponding to a
parameter ‘m’ controls the relationship between soil stiff-
reference pressure of 30 kPa, was chosen based on the soil
ness, E, and the corresponding confining stresses, p, as
stiffness employed by McCabe and Lehane [3] in OASYS
follows: ref ref ref
 0  SAFE. The default setting of Eur ¼ 3  E50 , where Eur is
E c cos u0 þ p sin u0 m the unloading–reloading soil stiffness, was maintained for
¼ 0 ð5Þ
Eref c cos u0 þ pref sin u0 the fill, silty sand and sleech material, while a value of

123
472 Acta Geotechnica (2014) 9:469–484

40 corresponding to Rinter values of 0.9 and 0.55. A non-por-

ous linear elastic model was chosen to represent the con-
crete pile material with a Young’s modulus, E = 30 GPa
and Poisson’s ratio, m = 0.15.
30
Deviator stress, q (kPa)

Since the input parameters in the present soil model

have been defined in terms of effective stress, it is neces-
sary to check that the present soil model accurately predicts
20 the undrained shear strength profile at the Kinnegar test site
which was measured by in situ shear vane testing. Cor-
M=1.33 rection factors for undrained strength reported by Bjerrum
[18] have been applied to the measured shear vane
10
strengths documented by McCabe [11] in Fig. 4. It can be
McCabe [11]; T16
PLAXIS HS
seen that predictions of the undrained shear profile deter-
PLAXIS HS -Failure envelope mined by the HS model, although illustrating a slightly
0 different gradient with depth, show reasonable agreement
0 10 20 30 40 with vane strengths reported by McCabe [11].
Mean effective stress, p' (kPa)

Fig. 3 HS predictions of triaxial behaviour

4 Finite element modelling
1 kPa for the cohesion of the fill and sleech was used for
4.1 Stages of analysis: single pile
numerical stability (e.g. [14]). It is recognised by the
authors that the permeabilities, k, of the sleech reduce with
The stages used in the analysis of a single pile are defined
increasing stress levels so average values at in situ stress
as follows:
levels are presented in Table 2.
In order to simulate the reduction in strength of the soil (i) Inclusion of interface elements in the soil model to
at the pile–soil interface arising from pile installation, allow for pile–soil slip.
PLAXIS allows for the input of the interface strength (ii) Initial stress generation by the K0 procedure, a special
reduction factor, Rinter, such that: calculation method available in PLAXIS.
tan u0inter ¼ Rinter tan u0soil ð6Þ
Einter ¼ R2inter Esoil ð7Þ
Although the present FE analyses are limited to ‘wished-in- Undrained Shear Strength (kPa)
0 10 20 30 40 50
place’ pile installation, the reduction in soil strength and 0
stiffness at the interface arising from the installation of
PLAXIS HS
each individual pile is considered in the interface strength -1
Vane tests [11]
reduction factor, Rinter. It is acknowledged that while the
-2
potential additional effects due to the driving of neigh-
bouring piles cannot be captured, in the case of the -3
aforementioned 5-pile group, radial total stress measure-
ments during and after group installation indicated that -4
Depth (m)

equalised values at the centre pile shaft were only mar-

-5
ginally lower than those on a single pile [7].
The database of interface strength reduction factors -6
documented by Potyondy [15] and Tiwari et al. [16] guided
the selection of an appropriate value of Rinter for the fill -7

material (chosen as Rinter = 0.9). The selection of appro-

-8
priate values of Rinter for the silty sand and sleech was
aided by site-specific operational interface friction angles, -9
d, documented by Strick van Linschoten [17]. These were
measured in ring shear interface tests with a concrete -10

interface. Values of d equal to 31° and 18.5° ± 3° were Fig. 4 Comparison between corrected vane strengths and predicted
reported for the silty sand and sleech, respectively, undrained shear strength profiles

123
Acta Geotechnica (2014) 9:469–484 473

(iii) Installation of the concrete pile reflected by changing E2 is the stiffness of the lower layer, and the boundary
appropriate elements to a linear elastic material with a between them occurs at a depth h below ground level,
Young’s modulus of 30 GPa and a Poisson’s ratio, m, where h is greater than or equal to the pile length L. The
of 0.15. depth below ground level to the bottom mesh boundary, H,
(iv) Pile loading by placing a compressive uniform was chosen as 3L so that the bottom mesh boundary had no
distributed load along the top surface of the pile effect on the output. Likewise, the lateral boundaries of the
material. FE model for each analysis were located at a distance such
(v) Recording of the pile head displacement versus pile that no influence was recorded on output results.
head load relative to the start of loading. Other features of the model are shown in Fig. 6 (for a
16-pile group). Symmetry was exploited to reduce the
number of elements used in the mesh and associated
4.2 Stages of analysis: two-pile interaction factors
computational time. In all analyses, the mesh was refined in
zones of high stresses near the piles. Coarse, medium and
The interaction factor (a) is defined by Poulos [19] as:
fine meshes were used to check mesh convergence for all
Sij analyses, but fine meshes were ultimately used. In the
a¼ ð8Þ
Sii present study, only free-standing pile groups are modelled,
where Sij is the additional settlement of pile i due to a so the pile cap does not come into contact with the ground
nearby loaded pile j and Sii is the settlement of pile i under surface.
its own load. For the calculation of interaction factors in
this paper, the settlement of a single pile under its own load
was first determined by PLAXIS analyses. A non-loaded 5 Validation of soil model
(i.e., no external pile head load applied) pile was then
incorporated within the settlement field of a loaded pile at a 5.1 Load displacement results
number of different pile spacing-to-diameter ratios (s/D).
The ‘interactive’ settlement of the non-loaded pile due to In the analyses, the square piles of width B were modelled
the presence of the neighbouring loaded pile was then as circular piles with an equivalent diameter, Deq, of
recorded for each value of s/D and the corresponding value 282 mm where Deq = 2B/p0.5. A pile-to-pile spacing of s/
of a calculated. Deq = 2.8 was adopted in the analyses. PLAXIS 3-D

4.3 Stages of analysis: pile groups with rigid caps

For the case of the pile groups with rigid caps, a slightly
different procedure was followed:
(i)–(iii) Similar procedure to that described in Sect. 4.1.
(iv) Excavation of soil to a depth of 0.5 m below the
pile heads.
(v) Installation of the pile cap (modelled as a ‘floor’
in PLAXIS) along the top of the pile group.
Floors in PLAXIS are composed of 6-node
triangular plate elements. The depth of the pile
cap was arbitrarily chosen as 0.4 m with the same
properties employed for the concrete pile.
(vi) Pile group loading by placing a compressive uniform
distributed load along the top surface of the pile cap.
(vii) Recording of the pile cap displacement versus
pile head load relative to the start of loading.

4.4 Finite element model parameters

The pile/soil parameters referred to in the study are illus-

trated in Fig. 5 where E1 is the stiffness of the upper layer, Fig. 5 Illustration of pile/soil parameters

123
474 Acta Geotechnica (2014) 9:469–484

70

60

50

Load (kN)
40

30 S; McCabe and Lehane [7]

20 S; PLAXIS HS
C; McCabe & Lehane [7]
10
C; PLAXIS HS
0
0 5 10 15 20 25
Displacement (mm)

Fig. 7 Predicted and measured load–displacement responses of the

single pile (S) and centre pile (C) of the 5-pile group

safety (FOS) on single pile capacity of 1.5 and 4.0,

respectively, where the pile capacity was defined nominally
at a pile head displacement of 0.1D. Although comparisons
with the data reported by Cooke [20] and Caputo and
Viggiani [21] are only indicative (since different soil and
pile properties as well as load levels will lead to differences
in interaction factors), PLAXIS results show good agree-
Fig. 6 Finite element mesh for 16-pile group ment to the measured field data beyond a value of
s/Deq = 2.5. More importantly, the agreement between
Foundation was used to simulate the single pile field test PLAXIS predictions and the measured data at the Belfast
(S) for the partially drained loading believed to have test site show significant agreement.
occurred in the test. This was carried out by increasing the From Fig. 8, it can also be seen that predictions deter-
pile head load in a ramp over a 6-h consolidation phase (i.e. mined by the present analyses and the 2-D nonlinear FE
the duration of the single pile load test). PLAXIS then analyses documented by Jardine et al. [22] show much
incrementally increases the load directly in proportion with improved agreement to field data than the predictions
time passed. The partially drained analysis appears to determined using the PIGLET computer program [23] and
predict the initial measured stiffness of the single pile, its the approach documented by Chen et al. [24] where the soil
nonlinearity and the pile load at 15 mm displacement very is idealised as a LE medium. The method employed by
well. Chen et al. [24] differs from conventional approaches in
The load–displacement response of the centre pile of the that a more rigorous approach to consider pile–soil
5-pile group (C) has also been plotted with that simulated
by a partially drained analysis in PLAXIS 3-D where the 1
duration of the 5-pile group load test in this case was 12 h. B = McCabe & Lehane [7]
It can be seen from Fig. 7 that PLAXIS 3-D under-predicts 0.8
C = Caputo & Viggiani [22]
Interaction factor α

the stiffness of the centre pile slightly but the nonlinear

load–displacement response and load at 15 mm displace- 0.6
ment are captured well.
0.4
5.2 Interaction factors
0.2
B B C
The present (HS) model has also been employed to com- B
C C
pare predictions of two-pile interaction factors to a range of 0
field data and existing prediction methods; the results of the 0 1 2 3 4 5 6 7 8

analyses have been plotted in the traditional form of a s/D, s/Deq

against s/D or s/Deq in Fig. 8. The variation in predictions PIGLET [24] Chen et al. [25]
Cooke [21] PLAXIS HS
of a with load level determined by PLAXIS is captured by Jardine et al. [23]
the shaded region in Fig. 8 where the upper and lower
bounds represent a load level corresponding to a factor of Fig. 8 Comparison of measured and predicted interaction factors

123
Acta Geotechnica (2014) 9:469–484 475

interaction is proposed using the fictitious pile-extended

half-space model.
It can be seen from Figs. 7 and 8 that PLAXIS results
compare well to the measured field data documented by
McCabe and Lehane [7], thus affirming the selection of D
material parameters and soil model outlined in Sect. 3. The
material and geometrical parameters serve as default values to
be varied during the parametric study. Since only pile settle-
ment after full completion of pore pressure dissipation is of

Dg
interest in the subsequent analyses, fully drained analyses are
used for the remaining analyses presented in this paper.

6 Stiffness efficiency of rigid pile groups

Fig. 9 Illustration of the term Dg (adapted from [8])
6.1 Overview

As already mentioned, existing empirical approaches and study, in the interests of managing the number of analyses
charts for the prediction of rigid pile group stiffness effi- to be performed. However, the plan density of soil
ciency are not general enough to take account of various replacement with piles is considered indirectly in the
factors influencing the value of gg. Thus, PLAXIS 3-D is composite term Dg/D (as used by [7]) which captures
used to obtain predictions of gg for parameters already variations in both pile diameter and spacing, where Dg is the
considered in existing empirical approaches such as N and equivalent diameter of the plan area of the pile group and
Dg/D [5, 7, 8] in addition to the parameters L/D, E2/E1 and D is the diameter of single pile as shown in Fig. 9. This
h/L and are presented in Sects. 6.4–6.7. assumption is shown to be appropriate in Sect. 7.2.
It is recognised that nonlinear pile interaction factors
(and thus gg) vary with load level; thus, a commonly 6.3 Effect of group configuration
employed FOS of 2.5 was maintained throughout the
parametric study. This FOS was conservatively chosen In the present study, the variation in gg for various group
since pile-to-pile interaction was shown in Fig. 8 to configurations of m 9 n piles is investigated, where m is
decrease with increasing load level. There is currently little the number of piles along a row and n is the number of
guidance in the literature on how the relative pile–soil piles along a column as shown in Fig. 10 for a 16-pile
stiffness (R) should be defined for a soil with stress- group. In Fig. 11, PLAXIS predictions of gg have been
dependent stiffness. In the present study, the value of R is plotted against m/n for group sizes (N) of 4, 16 and 36 piles
defined as the ratio of the Young’s modulus of the pile (30 having L/D = 25. It can be seen that the effect of group
GPa) to the average secant stiffness of the soil in drained configuration on gg is relatively insignificant for values of
 ref 
triaxial compression E50 at a reference pressure of m/n B 10. The authors have therefore not considered the
ref
p = 30 kPa. The soil properties in the present parametric effect of group configuration further in the subsequent
study are restricted to those of a soft clay; thus, the effect of parametric analyses where a conventional square pile
the relative pile–soil stiffness is not taken into account. group formation (m/n = 1) is used henceforth, erring very
From LE analyses [6], however, a variation in the value of marginally on the side of conservatism. Larger groups are
R between 1,000 (representing a stiff soil) and ? (repre- developed by placing an extra square of piles around each
senting a very soft soil) only had an appreciable effect on previous group as shown in Fig. 12.
pile interaction for unusually slender piles (i.e. L/D C 100).
The term ‘rigid pile groups’ is used loosely in the present 6.4 Effect of group size
paper since it is realistic to assume that slight deformation
of the pile cap may occur, particularly for large group sizes. It is well established that group size has a significant
impact on pile group stiffness efficiency and has been
6.2 Effect of pile spacing considered in many existing empirical approaches. Due to
the large number of elements associated with volume piles
A commonly employed pile spacing-to-diameter ratio in PLAXIS, a group size of N = 196 was the limiting value
(s/D) of 3 was also adopted throughout the parametric for this work. Details of the pile/soil parameters adopted in

123
476 Acta Geotechnica (2014) 9:469–484

Fig. 10 Example of a variation in group configuration for a 16-pile group

1 soil stiffness profile with depth has been employed in these

N=4 LE analyses. A value of h/L = 3 was adopted such that the
0.8 N=16
N=36 effect of a stiff bearing stratum below the pile group has no
0.6 effect on the value of gg, thus simulating a pile group sit-
ηg

0.4
uated in an infinite soil mass.
The effect of soil stiffness nonlinearity on predictions of
0.2 gg has also been investigated in Fig. 13b. Since the HS
0 model predicts a linearly increasing soil stiffness with
1 10 100 depth (i.e. a ‘Gibson’ profile), LE predictions with a similar
m/n Gibson profile have also been plotted for comparison. A
Fig. 11 Influence of a variation in group configuration (m/n) on
database of stiffness efficiency field data (from nine pub-
PLAXIS predictions of gg; s/D = 3; L/D = 25 lished pile group case histories) provided by McCabe and
Lehane [7] is also included in Fig. 13b which is repre-
sented by a dotted line and described by Eq. (4). Since the
database consisted mainly of pile group sizes ranging
between 4 and 9 piles (with the exception of two case
histories), Eq. (4) is represented only for a value of
N ranging between 4 and 9.
In Fig. 13c, predictions of gg using the HS model in
PLAXIS have been compared to existing nonlinear analy-
ses and empirical methods [8, 25].
From the results presented in Fig. 13a–c, the authors
have arrived at the following conclusions:
(i) Predictions of gg using a LE soil model in PLAXIS
appear to agree well with the empirical approach
derived by Fleming et al. [5] and results using
Fig. 12 Example of pile group configurations for up to 16 piles,
PIGLET which have been documented by Poulos
larger groups modelled by adding squares to the periphery
and Davis [6].
the numerical analyses represented in this section are (ii) Although predictions of gg using a Gibson soil profile
summarised in Table 3. with a LE soil model in PLAXIS showed improved
As a calibration exercise, the LE soil model in PLAXIS agreement to field data compared to a linear profile in
has first been used to compare predictions of gg to pre- Fig. 13b, it is clear that the assumption of a LE soil
dictions obtained from existing empirical approaches and medium under-predicts the value of gg compared to
charts based on LE theory [5, 6] (shown in Fig. 13a). A the nonlinear HS model.
value of e = 0.55 has been adopted for the empirical (iii) From Fig. 13c, it is evident that existing nonlinear
approach proposed by Fleming et al. [5] in Fig. 13a. The predictions of gg documented by Comodromos and
soil properties adopted in the LE soil model are similar to Bareka [25], using the finite difference code FLAC3D
those employed in the HS model. In addition, a constant [26], are slightly lower than field data, while

123
Acta Geotechnica (2014) 9:469–484 477

predictions documented by Castelli and Maugeri [8] (a) 1

are significantly greater. The exponent of 0.15 in the
0.8 PLAXIS LE
approach documented by Castelli and Maugeri [8] was Fleming et al. [5]
derived from a limited database of case histories, not 0.6 Poulos & Davis [6]

ηg
all of which were friction pile groups. Thus, the
0.4
overprediction of gg compared to field data for this
approach might be attributable to the influence of end- 0.2
bearing pile groups in the database.
0
(iv) PLAXIS results using the HS model appear to show 0 10 20 30 40
good agreement with both the stiffness efficiency of N
the load test carried out by McCabe and Lehane [7] in 1
(b)
Belfast and the database of stiffness efficiencies PLAXIS LE -Constant
0.8 PLAXIS LE -Gibson
documented by the same authors. The measured field PLAXIS HS
data documented by McCabe and Lehane [7] plot McCabe & Lehane [7]
0.6 McCabe & Lehane [7]; [measured]
slightly lower than predictions determined by the HS

ηg
model due to the difference in adopted group 0.4

configurations. Moreover, the HS model in PLAXIS 0.2

can be considered as a marked improvement on
existing approaches for the prediction of gg for 0
0 10 20 30 40
rigidly-capped pile groups.
N
(c) 1
6.5 Effect of L/D
0.8

A drawback of the simple empirical approaches developed 0.6

ηg

by Castelli and Maugeri [8] and McCabe and Lehane [7] is 0.4
that the effect of the value of L/D on stiffness efficiency is
0.2
not taken into consideration, while Butterfield and Douglas
[1] averaged values of gg for all values of L/D. In Fig. 14, 0
0 10 20 30 40
the effect of varying L/D between 21 and 100 was inves-
N
tigated using the HS model in PLAXIS and plotted along
Castelli & Maugeri [8] Comodromos & Bareka [25]
with the results by Comodromos and Bareka [25]. Poulos McCabe & Lehane [7] PLAXIS HS
[27] noted, however, that the reduction in the ‘near-pile’ McCabe & Lehane [7]; [measured]

soil stiffness results in reduced pile-to-pile interaction;

Fig. 13 a Comparison of LE predictions of gg against the number of
therefore, in the present analyses, the value of L/D was piles in the group. b Comparison between LE and nonlinear
varied by modifying the diameter of the piles instead of predictions of gg against the number of piles in the group.
length of the piles so as to maintain the same average value c Comparison of nonlinear predictions of gg against the number of
piles in the group
of Rinter over the length of the piles. A value of h/L = 3
was, again, maintained representing a pile group in an
infinitely deep soil mass. The corresponding load levels for
each value of L/D were calculated as a FOS of 2.5 against
the ultimate capacity of a single pile (defined in Sect. 5.2).
Table 3 Pile/soil parameters adopted in the numerical analyses From Fig. 14, it can be seen that for a value of
represented in Fig. 11a–c
R determined as &3,000 according to the approach set out
Reference Analysis s/D L/D R FOS in Sect. 6.1, the HS model in PLAXIS predicts a slight
LE Poulos and Davis PIGLET 3 25 1000 n/a
decrease in the value of gg with an increase in
[6] L/D. Although the results by Comodromos and Bareka [25]
Fleming et al. [5] PIGLET – – – – plot higher than the results of the present study due to the
Present study PLAXIS 3 25 3,000 n/a FOS adopted by those authors, the reduction in gg with
Nonlinear Comodromos FLAC 3D 3 25 4,000 2.5 increasing L/D is consistent with the findings of Como-
and Bareka [25] dromos and Bareka [25] from nonlinear finite difference
Castelli and Empirical 3 – – – analyses as shown in Fig. 14 (R & 4,000), although those
Maugeri [8]
authors document a slightly greater decrease in gg with an
Present study PLAXIS 3 25 3,000 2.5
increase in L/D from 25 to 50. Furthermore, numerous

123
478 Acta Geotechnica (2014) 9:469–484

1 1
PLAXIS HS; L/D=21 El Sharnouby & Novak [28] Poulos & Davis [6]
PLAXIS HS; L/D=50 Southcott & Small [29] PLAXIS LE
PLAXIS HS; L/D=100 0.8 PLAXIS HS
0.8 Comodromos & Bareka [20]; L/D=25
Comodromos & Bareka [20]; L/D=50
0.6 0.6

ηg
ηg

0.4 0.4

0.2
0.2

0
0 0 10 20 30 40 50
0 10 20 30 40
E2/E1
N

Fig. 14 Comparison of existing approaches for estimating the Fig. 15 Comparison of existing approaches for estimating the
variation in gg with an increasing value of L/D variation in gg with an increasing value of E2/E1; N = 9;
L/D = 25; s/D = 3; h/L = 1

authors have documented a similar decrease in gg with 0.8

El Sharnouby & Novak [28]
increasing L/D for incompressible pile groups from LE Poulos & Davis [6]
analyses [6, 21]. PLAXIS HS
0.6 PLAXIS LE

6.6 Effect of E2/E1

ηg
0.4

From LE analyses, numerous authors have documented a

0.2
significant dependence of the value of gg on the com-
pressibility of the soil bearing stratum [6, 28, 29]. Thus,
0
PLAXIS 3-D was used to obtain predictions of gg for 0 10 20 30 40 50
values of E2/E1 ranging between 1 and 50. For the purpose E2/E1
of these analyses, a stiff bearing stratum has been included
in the soil model where a value of h/L = 1 was adopted. Fig. 16 Comparison of existing approaches for estimating the
variation in gg with an increasing value of E2/E1; N = 25;
The soil properties of the stiff bearing stratum are identical L/D = 25; s/D = 3; h/L = 1
to those adopted for the soft clay except that the soil
stiffness has been multiplied by a factor of E2/E1.
In Figs. 15 and 16, typical predictions of the variation of reducing pile settlement. As the depth from the base of the
gg with an increasing value of E2/E1 determined by both piles to the stiff stratum increases, however, its effect on
the HS model and a LE soil model have been plotted for a the value of gg becomes less significant.
9-pile and 25-pile group, respectively. The LE soil model Predictions of the variation in the value of gg with
in PLAXIS appears to show good agreement to the pre- h/L and E2/E1 determined by the HS model in PLAXIS
dictions documented by El Sharnouby and Novak [28], shown in Fig. 17, however, suggest that the effect of an
Poulos and Davis [6] and Southcott and Small [29], also increasing value of E2/E1 greater than 5 has little influence
idealising the soil as a LE medium for both the 9-pile and on gg for a value of h/L = 1.1 or higher. It may be seen that
25-pile group. From both the LE and nonlinear results in for the case of a stiff soil stratum beneath the base of a rigid
Figs. 15 and 16, it can be seen that a reduction in the pile group, the value of E2/E1 is relatively insignificant
compressibility of the bearing stratum has the effect of beyond a value E2/E1 & 10. This is consistent for various
increasing the value of gg. pile group sizes and configurations and is thus not depen-
dent on the particular group formation employed in the
6.7 Effect of h/L present study. A value of E2/E1 = 10 has thus been
maintained in these analyses.
The effect of the depth to a stiff bearing stratum on the The variation in gg with h/L determined by the HS
value of gg has also been investigated. A stiff bearing model in PLAXIS is shown in Fig. 18 for a 4-, 16-, 36- and
stratum has again been included in the soil model as 64-pile group with an L/D ratio of 25. Predictions deter-
described in Sect. 6.6 with the value of h/L now ranging mined by LE analyses (i.e., the charts by Poulos and Davis
between 1.1 and 3. A value of h/L = 1 has already been [6] and Butterfield and Douglas [1], where values were
considered in Sect. 6.6. The presence of a stiff stratum at averaged over the range of L/D) have also been plotted in
the base of the piles (as in Sect. 6.6) has the effect of Fig. 18 for comparison. From predictions determined by

123
Acta Geotechnica (2014) 9:469–484 479

0.8 7.2 Pile groups in infinitely deep soil mass (h/L C 3)

1
1.1
1.2 h/L The first case considered is a floating pile group in an
0.6 1.5
2 infinitely deep soil mass, in which case the effect of a stiff
soil stratum beneath the pile group can be ignored. As
ηg

0.4
mentioned in Sect. 6.4, a value of h/L = 3 was deemed
sufficient to represent a pile group in an infinitely deep soil
0.2
layer. The equations developed in the present paper are
modifications of the form of the expression for gg in
0 McCabe and Lehane [7], defined in Eq. (4). As previously
0 10 20 30 40 50
mentioned, variations in pile spacing and diameter are
E2/E1
captured by the Dg/D term. Eq. (4) was modified in order to
Fig. 17 Predictions gg with increasing value of E2/E1 determined by provide a best fit to results obtained using the advanced HS
PLAXIS; N = 36 piles; L/D = 21; s/D = 3 model in PLAXIS 3-D Foundation, rather than to a data-
base of field data. The resulting expression for gg (which
now captures the effect of L/D in a modified exponent term
both the HS model in PLAXIS and existing LE predictions, A) is defined as follows where gf is used to denote the value
it can be seen that the value of gg reduces with an of gg for a floating pile group in an infinitely deep soil
increasing depth to a stiff bearing stratum beneath the pile mass:
group but levels off at a value of h/L & 2.   A
Dg D
gg ¼ gf ¼ ð9Þ
Nþ1
7 Formulation of new FE-based approach where A ¼ 0:83ðL=DÞ0:071 :
Figure 19 confirms that Eq. (9) provides an excellent fit
7.1 Pile group scenarios to results obtained using the HS model in PLAXIS
(although N = 196 was the limiting number for this work,
An empirical approach has been developed based on results only N = 100 is presented in Fig. 19 for clarity).
from the advanced HS model in PLAXIS which have been As previously mentioned, all PLAXIS output from
presented in the previous sections. For the formulation of which the empirical formulations were derived used a
this approach, an equation was developed for the prediction spacing of s/D = 3. However, Fig. 20 illustrates that
of pile group stiffness efficiency for three different sce- Eq. (9) matches PLAXIS output very well for a range of
narios defined as follows: s/D values and justifies the use of Dg/D as a composite
(i) Pile group in infinitely deep soil mass measure of spacing and number of piles. In Fig. 20,

(ii) Pile group in a finite soil mass gplax gEq:9 has been plotted against s/D for a 4-pile and 16-
(iii) Pile group end-bearing on a stiff soil stratum pile group where gEq.9 is the value of gg calculated using
Eq. (9) and gplax is the value of gg predicted directly by

PLAXIS. The ratio gplax gEq:9 falls no more than approx-
imately 2 % from unity in this figure.
1
PLAXIS HS; N=4 Poulos & Davis [6]; N=4
PLAXIS HS; N=36 Poulos & Davis [6]; N=16 0.6
PLAXIS HS; N=64 Poulos & Davis [6]; N=25
0.8 Butterfield & Douglas [1]; N=5 Butterfield & Douglas [1]; N=13
Butterfield & Douglas [1]; N=25 PLAXIS HS; L/D=21
PLAXIS HS; L/D=50
0.6 PLAXIS HS; L/D=100
0.4
Eq. (9)
ηg

ηg

0.4

0.2
0.2

0 0
1 1.2 1.4 1.6 1.8 2 2.2 2.4 0 20 40 60 80 100
h/L N

Fig. 18 Comparison of predictions determined by the HS model to Fig. 19 Comparison between PLAXIS results and fitted results using
existing predictions; s/D = 3; L/D = 25 Eq. (9) for a variation in the value of N and L/D

123
480 Acta Geotechnica (2014) 9:469–484

1.3
(a) 1
N=4 N=16
1.2
Perfect agreement +/- 10% 0.8
1.1
ηEq.9/ ηplax

0.6

ηg
1
0.4
0.9
N=4 N=16
0.2 N=36 N=64
0.8 N=100 N=196
Eq. 11
0.7 0
2 2.5 3 3.5 4 4.5 5 0 10 20 30 40 50
s/D E2/E1

Fig. 20 Variation in gpred/gcalc with s/D (b) 1

(a) 1 0.8
N=4 N=36 N=64
0.8 N=100 N=196 Eq. 10
0.6

ηg
0.6 0.4
ηg

N=4 N=16
0.4 N=36 N=64
0.2 N=100 N=196
Eq. 11
0.2
0
0 10 20 30 40 50
0 E2 /E1
1 1.5 2 2.5
h/L (c) 1
(b) 1
N=4 N=36 N=64 0.8
0.8 N=100 N=196 Eq. 10

0.6
ηg

0.6
ηg

0.4
0.4
N=4 N=16
0.2 N=36 N=64
0.2 N=100 N=196
Eq. 11
0
0 0 10 20 30 40 50
1 1.5 2 2.5 E2/E1
h/L
1
Fig. 22 a Comparison between PLAXIS results and fitted results
(c) using Eq. (11) for a variation in the value of E2/E1; L/D = 21.
N=4 N=36 N=64
N=100 N=196 Eq. 10 b Comparison between PLAXIS results and fitted results using
0.8
Eq. (11) for a variation in the value of E2/E1; L/D = 50. c Comparison
between PLAXIS results and fitted results using Eq. (11) for a
0.6 variation in the value of E2/E1; L/D = 100
ηg

0.4

7.3 Pile groups in a finite soil mass (1 \ h/L \ 3)

0.2

0 The expression developed for the stiffness efficiency of a

1 1.5 2 2.5 pile group in a finite soil mass is a modification of the
h/L expression for gf in Eq. (9) and is defined in Eq. (10) as
Fig. 21 a Comparison between PLAXIS results and fitted results follows:
using Eq. (10) for a variation in the value of h/L; L/D = 21.  
b Comparison between PLAXIS results and fitted results using 1 6
gg ¼ gf þ B ð10Þ
Eq. (10) for a variation in the value of h/L; L/D = 50. c Comparison h=L
between PLAXIS results and fitted results using Eq. (10) for a
variation in the value of h/L; L/D = 100 where B ¼ 0:147ðL=DÞ0:272  ln N.

123
Acta Geotechnica (2014) 9:469–484 481

Table 4 Database of measured values of gg

Reference Soil conditions Pile type N s/D L/D E2/E1 h/L gmeas Case

McCabe and Lehane [7] Soft organic clayey silt Driven concrete 5 2.5 21 – 1.4 0.48 ii
Tejchman et al. [31] Sand Driven concrete 264 3.5 27 – – 0.066 i
Tejchman et al. [31] Medium sand Driven concrete 72 4.5 44 – – 0.311 i
Tejchman et al. [31] Medium sand Bored concrete 292 5.4 26.5 5 – 0.164 i
Caputo et al. [32] Silty sand Bored concrete 241 2.9 21 2 1 0.13 iii
Goosens and Van Impe [33] Medium stiff clay DCIS concrete 697 4 26 – – 0.017 i
Briaud et al. [34] Medium dense sand Driven steel 5 3 33.5 – – 0.6 i
Thorburn et al. [35] Soft very silty clay Driven concrete 55 7.1 27 – – 0.15 i
Thorburn et al. [35] Soft very silty clay Driven concrete 97 7.1 107 – – 0.184 i
O’Neill et al. [36] Very stiff clay Driven steel pipe 4 3 48 2 1 0.78 iii
Bartolomey et al. [37] Soft plastic clay Unspecified 464 4.1 32 – – 0.122 i
Bartolomey et al. [[37] Shingle Unspecified 192 3.3 52.5 – – 0.42 i
Bartolomey et al. [[37] Tough plastic clay Unspecified 6 1.8 15.5 – – 0.23 i
Bartolomey et al. [[37] Tough plastic clay Bored 9 3 39 – – 0.6 i
Cooke et al. [38] London clay Cast in situ 351 3.5 29 – – 0.044 i
Trofimenkov [39] Stiff silty clay Driven concrete 7 6 14 – – 0.426 i
Trofimenkov [39] Stiff silty clay Driven concrete 9 3 30 – – 0.52 i
Koerner and Partos [40] Medium dense sand Driven concrete 132 6.9 19 5 2.3 0.1 ii
Brand et al. [41] Soft sensitive marine clay Driven timber 4 5 40 5 1.4 0.89 ii
Brand et al. [41] Soft sensitive marine clay Driven timber 4 4 40 5 1.4 1 ii
Brand et al. [41] Soft sensitive marine clay Driven timber 4 3 40 5 1.4 0.5 ii
Brand et al. [41] Soft sensitive marine clay Driven timber 4 2.5 40 5 1.4 0.6 ii
Brand et al. [41] Soft sensitive marine clay Driven timber 4 2 40 5 1.4 0.4 ii
Komornik et al. [42] Organic clay Driven concrete 61 8.1 28 2 1.6 0.368 ii
Koizumi and Ito [43] Organic silty clay Driven steel 9 3 19 2 2.2 0.42 ii
Vargas [44] Stiff plastic clay Cast in situ 317 3.5 23 2.5 1 0.05 iii
Vargas [44] Stiff plastic clay Cast in situ 143 3.5 29 2.5 1 0.12 iii
Vargas [44] Medium clayey sand Cast in situ 205 3.5 29 3 1 0.19 iii
Feagin [45] Medium dense sand Driven timber and concrete 239 2.9 34 *2.5 1 0.094 iii
Feagin [45] Medium dense sand Driven timber and concrete 186 2.8 36 *4 1 0.197 iii

The expression considers the effect of h/L, L/D and N. 7.4 Pile groups end-bearing on a stiff soil stratum
The influence of E2/E1, however, is not considered here (h/L = 1)
since the value of E2/E1 has little effect on stiffness effi-
ciency beyond a value approximately equal to 5 as shown The expression for the stiffness efficiency of a pile group
in Fig. 17. It must therefore be conceded that Eq. (10) may end-bearing on a stiff soil stratum is also a modification of
overestimate the value of gg if the value of E2/E1 is less gf defined in Eq. (9). The modified equation takes account
than 5 and it is recommended that for a value of E2/E1 of 5 for the value of E2/E1 and is defined as follows:
or less, the value of gg should be obtained using Eq. (9) in  C
E2
order to provide a more conservative estimate. Results gg ¼ gf  ð11Þ
E1
obtained using the HS model have been compared to those
provided by Eq. (10) in Fig. 21a–c where the value of where C ¼ 0:112 ln N  0:112.
h/L is varied between 1 and 2.5 in order to again check the Although the exponent C varies slightly depending on
self-consistency of the method when all variables are the group size, N, it is not dependent on the value of
brought together. Results appear to agree well for values of L/D. Fig. 22a–c illustrate the fit between PLAXIS results
L/D ranging between 21 and 100. and the results provided by Eq. (11) when all variables are

123
482 Acta Geotechnica (2014) 9:469–484

(a) 1 b Fig. 23 a Comparison of predicted values of gg using present

Perfect Agreement approach to measured values. Value of R2 = 0.94. b Comparison of
Clay; Floating
predicted values of gg using approach by McCabe and Lehane [7] to
0.8 Clay; End-bearing
Sand; Floating measured values. Value of R2 = 0.84. c Comparison of predicted
Sand; End-bearing values of gg using approach by Castelli and Maugeri [8] to measured
ηg predicted

0.6 values. Value of R2 = -0.05. d Comparison of predicted values of gg

using approach by Fleming et al. [5] to measured values. Value of
R2 = -1.33
0.4

brought together. Results show good agreement with a

0.2 variation in the values of E2/E1, L/D and N.
Therefore, for all three cases presented in Sect. 7.2–7.4,
0 Eqs. (9)–(11) can be considered as a sufficiently accurate
0 0.2 0.4 0.6 0.8 1
ηg measured representation of the results predicted by PLAXIS 3-D
Foundation.
1
(b) Perfect Agreement
Clay; Floating
Clay; End-Bearing
0.8 Sand; Floating 8 Comparison of present approach to field data
Sand; End-bearing
ηg predicted

0.6 A database has been compiled of measured stiffness effi-

ciencies of pile groups fixed to a rigid pile cap, obtained
0.4 from Mandolini et al. [30] and supplemented with other
case histories in Table 4. Where full load–displacement
0.2 curves were available, stiffness efficiencies represent a FOS
of 2.5. The measured field data were compared to predic-
0 tions of gg obtained using the present equations, i.e., Eqs.
0 0.2 0.4 0.6 0.8 1
ηg measured (9)–(11). In calculating values of gg, each case history was
categorised as a case (i), (ii) or (iii) as defined in Sect. 7.1.
(c) 1
Predictions of gg obtained from the present equations
were then plotted against the documented measured values
0.8
in Fig. 23a, while the approaches by McCabe and Lehane
ηg predicted

0.6
[7] (Fig. 23b), Castelli and Maugeri [8] (Fig. 23c) and
Fleming et al. [5] (Fig. 23d) have also been included for
0.4 reference. It is clear that:
Perfect Agreement
Clay (1) Predictions of gg determined using the present
0.2
Sand approach show good agreement to measured field
data where the coefficient of determination (calcu-
0
0 0.2 0.4 0.6 0.8 1 lated as R2 = 0.94 for the clay data) is significantly
ηg measured higher than the corresponding values calculated for
existing empirical approaches in Fig. 23b–d. Future
(d) 1
published case histories may allow the expressions to
be tested over a broader range of the primary
0.8
influencing variables in Table 4.
(2) Although it is recognised by the authors that the
ηg predicted

0.6
parametric study in the present paper is developed
from the starting point of soft clay parameters and
0.4 does not take into account relative pile–soil stiffness,
Perfect Agreement the equations developed appear to show satisfactory
Clay
0.2
Sand
agreement with measured data associated with vari-
ous clay types described in Table 4.
0 (3) It can be seen from Fig. 23a that although the
0 0.2 0.4 0.6 0.8 1
database of measured values of gg presented in
ηg measured Table 4 consists primarily of friction pile groups, the

123
Acta Geotechnica (2014) 9:469–484 483

present approach provides improved agreement with (4) Even though the database of measured values of gg,
end-bearing field data compared to the approach by against which the current approach was appraised,
McCabe and Lehane [7] (Fig. 23b). consisted primarily of friction pile groups, it has been
(4) Although it is not theoretically justifiable to apply this shown that the present approach provides improved
method to piles in sands (as potential densification agreement with end-bearing field data compared to
during installation and dilation, for example, have not the approach by McCabe and Lehane [7].
been considered), data from pile groups in sand are (5) Although the adopted parameters in the present paper
also included as the match is remarkably good. This are representative of soft clay, predictions of gg
suggests that the geometrical factors captured in the obtained using the present approach show good
form of the Eqs. (9)–(11) developed have a significant agreement to a database of field data for a range of
bearing on the stiffness efficiency of pile groups in clay types.
sands. (6) The ability of the empirical equations to predict the
(5) It should be noted that although some of the N values in stiffness efficiency of pile groups in sand reasonably
the Table 4 database extend beyond the N = 196 scope well is likely to be an indication that pile group
of the equations developed, the effect of N on gg is geometry is important and that the equations devel-
much reduced beyond N = 196. The database other- oped capture the influence of geometry very well.
wise provides a good spread of parameter values across
the ranges for which the empirical equations were Acknowledgments The first author gratefully acknowledges the
developed, although further cases to test their validity College of Engineering and Informatics, NUI Galway, for providing a
(and possibly extent their range) will be useful. fellowship for this research.

9 Conclusions References

1. Butterfield R, Douglas RA (1981) Flexibility coefficients for the

A numerical study on rigid pile group stiffness efficiency design of piles and pile groups. Paper presented at the CIRIA,
has been presented and an approach derived from results of London
drained analyses using the advanced nonlinear Hardening 2. Basile F (2003) Analysis and design of pile groups. In: Bull JW
Soil model in PLAXIS 3-D Foundation. The formulae (ed) Numerical analysis and modelling in geomechanics, London,
pp 278–315
allow for an input of a range of parameters that have not 3. McCabe BA, Lehane BM (2006) Prediction of pile group
been considered in approaches developed by other authors. response using a simplified non-linear finite element model. In:
The following conclusions have emerged from this study: Proceedings of the 6th European conference on numerical
methods in geotechnical engineering, Austria, pp 589–594
(1) Although pile installation is not explicitly modelled in 4. Klar A, Leung YF (2009) Simple energy-based method for
PLAXIS, the present equations derived from results nonlinear analysis of incompressible pile groups in clays. J Geo-
tech Geoenviron Eng 135(7):960–965
using the HS model were shown to be a significant 5. Fleming WGK, Weltman AJ, Randolph MF, Elson WK (2009)
improvement on existing empirical methods for Piling engineering, 3rd edn. Taylor & Francis, Oxon
predicting rigid pile group stiffness efficiency of up 6. Poulos HG, Davis EH (1980) Pile foundation analysis and design.
to 196 piles based upon a database provided. Wiley, New York
7. McCabe BA, Lehane BM (2006) Behavior of axially loaded pile
(2) A conventional square pile group configuration was groups driven in clayey silt. J Geotech Geoenviron Eng 132(3):
used as the basis of the present parametric study in 401–410
order to err on the side of conservatism. The authors 8. Castelli F, Maugeri M (2002) Simplified nonlinear analysis for
recommend, however, that the application of present settlement prediction of pile groups. J Geotech Geoenviron Eng
128(1):76–84
approach should be limited to groups of m 9 n piles 9. Brinkgreve RBJ (2007) Plaxis 3D—foundation reference manual
with m/n B 10 to avoid overconservative estimates of Version 2. Plaxis bv, The Netherlands
pile group stiffness efficiency. Similarly, while the 10. Lehane BM (2003) Vertically loaded shallow foundation on soft
authors believe the adopted FOS of 2.5 represents a clayey silt. Proc ICE Geotech Eng 156(1):17–26
11. McCabe BA (2002) Experimental investigations of driven pile
commonly employed load level in pile group design, group behaviour in Belfast soft clay. PhD, Trinity College, Dublin
the application of the present approach to lower FOS 12. Schanz T, Vermeer PA, Bonnier PG (1999) The hardening soil
values will provide conservative predictions. model: Formulation and verification. In: Brinkgreve RBJ (ed)
(3) It is acknowledged that while the variability of the Beyond 2000 in computational geotechnics. Ten years of
PLAXIS International. Proceedings of the international sympo-
parameter E2/E1 in the present database is limited, the sium, A.A. Balkema, Amsterdam, pp 281–296
range of stiffness efficiencies represented, however, is 13. Mayne PW, Kulhawy FH (1982) Ko-OCR relationships in soil.
quite broad. J Geotech Eng Div ASCE 108(GT6):851–872

123
484 Acta Geotechnica (2014) 9:469–484

14. Killeen MM, McCabe BA (2010) A numerical study of factors 32. Caputo V, Mandolini A, Viggiani C (1991) Settlement of a piled
governing the performance of stone columns supporting rigid foundation in pyroclastic soils. In: Proceedings of the 10th
footings on soft clay. Paper presented at the Numerical Methods European conference on soil mechanics and foundation engi-
in Geotechnical Engineering, Trondheim neering, Firenze, pp 353–358
15. Potyondy JG (1961) Skin friction between various soils and 33. Goosens D, Van Impe WF (1991) Long term settlements of a pile
construction materials. Geotechnique 11(4):339–353 group foundation in sand, overlying a clay layer. In: Proceedings
16. Tiwari B, Ajmera B, Kaya G (2010) Shear strength reduction at of the 10th European conference on soil mechanics and founda-
soil structure interface. Paper presented at the advances in anal- tion engineering, Florence, pp 425–428
ysis, modelling and design, Florida 34. Briaud JL, Tucker LM, Ng E (1989) Axially loaded 5 pile group
17. Strick van Linschoten CJ (2004) Driven pile capacity in organic and single pile in sand. In: Proceedings of the 12th international
clays, with particular reference to square piles in the Belfast conference on soil mechanics and foundation engineering, Rio de
sleech. [Link]. Thesis, University of London (Imperial College) Janeiro, Br, Publ by A.A. Balkema, pp 1121–1124
18. Bjerrum L (1972) Embankments on soft ground. In: Proceedings 35. Thorburn S, Laird C, Randolph MF (1983) Storage tanks founded
of the specialty conference on performance of earth and earth- on soft soils reinforced with driven piles. In: Proceedings of the
supported structures, New York, ASCE, pp 1–54 conference of recent advances in piling and ground treatment for
19. Poulos HG (1968) Analysis of the settlement of pile groups. foundations, ICE, London, pp 157–164
Geotechnique 18(4):449–471 36. O’Neill MW, Hawkins RA, Audibert JME (1982) Installation of
20. Cooke RW (1974) The settlement of friction pile foundations. pile group in overconsolidated clay. J Geotech Eng Div ASCE
Paper presented at the conference on tall buildings, Kuala Lum- 108 (GT11):1369–1386
pur, Malaysia 37. Bartolomey AAea Pile foundation settlement (1981) In: Pro-
21. Caputo V, Viggiani C (1984) Pile foundation analysis: a simple ceedings of the 10th international conference on soil mechanics
approach to nonlinearity effects. Riv Ital Geotec 18(1):32–51 and foundation engineering, Stockholm, Sweden, pp 611–614
22. Jardine RJ, Potts DM, Fourie AB, Burland JB (1986) Studies of 38. Cooke RW, Bryden Smith DW, Gooch MN, Sillett DF (1981)
the influence of non-linear stress-strain characteristics in soil- Some observations of the foundation loading and settlement of a
structure interaction. Geotechnique 36(3):377–396 multi-storey building on a piled raft foundation in london clay.
23. Randolph MF (2003) PIGLET: analysis and design of pile groups Proc Instn Civ Engrs 7(1):433–460
users’ manual. Perth, Australia 39. Trofimenkov J (1977) Panel contribution, Session 2, Behaviour of
24. Chen SL, Song CY, Chen LZ (2011) Two-pile interaction factor foundation and structures. In: Proceedings of the 9th international
revisited. Can Geotech J 48(5):754–766 conference on soil mechanics and foundation engineering, Tokyo,
25. Comodromos EM, Bareka SV (2009) Response evaluation of pp 370–371
axially loaded fixed-head pile groups in clayey soils. Int J Numer 40. Koerner RM, Partos A (1974) Settlement of building on pile
Anal Methods Geomech 33(17):1839–1865 foundation in sand. J Geotech Eng Div ASCE 100(GT3):265–278
26. Itasca Consulting Group Inc (2000) FLAC 3D: fast Lagrangian 41. Brand EW, Muktabhant C, Taechathummarak A (1972) Load
analysis of continua. User’s manual, Minneapolis tests on small foundations in soft clay. In: Proceedings on the
27. Poulos HG (1988) Modified calculation of pile group settlement performance of earth and earth-supported structures, ASCE,
interaction. Research report, vol R582. University of Sydney, pp 903–928
School of Civil and Mining Engineering 42. Komornik A, Wiseman G, Zeitlen JG (1972) Buidling settlement
28. El Sharnouby B, Novak M (1990) Stiffness constants and inter- on end-bearing driven piles. In: Proceedings on the performance
action factors for vertical response of pile groups. Can Geotech J of earth and earth-supported structures, ASCE, pp 1135–1154
27(6):813–822 43. Koizumi Y, Ito K (1967) Field tests with regard to pile driving
29. Southcott PH, Small JC (1996) Finite layer analysis of vertically and bearing capacity of piled foundations. Soil Found 7:30–53
loaded piles and pile groups. Comp Geotech 18(1):47–63 44. Vargas M (1948) Building settlement observations in Sao Paulo.
30. Mandolini A, Russo G, Viggiani C (2005) Pile foundations: In: Proceedings of the 2nd international conference on soil
experimental investigations, analysis and design. Ground Eng mechanics and foundation engineering, Rotterdam, p 13
38(9):34–35 45. Feagin LB (1948) Performance of pile foundations of navigation
31. Tejchman A, Gwizdala K, Dyka Y (2001) Analysis of settlements locks and dams on the upper mississippi river. In: Proceedings of
of piled foundations. In: Proceedings of the 15th international the 2nd international conference on soil mechanics, Rotterdam,
conference on soil mechanics and foundation engineering, pp 98–106
Istanbul, pp 1025–1030

123