Soils and Foundations 2014;54(4):760–776

HOSTED BY The Japanese Geotechnical Society

Soils and Foundations

www.sciencedirect.com
journal homepage: www.elsevier.com/locate/sandf

Settlement performance of pad footings on soft clay supported

by stone columns: A numerical study
Micheál M. Killeena,1, Bryan A. McCabeb,n
a
Arup Consulting Engineers, Ringsend Road, Dublin, Ireland
b
College of Engineering and Informatics, National University of Ireland, Galway, Ireland
Received 14 March 2013; received in revised form 10 April 2014; accepted 22 April 2014
Available online 28 July 2014

Abstract

The settlement behaviour of small loaded areas (such as pad and strip footings) on soft soil supported by stone columns is poorly understood.
The lack of confinement associated with peripheral columns and the sharp stress decay with depth are features which are incompatible with the
widely-used unit cell method for infinite column groups thereby rendering small group behaviour particularly difficult to analyse. Useful field
data is virtually non-existent and although innovative laboratory modelling has been insightful, the findings cannot easily be extrapolated to field
scale. In this study, a 3-D finite element analysis in conjunction with an elastic–plastic soil model is used to identify the effect of variables in the
design process and interactions between them: these include column arrangement, spacing, length, and Young's modulus of the column material.
A simplified method is proposed to relate the settlement of small groups to a reference unit cell settlement predicted by current analytical
approaches.
& 2014 The Japanese Geotechnical Society. Production and hosting by Elsevier B.V. All rights reserved.

Keywords: Footings; Stone columns; Settlement improvement factor

1. Introduction stiffness and strength of the compacted stone serve to

It is well established that vibro-replacement stone col-
umns provide an effective means of increasing the bearing
capacity and reducing the compressibility of soft soils. The
benefits arise from the fact that there is partial replacement
of the host soil by a more competent material; the enhanced
n
Corresponding author. Tel.: þ353 91 492021; fax: þ 353 91 494507.
E-mail address: bryan.mccabe@nuigalway.ie (B.A. McCabe).
1
Formerly of the College of Engineering and Informatics, National
University of Ireland, Galway, Ireland.
Peer review under responsibility of The Japanese Geotechnical Society.
relieve the soil of some of its load. Stone columns can be
used in a variety of scenarios ranging from small loaded
areas (e.g. footings) to wide loaded areas (e.g. floor slabs,
embankments).
The optimum deployment of stone columns beneath
small loaded areas (such as pad and strip footings) is
arguably the most challenging aspect of stone column
settlement prediction in soft soils. Sexton et al. (2013)
have used 2-D axisymmetric Finite Element (FE) modelling
to demonstrate that some recent analytical methods (Castro
and Sagaseta, 2009 and Pulko et al., 2011 in particular)
are successful in capturing the key factors influencing

http://dx.doi.org/10.1016/j.sandf.2014.06.011
0038-0806/& 2014 The Japanese Geotechnical Society. Production and hosting by Elsevier B.V. All rights reserved.
 M.M. Killeen, B.A. McCabe / Soils and Foundations 54 (2014) 760–776
the performance of infinite column grids, which can be
implemented mathematically using a unit cell model.
However, unit cell models cannot capture the behaviour
of peripheral group columns beneath footings that are not
equally confined on all sides. The smaller the group, the
more influential these peripheral columns become. Further-
more, the increment of vertical stress beneath footings
decays much more acutely with depth than that beneath
widely loaded areas, allowing partial depth treatment to be
used in practice even though analytical methods are poorly
developed in this respect.
A database of measured field settlement improvement
factors in fine-grained soils compiled by McCabe et al.
(2009) highlights a dearth of such data for strip and pad
footings. High quality physical models of footings on soft
clay supported by stone columns (i.e. Muir Wood et al.,
2000; McKelvey et al., 2004; Black et al., 2011; Sivakumar
et al., 2011; Shahu and Reddy, 2011) have been informa-
tive, although there are obvious difficulties in extrapolating
model test findings to field scale, and the proportion of the
area under each footing that has been replaced with stone in
some of these tests has tended to lie at the high end of what
might commonly be used in practice. Research in which the
finite element method has been used to model ground
improved with stone columns relates to wide area loading
in the main, using either unit cell (e.g. Domingues et al.,
2007a) or 2-D axisymmetric (e.g. Elshazly et al., 2008a)
approximations. Some 3-D modelling of large column grids
has also been carried out (i.e. Gäb et al., 2008); however,
other than Kirsch (2008) no 3-D modelling of column-
supported footings, carried out in conjunction with an
advanced constitutive soil model, has been published. In
this paper, PLAXIS 3D Foundation is used to model the
behaviour of rigid square pad footings supported by various
stone column configurations. The soil profile at the former
UK geotechnical test site at Bothkennar, Scotland, is used
as the basis of the modelling as it is representative of many
soft soil profiles internationally for which the applicability
of stone columns is of growing interest. This research
identifies the roles and interactions of key factors relevant
to the settlement design of small groups of stone columns,
such as column arrangement, spacing and length, as well as
Fig. 1. Typical layout of a 3  3 group of columns beneath a 3 m square footing in the Bothkennar soil profile (see Section 4.1).
761
the strength and stiffness of the column material. A simple
design method predicting the settlement of groups of
floating (partial depth) or end-bearing (full depth) columns
is proposed, based upon geometrical corrections to unit cell
settlement predictions.
2. General finite element modelling details
2.1. Column group variables
PLAXIS 3D Foundation FE software (Version 2.2) has
been used in this study to capture the three-dimensional
nature of small groups of columns supporting pad footings.
The settlement performance of various configurations of
columns was investigated by varying the number of columns
(N), column spacing (s) on a square grid (with one
exception) and column length (L). The combination of N
and s dictated the footing width (B). The diameter of stone
columns (d) is normally not a design variable as the poker is
of fixed diameter (430 mm used for the bottom feed system)
and a column diameter of d ¼ 600 mm was adopted for the
subsequent Finite Element Analysis (FEA) which is a
typical average diameter of columns formed in soft soils.
The area replacement ratio or area ratio is a parameter which
measures the proportion of in situ soil replaced with stone
columns. For a finite group of columns, the area ratio is
defined as the ratio of the footing area (A) to the total area of
columns beneath the footing (AC):
A 4B2
¼ ð1Þ
AC Nπd2
However, for an infinite grid of columns, the area ratio is
defined as the ratio of the area of the zone of influence of one
column (A ¼ s2 for a square grid) to the area of a single column
(AC):
A 4s2
¼ 2 ð2Þ
AC πd
The column length was increased in 1 m increments in
the subsequent FEA from L ¼ 0 m (i.e. an untreated footing)
to L ¼ 13 m to examine the settlement performance corre-
sponding to partial depth treatment. The final analysis at
762 M.M. Killeen, B.A. McCabe / Soils and Foundations 54 (2014) 760–776

L ¼ H ¼ 13.9 m (H ¼ depth of the Carse clay modelled) 2.2. Column configurations

represents a full length stone column founded on a rigid
layer. The various column group parameters are illustrated
for a 3  3 group of columns beneath a 3 m square footing
in Fig. 1.
The influence of other design parameters such as column
0
stiffness (Ecol), column friction angle ðϕcol Þ and potential
increased coefficient of lateral earth pressure (K) arising from
installation were considered, as well as the presence of a stiff
crust typical of soft clay profiles and the position of columns
relative to the edge of the footing. These variables were
investigated for groups of four, five and nine columns, a subset
of the full parametric study.
2.2.1. Influence of column confinement
The degree of lateral support or confinement experienced
by columns in a group is dependent upon the number of
columns and the area ratio. The influence of confinement
upon the settlement performance of small groups of columns
was investigated by analysing single columns, 2  2, 3  3,
4  4 groups and infinite column grids at various spacings,
which represent the spectrum of A/AC values typically
encountered in practice, as identified in the McCabe et al.
(2009) database (Fig. 2 and Table 1). Single columns and
infinite grids of columns provide a useful frame of reference

Fig. 2. Column configurations to examine influence of column confinement.

Table 1
Column configurations to investigate factors influencing column confinement.
0
No. of columns, N s B A/AC L E50,col ϕcol
(m) (m) (–) (m) (MPa) (1)

1 1, 1.5, 2 1, 1.5, 2 3.5, 8.0, 14.1 1.0–13.9 70 45

4 1, 1.5, 2 2, 3, 4 3.5, 8.0, 14.1 1.0–13.9 70 45
9 1, 1.5, 2 3, 4.5, 6 3.5, 8.0, 14.1 1.0–13.9 70 45
16 1, 1.5, 2 4, 6, 8 3.5, 8.0, 14.1 1.0–13.9 70 45

Fig. 3. Column configurations to examine influence of column position relative to footing edge.

Fig. 4. Column configurations to examine the influence of column compressibility, strength, installation effects and the presence of a stiff crust.
 M.M. Killeen, B.A. McCabe / Soils and Foundations 54 (2014) 760–776 763

Table 2
Column configurations to investigate influence of other parameters.
0
No. of columns, N s B A/AC L E50,col ϕcol
(m) (m) (–) (m) (MPa) (1)

4 1.0 3 8.0 1.0–13.9 70 45

4 1.5 3 8.0 1.0–13.9 70 45
4 2.0 3 8.0 1.0–13.9 70 45

4 2.0 3 8.0 1.0–13.9 30, 50, 70 45

5 1.0 3 6.4 1.0–13.9 30, 50, 70 45
9 1.0 3 3.5 1.0–13.9 30, 50, 70 45

4 2.0 3 8.0 1.0–13.9 70 40, 45, 50

5 1.0 3 6.4 1.0–13.9 70 40, 45, 50
9 1.0 3 3.5 1.0–13.9 70 40, 45, 50

for the other configurations as they correspond to zero and

full lateral confinement, respectively. Columns are posi- CL
tioned on a square grid and the edge of footings is located at
a distance of half the column spacing (0.5  s) from the 8B

centre-line of the outer row of columns. CL

½B

2.2.2. Influence of other parameters

The influence of column position relative to the footing
edge was investigated by examining the settlement perfor-
mance of a group of four columns which are progressively
moved towards the footing edge beneath a 3 m square
footing (Fig. 3).
The influence of column stiffness, friction angle, coeffi-
cient of lateral earth pressure and the presence of the stiff
crust was investigated for three configurations of columns
illustrated in Fig. 4. The influence of column stiffness was
investigated by varying E50 (E50 ¼ secant Young's modulus
Fig. 5. Typical mesh adopted for FEA employing symmetry and with locally
at 50% of the deviator stress) from 30 to 70 MPa, while the
refined mesh around footing.
friction angle was varied from 401 to 501 in accordance with
the experience of McCabe et al. (2009) for the bottom-feed
system.
The changes to the in situ stress regime due to the
installation of stone columns are difficult to quantify due to 2.3. Pad footing
large strains and vibrations associated with poker penetra-
tion. Many authors (Kirsch, 2006; Castro, 2007; Gäb et al., Pad footings of 0.6 m thickness and founded at 0.6 m below
2007) have reported an increase in pore pressure, horizontal ground level (Fig. 1) are modelled as a linear elastic material
stress and stiffness in the surrounding soil. However, the with properties representative of concrete (i.e. unit weight,
increase in horizontal stress and soil stiffness is negated by γ¼ 24 kN/m3; Young's modulus, E¼ 3  107 kPa and Poisson's
remoulding and dynamic excitation of the soil close the ratio, υ¼ 0.15). The pad footings act as a rigid loading cap on
column. Despite this complexity, Elshazly et al. (2008a) top of the stone columns. Since the settlement performance of
attempt to capture column installation effects by increasing stone columns is usually the governing criterion for the design
the coefficient of lateral earth pressure (K0) in the surround- of foundations on soft soils (e.g. Priebe, 1976) and is also
ing soil. The authors conducted a series of axisymmetric the focus of this paper, a typical working pressure of 50 kPa
FEA to back-analysis field load tests described by Mitchell (i.e. within the range quoted by Serridge and Sarsby, 2008) has
and Huber (1985) and suggest conservative estimates of K0 been applied to each footing; this pressure corresponds to a
in the range 0.85–1.7, which increase at lower area ratios factor of safety of approximately 2.75 on the ultimate bearing
(A/AC ¼ 2.5–4.8). capacity of an unreinforced footing at the Bothkennar site
A summary of the parameters varied is outlined in Table 2. (Jardine et al., 1995).
764 M.M. Killeen, B.A. McCabe / Soils and Foundations 54 (2014) 760–776
2.4. Meshing and boundaries
The FE mesh comprised 15-noded wedge isoparametric
elements and the necessary detailed mesh sensitivity studies
are reported in Killeen (2012). The lateral boundaries of the
computational domain allow vertical movement but restrict
radial movement, and boundary sensitivity studies indicated
that these should be positioned at a distance of 8B from the
centre of the pad to avoid influencing the results. The bottom
boundary is influenced by the ground profile which is
discussed in Section 4. In general, it was possible to exploit
symmetry to save computation time, as illustrated by the mesh
in Fig. 5.
2.5. Column–soil interface
The column compaction process involves forcing the stone
laterally into the host soil. Since the stone becomes tightly
interlocked with the surrounding soil, no discrete interface
zone exists as might be expected for a pile. Perfect adhesion is
assumed at the column–soil interface and, in keeping with
standard practice for modelling stone columns, interface
elements are not used (e.g. Kirsch, 2006; Gäb et al., 2008;
Shahu and Reddy, 2011). Many authors report the develop-
ment of a smear zone along the column–soil interface due to
column installation (Han and Ye, 2001; Weber et al., 2008;
Indraratna et al., 2013). This smear zone is characterised by
low permeability, which adversely affects the drainage capa-
city of stone columns and reduces the rate of consolidation.
However, since this paper focuses on the long term (i.e.
drained) settlement performance of stone columns, the con-
solidation rate is not relevant. Other than field trials by Kirsch
(2006), there is no information on the extent to which column
installation increases lateral earth pressures in the surrounding
soil, but increased K0 (coefficient of lateral earth pressure
at rest) values are considered as a variable in the parametric
study.
Settlement improvement factor, n
1.0 1.5 2.0 2.5 3.0 3.5
0 where σ is the confining stress and Eref is the reference soil
4 cols - s = 2.0 m
5 cols - s = 1.0 m
2 9 cols - s = 1.0 m
Column length, L (m)
Drained
4 Undrained &
Consolidation
6
10
12
14
Fig. 6. Influence of analysis type upon settlement improvement factors for
various groups of columns.
2.6. Drainage conditions
PLAXIS 3D enables long-term behaviour to be modelled in
two ways: (i) drained analysis, using effective stress para-
meters and (ii) undrained analysis, using effective stress
parameters, followed by consolidation. The use of effective
stress parameters allows the increase of shear strength with
effective stress to be captured. Method (ii) is a closer
simulation of the process in reality as a large bulk modulus
is added to the soil stiffness matrix in order to mimic undrained
conditions, and subsequently removed during the consolidation
phase to allow the transfer of stress to soil particles. However,
runs based upon method (ii) are far more time-consuming and
method (i) was preferred in this research. Fig. 6 shows that for
a square footing with B ¼ 3 m, loaded to 50 kPa and supported
by various configurations of stone columns outlined in the
inset to the figure, approaches (i) and (ii) produce quite similar
(if not perfectly consistent) results.
3. Hardening Soil model
Both the stone columns and the host soft clay are modelled
using the elasto-plastic Hardening Soil (HS) model. An
extension of the hyperbolic model developed by Duncan and
Chang (1970), the HS model includes soil dilatancy, captures
unload–reload behaviour and accounts for shear and volu-
metric hardening. Schanz et al. (1999) advocate its suitability
for both soft and stiff soils.
The basis of the HS model is that the stress–strain curve for
a standard drained triaxial test may be approximated by a
hyperbola, which is defined by a secant Young's modulus (E50)
and an unload–reload modulus (Eur). The Mohr–Coulomb
failure criterion defines yielding and plasticity is accounted
for using stress-dilatancy theory developed by Rowe (1962).
The stiffness of soils (E) tends to increase with confining stress
and this is captured by the HS model using a power law:

m
E σ
¼ ð3Þ
Eref pref
4.0
stiffness for a given confining pressure, pref. The exponent, m,
controls the relationship between soil stiffness and confining
pressure and Brinkgreve and Broere (2006) propose a value of
1.0 for soft soils.
Fig. 7. Bothkennar test site location (Nash et al., 1992b).
 M.M. Killeen, B.A. McCabe / Soils and Foundations 54 (2014) 760–776 765

Table 3
Summary of material parameters adopted for Bothkennar test site.

Soil parameter Crust Upper Carse clay Lower Carse clay

Depth (m) 0.0–1.5 1.5–2.5 2.5–14.5

Bulk unit weight, γ (kN/m3) 18.0 16.5 16.5
Over-consolidation ratio – – 1.5
Pre-overburden stress (kPa) 15 15 –
Coefficient of lateral earth pressure, K0 1.5 1.0 0.75
Effective cohesion, c' (kPa) 3 1 1
Angle of internal friction, ϕ' (1) 34 34 34
Initial voids ratio, e0 1.0 1.2 2.0
Compression index, CC 0.07 0.25 1.12
Swelling index, CS 0.01 0.03 0.16
Reference pressure, pref (kPa) 13 20 30
Vertical coefficient of permeability, kvert (m/day) 6.9  10  5 6.9  10  5 6.9  10  5
Horizontal coefficient of permeability, khorz (m/day) 1.0  10  4 1.0  10  4 1.0  10  4

The HS model does not account for the effects of secondary
settlement and, consequently, this paper focuses on the long
term primary settlement of small groups of stone columns.
4. Development of soil profile and material properties
for HS model
4.1. Bothkennar test site
The Bothkennar test site is located along the Firth of Forth,
near Grangemouth in Scotland (Fig. 7). The site is located
in a historic tidal estuary and primarily consists of soft uni-
form clay, which is extensively characterised in Géotechnique
(1992). The clay, often referred to as Carse clay, rests upon
Bothkennar gravel and is overlain by a stiff crust. The Carse
clay varies in thickness from 13 to 19 m across the site and is
classified by Paul et al. (1992) as a silty clay with an organic
content of 3–8%. Piezometers indicate that the groundwater
conditions are hydrostatic and the water table varies from 0.5
to 1.0 m below ground level.
The soil profile for the subsequent FEA is divided into three
layers: crust; upper Carse clay and lower Carse clay. The
development of material properties, outlined in Table 3, for
each of these layers is described in the remainder of this
section.
Hight et al. (1992) suggest that the Bothkennar test site may
have been subjected to various post-depositional processes
such as erosion, changes in groundwater level and bonding.
These processes correspond to a 15 kPa drop in effective
overburden stress, which agrees favourably with yield stress
ratios (equivalent to over consolidation ratios, OCR) measured
in the upper layers by Nash et al. (1992a) in a series of
oedometer tests. In this paper, the stress history of the upper
layers is modelled using a single pre-overburden pressure
(POP) value of 15 kPa (POP is defined as the difference
between the maximum vertical effective stress experienced
0
previously ðσ max Þ and the current in situ vertical effective stress
0 0 0
ðσ 0 Þ). However, a higher stress state ðOCR ¼ σ max = σ 0 ¼ 1:5Þ
is more appropriate for depths greater than 4 m, which
indicates that other post-depositional processes such as ageing
have occurred at the Bothkennar test site. Nash et al. (1992b)
measured the in situ coefficient of earth pressure at rest (K0)
and report a variation from 1.5 in the upper layers to 0.75
at depth.
Allman and Atkinson (1992) conducted a series of triaxial
tests to determine the strength characteristics of reconstituted
Carse clay. The Carse clay was formed into a slurry (at a water
content of 1.25 times the liquid limit) and then one-
dimensionally compressed into normally-consolidated or (with
swelling) into slightly over-consolidated states before shearing.
A critical state friction angle of ϕ' ¼ 341 is reported, which is
larger than would be expected for a high plasticity clay but
attributable to the high proportion of angular silt particles. In
order to ensure numerical stability, nominal values of 3 kPa
and 1 kPa were adopted for the effective cohesion (c0 ) in the
stiff crust and Carse clay layers, respectively.
The one-dimensional stiffness properties of the Carse clay
were investigated by Nash et al. (1992a) through a series of
incremental load, restricted flow and constant rate of strain
tests. The initial voids ratio (e0) and coefficient of compression
(CC) are reported for various depths and these can be readily
converted into oedometric moduli (Eoed) for the HS model
using Eq. (4) (Brinkgreve and Broere, 2006):
2:3ð1 þ e0 Þpref
oed ¼
Eref ð4Þ
CC
The relationship between E50 and Eoed is defined by:
E50 ¼ 1:25E oed ð5Þ
While the coefficient of swelling (CS) was not routinely
measured by Nash et al. (1992a), Allman and Atkinson (1992)
found the slopes of the normal compression (λ) and swelling
(κ) lines to be 0.181 and 0.025, respectively. This ratio was
used to determine the coefficient of swelling, i.e. λ/κ ¼ CC/
CS ¼ 7.2. The coefficient of swelling can be converted into an
unload–reload modulus using Eq. (6) (Brinkgreve and Broere,
766 M.M. Killeen, B.A. McCabe / Soils and Foundations 54 (2014) 760–776

2006): 17
Plaxis 3D Foundation
15 Priebe (1995): n0

Settlement improvement factor, n

2:3ð1 þ e0 Þð1þ νÞð1 2υÞpref
Priebe (1995): n2

ur ¼
E ref ð6Þ Pulko et al. (2011)
13
C S ð1  νÞ Balaam & Booker (1981)
11

9
4.2. Stone backfill
7

A bulk unit weight of γ ¼ 19 kN/m is adopted for the stone 3 5

backfill which is similar to values adopted by Mitchell and 3

Huber (1985), Gäb et al. (2008) and Domingues et al. (2007a). 1
An angle of internal friction of ϕ' ¼ 451 is adopted as the 2 4 6 8 10 12 14 16
default for the stone backfill on the basis of the finding that Area ratio, A/Ac

ϕ' ¼ 401 is conservative for stone columns formed using the Fig. 9. Comparison of settlement improvement factors for an infinite grid of
modern dry bottom feed system (McCabe et al., 2009). Kirsch end-bearing columns with analytical design methods.
(2004) proposed similar values. The angle of dilatancy is
related to the angle of internal friction through the well-known
empirical relationship, ψ0 ¼ ϕ0 –301. A nominal value of effec-
exceeded 8 mm/hr. Pad A was founded at 0.8 m below ground
tive cohesion (c0 ¼ 1 kPa) is adopted to ensure numerical
level and was 2.2 m wide. A second pad (Pad B) was not
stability.
modelled as significant creep settlements are incorporated in
Barksdale and Bachus (1983) observed that Young's moduli
the settlements recorded over a period in excess of 2 years.
for stone columns, which were back-calculated from measured
The load test on Pad A was simulated with PLAXIS 3D
settlements and recommended by other authors, vary from 30
Foundation as an undrained loading (with effective stress
to 58 MPa. However, since columns installed using the dry
material parameters) due to the short duration of the load test.
bottom feed system have been shown by McCabe et al. (2009)
However, the crust in the adopted soil profile is modelled as a
to exhibit better settlement performance than other systems, it
drained material as the lower part of the crust contains a
was deemed appropriate to adopt a higher Young's modulus
significant proportion of shelly fragments (Nash et al., 1992b)
for the stone backfill. Default values of E50 = 70 MPa and
and the upper part of the crust is above the groundwater level.
Eur ¼ 210 MPa were adopted. The exponent, m, which controls
In Fig. 8, it is clear that PLAXIS 3D Foundation captures
the stress dependency of soil stiffness for the HS model was
the trend of settlement versus applied pressure quite well,
chosen as m ¼ 0.3. These stiffness parameters (i.e. E50, Eur and
despite slightly over-predicting the settlement at low working
m) are in keeping with those adopted by Gäb et al. (2008).
loads. This may be explained as the location where the load
tests were conducted by Jardine et al. (1995) had a localised
4.3. Validation of Bothkennar soil profile shelly layer 0.3 m in thickness beneath the base of the footing,
which would have enhanced the settlement performance of the
Instrumented field load tests on unreinforced rigid pad footing.
footings at the Bothkennar test site, documented by Jardine
et al. (1995), were modelled as an exercise in validating the HS 5. Comparison of PLAXIS 3D and analytical design
parameters adopted for the soil. Pad A was loaded to failure method output
using kentledge in less than five days, with pauses in loading
occurring overnight and whenever the rate of settlement The settlement performance of an infinite grid of end-
bearing stone columns from PLAXIS 3D Foundation is
compared with various analytical design methods in Fig. 9.
140
The settlement performance is evaluated using a settlement
120 improvement factor, n, which is defined as the ratio of the
Jardine et al. (1995) PLAXIS 3D Foundation
Applied pressure (kPa)

100
settlement of an untreated footing (i.e. no stone columns) to
that of a treated footing (i.e. with columns). Settlement
80
improvement factors in Fig. 9 correspond to the application
60 of a 50 kPa uniform pressure on a concrete footing. A brief
summary of the design methods chosen for comparison,
40
adapted from Sexton et al. (2013), is presented in Table 4.
20 Another design method worthy of note (which is not listed
0 in Table 4) is that proposed by Castro and Sagaseta (2009).
0 20 40 60 80 100 120 140 160 180 200 The authors develop a comprehensive closed-form solution
Vertical displacement (mm) which accounts for the effects of both radial consolidation
Fig. 8. Comparison of PLAXIS 3D Foundation with actual load–displacement and loading history upon the settlement of stone columns.
behaviour for a pad footing observed by Jardine et al. (1995). However, for monotonic loading, as is the case with the FEA
 M.M. Killeen, B.A. McCabe / Soils and Foundations 54 (2014) 760–776
Table 4
Summary of analytical design methods chosen for comparison.
Author Column behaviour
Balaam and Booker (1981)  Elastic
 Accounts for radial expansion due to bulging
Priebe (1995)  Rigid-plastic and incompressible
 Modification factors introduced to account for:
(i) Column compressibility
(ii) Overburden stress
Pulko et al. (2011)  Elastic–plastic with a non-associated flow rule based on a constant angle of dilation
 Accounts for radial expansion of column and initial stress state in surrounding soil.
in this paper, the solution proposed by Castro and Sagaseta
(2009) is similar to Pulko et al. (2011) at the final state.
All of the aforementioned design methods are based on the
unit cell concept which is implemented in the FEA. However,
the design methods are based upon more simplistic material
models and it is necessary to select appropriate material
parameters which are consistent with the FEA:
 Priebe (1995) and Pulko et al. (2011) both account for
the initial stress state of the surrounding soil using the
coefficient of lateral earth pressure at rest, K0. While Pulko
et al. (2011) allow for a direct input of K0, Priebe (1995)
assumes that K0 ¼ 1 as a result of column installation.
Consequently, it was necessary to modify Priebe (1995) to
allow for alternative values of K0 for the soil for compara-
tive purposes.
 While the HS model requires the selection of two material
parameters (E50 and Eur) to capture the stress path depen-
dency of soil stiffness, the design methods are based on a
single stiffness modulus. E50 is adopted for the stone
column as this is suitable for primary loading (Brinkgreve
and Broere, 2006). However, a higher stiffness modulus is
adopted for the surrounding soil, as the soil is subject to
lower stress levels (due to stress concentrations which
develop in the stiffer column) and the fact that soils tend
to exhibit increasing stiffness at low strain levels. An
average of E50 and Eur is chosen to represent the stiffness
modulus of the surrounding soil.
 The stiffness of soil tends to increase with confining
pressure and this is captured by the HS model using the
power law outlined in Eq. (3). The stress level dependency
of soil stiffness is accounted for in the analytical design
methods by averaging the minor principal stresses, and
subsequently determining the average stiffness, in each
layer. The minor principal stress is adopted for the confin-
ing stress in accordance with Brinkgreve and Broere (2006),
as a FEA by Balaam and Booker (1985) shows that an
infinite grid of columns are in a triaxial stress state during
loading.
It can be seen in Fig. 9 that Balaam and Booker (1981) over-
estimate n values, which is due to the assumption of linear
elasticity for the stone column (Castro and Sagaseta, 2009).
767
Column yielding
Not considered
Mohr–Coulomb
Mohr–Coulomb
Pulko et al. (2011) is an extension of Balaam and Booker
(1981), as it accounts for column yielding, thereby predicting
more realistic n values. It appears that columns are in an elastic
state at low area ratios as Pulko et al. (2011) and Balaam and
Booker (1981) predict similar n values. However, as area ratio
increases the extent of plasticity becomes more pronounced
and Pulko et al. (2011) predict lower n values which agree
favourably with PLAXIS 3D Foundation.
The basic (n0) and modified (n2) settlement improvement
factors developed by Priebe (1995) are also shown in Fig. 9.
The basic settlement improvement factors under-predict n
values, whereas the modified settlement improvement factors,
which account for column compressibility and overburden
stress, are in closer agreement with PLAXIS 3D Foundation. It
can be seen that Priebe (1995) under-estimates n values at low
area ratios. This is due in part to the assumption of rigid-plastic
behaviour for the columns, which neglects strain developed
when columns are in an elastic state. The effect of this
assumption diminishes with increasing area ratio, as the extent
of plasticity becomes more pronounced and plastic strains
dominate the settlement of stone columns. At high area ratios,
n2 values slightly over-estimate PLAXIS 3D Foundation. This
may be explained as Priebe's modification for column com-
pressibility is based upon the constrained modular ratio of the
stone column and soil (i.e. oedometric conditions). This
assumption over-estimates n values as it does not account for
column bulging.
This exercise gives early confidence in the ability of
PLAXIS 3D Foundation to capture stone column behaviour.
6. Results of FEA
6.1. Column confinement
The effect of column configuration was examined as
outlined in Section 2.2. Various column lengths were
considered which allows the relationship between column
length, area ratio and the number of columns to be
investigated.
6.1.1. Infinite grid of columns
The influence of column length and area ratio upon the
settlement performance of an infinite grid of columns is shown
in Fig. 10. A negligible reduction in footing settlement is
768 M.M. Killeen, B.A. McCabe / Soils and Foundations 54 (2014) 760–776

Footing settlement (mm) Settlement improvement factor, n

0 200 400 600 1 3 5 7 9 11
0 0
Unit Cell
A/Ac = 3.5
2 2
A/Ac = 8.0
Column length, L (m)

Column length, L (m)

A/Ac = 14.1
4 4

6 6

8 8

10 10
Unit Cell
12 A/Ac = 3.5 12
A/Ac = 8.0
A/Ac = 14.1
14 14
(i) (ii)
Fig. 10. Variation of (i) footing settlement and (ii) settlement improvement factor with column length for an infinite grid of columns.

Settlement improvement factor, n Settlement improvement factor, n Settlement improvement factor, n

1 3 5 7 9 11 1.0 1.5 2.0 2.5 3.0 1.0 1.2 1.4 1.6 1.8
0 0 0
A/A = 3.5 A/A = 8.0 A/A = 14.1
1 col 1 col 1 col
2 2 2
2×2 2×2 2×2

Column length, L (m)

Column length, L (m)

Column length, L (m)

3×3 3×3 3×3

4 4×4 4 4×4 4 4×4
Unit cell Unit cell Unit cell
6 6 6

8 8 8

10 10 10

12 12 12

14 14 14

Settlement improvement factor, n Settlement improvement factor, n Settlement improvement factor, n

1.0 2.0 3.0 4.0 5.0 6.0 1.0 1.3 1.6 1.9 2.2 2.5 1.0 1.1 1.2 1.3 1.4 1.5
0 0 0
Normalised column length, L/B
Normalised column length, L/B

Normalised column length, L/B

2 2 2

4 4 4

6 6 6

8 8 8

10 A/A = 3.5 10 A/A = 8.0 10 A/A = 14.1

1 col 1 col 1 col
2×2 2×2 12 2×2
12 12
3×3 3×3 3×3
4×4 4×4 4×4
14 14 14

Fig. 11. Variation of settlement improvement factor with (a) column length and (b) normalised column length for small groups of columns spaced at area ratios of
(i) 3.5, (ii) 8.0 and (iii) 14.1.

observed in Fig. 10(i) with increasing column length up to 6.1.2. Small groups of columns
L E2 m. This is attributed to the stiff crust, which extends to The relative settlement performances of the various stone
1.9 m, as installing stone columns in an already competent column group configurations identified in Table 1 are shown in
layer will do little to enhance the settlement performance. a plot of n against column length (L) in Fig. 11(a). As with an
A significant reduction in footing settlement, which yields a infinite grid of columns, a negligible increase in n values is
corresponding increase in n values (Fig. 10(ii)), is observed observed for all columns shorter than L E2 m. However, n
with increasing column length thereafter in the Carse clay. values increase with column length thereafter and it appears
It appears that the option of using of longer columns will be that the influences of area ratio and column length upon the
most advantageous at low area ratios. settlement performance of small groups of stone columns are
 M.M. Killeen, B.A. McCabe / Soils and Foundations 54 (2014) 760–776
somewhat coupled, as the influence of column length becomes
more pronounced at low area ratios.
A continuous increase in n values with column length is
observed in Fig. 11(a–i) for columns at low area ratios
(A/AC ¼ 3.5). An abrupt increase in n values is observed when
moving from floating to end-bearing columns (i.e. from
L ¼ 13 m to L ¼ 13.9 m). This suggests that closely-spaced
columns transfer a large proportion of the applied load to the
base of columns. Furthermore, the magnitude of the abrupt
increase in n values increases with the number of columns,
which indicates that the proportion of the applied load
transferred to the base of columns increases for larger groups
of columns. This is akin to a punching mode of deformation
and is consistent with findings by Muir Wood et al. (2000) and
Black et al. (2011) who conducted a series of small scale
laboratory studies on small groups of stone columns. Muir
Wood et al. (2000) observed four modes of deformation
(bulging, shearing, punching and bending) and noted that
punching was most pronounced for short columns at low area
ratios. Similar findings are reported by Black et al. (2011) who
observed the development of ‘block failure’ for a small group
of long, closely-spaced columns (A/AC ¼ 2.5), which acted as a
single entity and punched into the underlying soil.
A constant increase in n values with column length is
evident up to approximately L E 5 m for columns at A/AC ¼ 8.0
and 14.1 in Fig. 11(a–ii) and (a–iii), respectively. However, the
rate of increase in n values with column length slows down
thereafter, which suggests that the mode of deformation may
have transitted from punching to bulging at higher area ratios.
Similar observations were made by Hughes and Withers
(1974), McKelvey et al. (2004) and Black et al. (2011) in
small scale laboratory tests on single and small groups of
(×10-3)
Low strain in
soil bounded by
outer columns
Uniform strain at
base of columns
(i)
Fig. 12. Distribution of total shear strains through a cross-section of a 3  3 group of 8 m long columns at A/AC of (i) 3.5, (ii) 8.0 and (iii) 14.1.
769
columns, as the combined shaft and base resistance surpasses
the lateral confining stress with increasing column length.
Columns which are bulging cannot transfer the applied load to
depth and, thus, do not yield a significant increase in n values
with column length.
The deformed shape of a 3  3 group of 8 m long columns
at A/AC ¼ 3.5, 8.0 and 14.1 is illustrated through plots of
total shear strain in Figs. 12(i), 12(ii) and 12(iii), respec-
tively. It can be seen in Fig. 12(i) that limited bulging occurs
in the group of closely-spaced columns, which reflects the
enhanced lateral restraint experienced by columns that
precipitates a punching mode of deformation. In contrast,
columns at higher area ratios tend to bulge in the upper
layers and bend outwards from the central columns (Fig. 12
(iii)). Punching is reduced at higher area ratios as column
bulging minimises the magnitude of applied load transferred
to the base of columns.
The increased level of lateral confinement associated with
larger groups of columns is evident in Fig. 11 as an increase in
the number of columns (for a given area ratio) yields higher
n values. However, this trend of increasing n values with group
size does not hold for an infinite grid of floating columns
shorter than L E 7 m. This can be attributed to the constant
increment of vertical stress with depth associated with an
infinite grid of columns, which induces significant settlement
beneath the base of floating columns. The increase in n values
with column length is most pronounced for an infinite grid of
columns at all area ratios and the positive effects of confine-
ment are evident for longer columns. The different variations
of n values with column length observed for small groups and
infinite grids of columns highlights the difficulty of employing
settlement design methods based on the unit cell model to
(×10-3) (×10-3)
Development
of shear zone
(ii) (iii)
770 M.M. Killeen, B.A. McCabe / Soils and Foundations 54 (2014) 760–776

Settlement improvement factor, n Settlement improvement factor, n

1.0 1.2 1.4 1.6 1.8 1.0 2.0 3.0 4.0
0 0
4 cols - s = 1.0 m 4 cols - s = 2.0 m
4 cols - s = 1.5 m 5 cols - s = 1.0 m
2 4 cols - s = 2.0 m 2
9 cols - s = 1.0 m
Column length, L (m)

Column length, L (m)

Footing width, B = 3 m E50,col = 30 MPa
4 4 E50,col = 50 MPa
E50,col = 70 MPa
6 6

8 8

10 10

12 12

14 14
Fig. 13. Influence of column position upon settlement improvement factors. Fig. 14. Influence of column stiffness upon settlement improvement factors.

capture the settlement performance of small groups of stone

columns.
The same data are presented in Fig. 11(b), this time with the
column length normalised by the footing width (i.e. L/B); n
values are not presented for the ‘unit cell’ model as the footing
width is theoretically infinite. The classical Boussinesq (1885)
solution for the distribution of vertical stress in an elastic
medium indicates that the increment of stress applied at the
surface is considered negligible at a depth of 2B (B ¼ footing
width). Muir Wood et al. (2000) suggest that the dominant
strain in columns is dependent upon footing width and
postulate the existence of a critical length, beyond which the
mode of deformation changes from punching to bulging.
However, Fig. 11(b) shows that the settlement performance
of small groups of stone columns is dependent upon a
combination of area ratio, column length and number of
Fig. 15. Influence of column strength upon settlement improvement factors.
columns, rather than footing width alone. The laboratory tests
undertaken by Muir Wood et al. (2000) incorporate a weaker
0
sand column ðϕcol ¼ 301Þ, which is loaded to failure in a and thus closer to these shear zones, may also explain the
homogenous soil profile. All of these factors increase the enhanced settlement performance.
likelihood of bulging failure, which may have guided the
author's interpretation of a critical length. 6.2.2. Column stiffness
The influence of column stiffness upon the settlement perfor-
mance of various configurations of columns (see Fig. 4) is shown
6.2. Parametric study
in Fig. 14. As expected, an increase in n values is observed with
increasing column stiffness. This is consistent with the work of
6.2.1. Column position
Domingues et al. (2007a) who conducted a FEA on a large group
The influence of column position relative to the footing
of columns (A/AC ¼ 5.3) supporting an embankment. The increase
edge (see Fig. 3) upon n values is shown in Fig. 13. It
in n values with column stiffness is most pronounced for columns
appears that small benefits can be gained by positioning
at low area ratios, as the degree of plasticity within columns is
columns closer to the footing edge, confirming the merit of
reduced due to the enhanced column confinement. Consequently,
this strategy which is sometimes used in practice. This may
columns at low area ratios are more susceptible to changes in
be attributed to stress concentrations which develop beneath
elastic stiffness moduli.
the edges of rigid footings. Columns positioned closer to
this zone have the potential to absorb a larger proportion of
the applied load and, hence, develop higher n values. Wehr 6.2.3. Column friction angle
0
(2004) conducted a FEA on small groups of stone columns The influence of the angle of internal friction ðϕcol Þ upon the
and observed the development of shear zones which extend settlement performance of various configurations of stone
from the edges of footings to a depth beneath the centre of columns (see Fig. 4) is shown in Fig. 15. The angle of internal
footings. Positioning columns closer to the footing edge, friction controls the plastic deformation of columns, i.e.
 M.M. Killeen, B.A. McCabe / Soils and Foundations 54 (2014) 760–776
deformation which occurs post-yielding. Increasing ϕcol from
the default value of 451 to 501 has the lowest influence on the
group of nine columns, as columns at low area ratios exhibit the
0
lowest degree of plasticity. Decreasing ϕcol from the default
value of 45–401 is most pronounced for columns at low area
ratios, as the groups of four and five columns (at higher area
ratios) are already in an advanced state of plasticity.
6.2.4. Coefficient of lateral earth pressure
Fig. 16 shows the influence of K0 for the lower Carse clay
layer upon the settlement performance of various configurations
of columns outlined in Fig. 4. On the basis of previous research
outlined in Section 2.2, the in situ stress state of the upper Carse
clay and crust is deemed sufficiently high (K0 Z1.0) and has not
been modified. An increase in K0 generally has a positive
influence on n values for all configurations of columns. The
influence of K0 upon n values is most pronounced for groups of
four and five columns, as columns at high area ratios tend to
exhibit bulging and, hence, are dependent upon the lateral
support provided by the surrounding soil. In contrast, columns
at low area ratios transfer the applied load to the base of
columns by punching and are less dependent upon K0 of the
surrounding soil.
Settlement improvement factor, n
1.0 2.0 3.0 4.0
0 stiff upper layers. Fig. 19 highlights the importance of the stiff
4 cols - s = 2.0 m
5 cols - s = 1.0 m
2
9 cols - s = 1.0 m
Column length, L (m)
K0,soil = 0.75
4 K0,soil = 1.00
K0,soil = 1.25
6 ment (Karstunen and Yin, 2010)) and should be taken into
8
10
12
14
Fig. 16. Influence of the coefficient of lateral earth pressure upon settlement
improvement factors.
Fig. 17. Soil profile adopted to investigate influence of lower Carse clay (Profile 2) and stiff crust (Profile 3).
771
0
6.2.5. Presence of stiff crust and lower Carse clay layer
The influence of the lower Carse clay and stiff crust upon
the settlement performance of small groups of stone columns
(see Fig. 4) is analysed by adjusting the standard soil profile in
two steps, as illustrated in Fig. 17. The importance of soil
stiffness variation with depth is clearly evident in Fig. 18 as n
values are lowest for Profile 1, which contains the weak lower
Carse clay (not present in Profiles 2 and 3). A sharp increase in
n values occurs with increasing column length up to L ¼ 3 m
for Profile 3. However, the rate of increase in n values with
column length reduces thereafter. This is most pronounced for
smaller groups of columns (i.e. higher area ratios) and suggests
that a critical length may exist for columns formed in Profile 3.
The distribution of vertical strain along the length of a 2  2
group of 4 m long columns is shown in Fig. 19. A comparison
of Profiles 1 and 2 with Profile 3 reveals that column bulging
becomes shallower and more pronounced with the absence of a
stiff crust. It can also be seen that the magnitude of vertical
strain in the underlying soil (i.e. below a depth of 4 m) is
higher for Profile 1 relative to Profiles 2 and 3, which indicates
the increased potential for column punching in weaker soil
strata. Profile 3 is a homogenous soil profile and is similar to
that used in laboratory studies. The low lateral support
provided by the upper layers, and subsequent bulging, may
explain why authors (Muir Wood et al., 2000; McKelvey et al.,
2004; Black et al., 2011) have postulated critical lengths in the
past. However, it can be seen that critical lengths are less well
5.0 defined for more realistic soft soil profiles which incorporate
crust upon the settlement performance of stone columns. This
is a salient feature of many soft soil profiles (e.g. Väsby and
Skå-Edeby test fill areas (Larsson, 1986); Ellingsrud test fill
area (Karlsrud and Haugen, 1979); and Murro test embank-
account when designing stone columns in soft soils.
7. Development of a simplified design framework
It was shown in the previous section that column confine-
ment has a significant influence upon the settlement perfor-
mance of small groups of stone columns. The degree of
column confinement depends upon many factors such as area
ratio, the number of columns and column length. The influence
772 M.M. Killeen, B.A. McCabe / Soils and Foundations 54 (2014) 760–776

of the number of columns upon the settlement performance of Settlement improvement factor, n
small groups of stone columns is examined using a settlement 1.0 1.5 2.0 2.5 3.0 3.5
ratio which is defined as the ratio of the settlement of a small 0 Profile: 1 2 3

group (s) to that of an infinite grid (suc) of columns. Crust

UCC
Crust
UCC
UCC
2 LCC

Column length, L (m)

4
7.1. Previous research using settlement ratios removing Crust
6 removing LCC
7.1.1. Numerical methods
Elshazly et al. (2008b) conducted a series of axisymmetric 8
FEA to examine the settlement performance of small groups of
10
end-bearing stone columns. The authors use s/suc to examine
4 cols - 2.00 m
the trade-off between the decay of vertical stress with depth 12 Profile 1
Profile 2
and loss of lateral confinement associated with small groups of Profile 3
stone columns. For small groups of stone columns, the decay 14
of vertical stress with depth dominates due to the small footing
widths and, consequently, s/suc o 1. However, as the number Settlement improvement factor, n
of columns increases, the distribution of vertical stress with 1.0 1.5 2.0 2.5 3.0 3.5
0
depth beneath footings becomes closer to that for unit cell Profile: 1 2 3
Crust Crust UCC
conditions and the influence of the loss of lateral confinement 2 UCC UCC
LCC
becomes more prevalent. This yields s/suc 41 which indicates

Column length, L (m)

that settlement design methods based on the unit cell concept 4
are unconservative. These ratios are difficult to justify; the high removing Crust
6
s/suc values may be due in part to the flexible loading condition
and weak surrounding soil (i.e. low lateral support). 8
removing LCC

10
7.1.2. Analytical methods 5 cols - 1.00 m
Priebe (1995) presents a design chart which relates s/suc to 12 Profile 1
Profile 2
normalised column length (L/d). The s/suc values are devel- Profile 3
oped from numerous calculations which consider the reduction 14
of vertical stress with depth beneath small loaded areas and, Settlement improvement factor, n
also, a lower bearing capacity for the outer row of columns in a 1.0 2.0 3.0 4.0 5.0
group. No direct reference is made to footing area (A) in the 0 Profile: 1 2 3
design chart and it appears that for a given number of columns Crust Crust UCC
UCC UCC
s/suc does not depend upon area ratio. However, an increase in 2 LCC

area ratio results in higher settlements for small groups and

Column length, L (m)

4
infinite grids of columns. Priebe (1995) claims that the increase
in ‘s’ and ‘suc’ compensate such that s/suc is acceptable for 6 removing
A/AC o 10. Crust
8
removing LCC
7.2. Comparison of s/suc from Priebe (1995) and PLAXIS 3D 10
foundation for end-bearing columns 9 cols - 1.00 m
12 Profile 1
Profile 2
Settlement ratios determined from PLAXIS 3D Foundation Profile 3
14
and those of Priebe (1995) are shown in Fig. 20. It can be seen
Fig. 18. Influence of the lower Carse clay and the stiff crust upon settlement
that s/suc increases for larger groups of columns. An increase in
improvement factors for (a) four, (b) five and (c) nine columns beneath a 3 m
the number of columns leads to higher levels of lateral square footing.
confinement and also higher levels of vertical stress with
depth (due to larger footing widths). Therefore, the boundary
conditions approach unit cell conditions with an increasing 7.3. Settlement ratios for floating stone columns
number of columns and s/suc tends towards unity. It can also
be seen that s/suc increases with area ratio. The design curve Currently no study has investigated the variation of s/suc for
developed by Priebe (1995) captures the variation of s/suc quite floating stone columns. The effect of floating columns is
well for A/AC ¼ 3.5, but diverges at high area ratios. This introduced into Fig. 21 using the L/H term, where H is the
confirms that Priebe's (1995) method is only applicable for low thickness of the soil deposit (H ¼ 13.9 m for the Bothkennar
area ratios and is unconservative for widely spaced columns. deposit in this study). The influence of area ratio upon s/suc for
 M.M. Killeen, B.A. McCabe / Soils and Foundations 54 (2014) 760–776 773

floating columns is also shown in Fig. 20. For clarity, only Vertical strain, εy (10-3)
data specific to three column lengths are presented, namely: 0 10 20 30
L/H ¼ 0.22, 0.58 and 1.00, which correspond to L ¼ 3, 8 and 0 Bulging becomes shallower
13.9 m, respectively. 1 and more pronounced with

Depth beneath footing base (m)

3
2 absence of stiff crust
2
For a given B/L, it can be seen that s/suc reduces with
COLUMN
decreasing L/H. A decrease in L/H corresponds to a thicker soil 4
2 + 3

deposit beneath the base of columns. The settlement of the soil SOIL
1 Increased punching in
beneath the base of columns is much higher for an infinite grid 6 weaker deposits at depth
of columns compared to small groups due to the constant
increment of vertical stress with depth. Therefore, a decrease in 8
L/H leads to higher suc and, consequently, lower s/suc. 10 Soil Profiles:
1 2 3
It appears that area ratios only influence s/suc for end- Crust Crust UCC 4 cols - 2.00 m
bearing columns (L/H ¼ 1.0), which is due to the large 12 UCC
LCC
UCC Profile 1
Profile 2
reduction in suc for an infinite grid of long closely-spaced Profile 3
columns (Fig. 10(i)). In contrast, it can be seen that area ratios
do not influence s/suc for floating columns. For a given L/H, it Fig. 19. Distribution of vertical strain within and directly beneath 4 m long
columns, which are part of a 2  2 group in Profiles 1, 2 and 3.
appears that s/suc does not depend on the configuration of
columns and is directly proportional to the footing width (i.e. B/L).
The variation of s/suc for additional column lengths is shown 0.5

in Fig. 22. A somewhat linear relationship appears to exist

between s/suc and B/L for floating stone columns. However, 0.4

Settlement ratio, s/suc

care must be exercised in assuming a linear relationship for
0.3
end-bearing columns as s/suc from Priebe (1995) approach
unity somewhat asymptotically at high B/L. The linear 0.2
Priebe (1995)

relationship between s/suc and B/L for floating stone columns

is justified by the high coefficients of determination (R2). 0.1
PLAXIS 3D Foundation
A/Ac=3.5
Assuming a linear relationship between s/suc and B/L, the A/Ac=8.0
A/Ac=14.1
generalised relationship may be defined by: 0.0

 0 5 10 15 20
s B
¼α ð7Þ No. of columns
suc L
Fig. 20. Comparison of s/suc for groups of end-bearing columns from PLAXIS
The variation of α with L/H is shown in Fig. 23 and can be 3D Foundation with Priebe (1995).
captured by a quadratic equation:

2
 1.0
L L PLAXIS 3D Foundation
α ¼ 0:61 þ 0:1 þ 0:06 ð8Þ A/Ac = 3.5
H H 0.8
A/Ac = 8.0
A/Ac = 14.1
Settlement ratio, s/suc

Priebe (1995)
Therefore, it is possible to relate s/suc to B/L and L/H by the A/Ac = 3.5
0.6
following equation:
"
2
 #
 L/H = 1.00
s L L B 0.4
¼ 0:61 þ 0:1 þ 0:06 U ð9Þ L/H = 0.58
L/H = 0.22
suc H H L
0.2

The relevant value of suc in Eq. (9) will depend on whether

the columns are floating or end-bearing. However, analy- 0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
tical design methods have only been developed for the end- Normalised footing width, B/L
bearing scenario (suc,eb). Therefore, the dependence of the
ratio suc,eb/suc on L/H has been investigated from the FEA, Fig. 21. Variation of s/suc with normalised footing width and column length.
with a view to rendering Eq. (9) amenable for practical use.
Suitable curve-fitted relationships are shown in Fig. 24 in the
form of Eq. (10): The settlement (s) of a group of stone columns can then be

 determined by combining Eqs. (9)–(11) as follows:
L suc; eb
¼ β ln þ1 ð10Þ
H suc
s ½0:61ðL=HÞ2 þ 0:1ðL=HÞþ 0:06 UðB=LÞ
¼ ð12Þ
The coefficients β are shown in Fig. 25 to vary approxi- suc; eb e8:71ððL=HÞ  1ÞðAC =AÞ
mately linearly with A/AC, as given by the following equation:

 It is recommended that either Pulko et al. (2011) or Castro
A
β ¼ 0:1148 ð11Þ and Sagaseta (2009), rather than Priebe (1995), is used as the
AC basis of estimating suc,eb in Eq. (12).
774 M.M. Killeen, B.A. McCabe / Soils and Foundations 54 (2014) 760–776

1.0 Area ratio, A/AC

0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0
Normalised settlement, s/suc

0.8 0.0

0.2
β = 0.1148(A/Ac)
0.6

β = (L/H - 1) / ln(suc,eb/suc)
0.4

0.4
0.6

0.8
0.2
1.0

1.2
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 1.4
Normalised footing width, B/L
1.6
Fig. 22. Influence of normalised footing width and column length upon s/suc 1.8
for groups of columns.
Fig. 25. Relationship between β (which relates suc,eb/suc and L/H) with A/AC.

1.0
While the settlement of a large group of columns was not
measured at the same site, it was possible to estimate suc,eb ¼
0.8
47 mm using Pulko et al. (2011). This results in s/suc,eb ¼ 0.53,
α = 0.61(L/H) + 0.1(L/H) + 0.06
which is higher than s/suc,eb ¼ 0.29 determined using Eq. (12)
α = (s/suc) / (B/L)

R =1
0.6
(i.e. for L/H ¼ 0.45, B/L ¼ 0.33 and A/AC ¼ 3.58). The high
s/suc,eb values observed for Kirsch (2004) may be due in part to
0.4
undrained response of the surrounding soil as a result of the
short duration of the load test; the long term settlement of an
0.2
infinite grid (suc,eb) is much higher than a small group
of columns (s) as the soil is stressed to the full depth of the
0.0
0.0 0.2 0.4 0.6 0.8 1.0 1.2 soil profile and, hence, will develop larger consolidation
Normalised column length, L/H settlement.
The availability of further data will allow the form of
Fig. 23. Relationship between α (which relates s/suc and B/L) with L/H.
Eq. (12) to be appraised more closely and modified as appro-
priate. However, this approach provides a useful framework
suc,eb / suc through which the behaviour of small groups of columns can
0.0 0.2 0.4 0.6 0.8 1.0 be distinguished from larger groups.
0.0
Normalissed column length, L/H

8. Conclusions
0.2

The settlement performance of various configurations of

0.4
columns beneath rigid square pad footings in the Bothkennar
soil profile is examined in this study.
0.6

8.1. Influence of column confinement

0.8

 The influences of area ratio, column length and the number

1.0
of columns upon the settlement performance of stone
Fig. 24. Relationship between suc,eb/suc with normalised column length (L/H). columns are somewhat inter-related, as the influence of
column length becomes more pronounced at low area ratios.
7.4. Comparison of simplified design framework  The unit cell model does not capture the reduction of
with field data vertical stress with depth beneath small loaded areas and,
hence, under-predicts n values for floating columns shorter
There is a shortage of field data with which the validity of than L≈7 m, corresponding to L/H ¼ 0.5 in this study.
Eq. (12) could be assessed – only one partially-suitable case
history is known to the authors. Kirsch (2004) reports on a
load test on a group of 5 columns beneath a 3 m square footing 8.2. Influence of other parameters
in Kuala Lumpur, Malaysia. Columns were installed to 9 m in
a 20 m deep soil deposit, which primarily consisted of soft clay  For a given number of columns, their position relative to the
underlain by sandy silt. The footing was loaded to 118 kPa in footing edge has a minor influence upon the settlement
under 12 days, and 25 mm settlement was observed at 50 kPa. performance.
 M.M. Killeen, B.A. McCabe / Soils and Foundations 54 (2014) 760–776
 Increasing the column stiffness and strength increases
n values. The influence of these parameters is related to
the degree of plasticity within columns, as increasing the
stiffness and strength is most significant at low and high
area ratios, respectively.
 Accounting for column installation effects by increasing the
coefficient of lateral earth pressure at rest (K0) in the lower
Carse clay yields higher n values. The influence of K0 upon
the settlement performance of stone columns appears to be
dependent upon the mode of deformation and is most
pronounced at high area ratios, i.e. column bulging.
 The presence of the stiff crust has a significant influence
upon the settlement performance of stone columns. The stiff
crust confines columns in the upper layers and forces
bulging to occur deeper, in weaker layers. Conclusions
about critical column length quoted in previous studies
appear to be specific to ultimate conditions and due to the
unrealistic simplified ground profile modelled rather than
mechanisms particular to stone column behaviour.
8.3. Simplified design framework
A simplified design framework is proposed in which the
settlement (s) of either floating or end-bearing small groups
can be estimated from the settlement of infinite grids of end
bearing columns (suc,eb) and geometrical factors such as area
ratio (A/AC), normalised footing width (B/L) and normalised
column length (L/H).
Acknowledgements
The authors wish to acknowledge the support of the Irish
Research Council for Science, Engineering and Technology
EMBARK Initiative for funding the first author’s research.
The authors are also extremely grateful for the input of Keller
Foundations (UK) and Mr.Brian Sexton (NUI Galway) to
this study.
References
Allman, M.A., Atkinson, J.H., 1992. Mechanical properties of reconstituted
Bothkennar soil. Géotechnique 42 (2), 289–301.
Balaam, N.P., Booker, J.R., 1981. Analysis of rigid rafts supported by granular
piles. Int. J. Numer. Anal. Methods Geomech. 5, 379–403.
Balaam, N.P., Booker, J.R., 1985. Effect of stone column yield on settlement
of rigid foundations in stabilised clay. Int. J. Numer. Anal. Methods
Geomech. 9, 331–351.
Barksdale, R.D., Bachus, R.C., 1983. Design and Construction of Stone
Columns Report No. FHWA/RD-83/026. National Information Service,
Springfield, Virginia, USA.
Black, J.A., Sivakumar, V., Bell, A., 2011. The settlement performance of
stone column foundations. Géotechnique 61 (11), 909–922.
Boussinesq, J., 1885. Applications des Potentiels à L'étude de L'équilibre et du
Mouvement des Solides Élastiques, Gauthier-Villars, Paris.
Brinkgreve, R.B.J., Broere, W., 2006. PLAXIS 3D Foundation Manual
Version 2. PLAXIS BV.
Castro, J., 2007. Pore pressures during stone column installation. In: Proceed-
ings of the18th European Young Geotechnical Engineers' Conference,
Ancona.
775
Castro, J., Sagaseta, C., 2009. Consolidation around stone columns. Influence
of column deformation. Int. J. Numer. Anal. Methods Geomech. 33 (7),
851–877.
Domingues, T.S., Borges, J.L., Cardoso, A.S., 2007. Parametric study of stone
columns in embankments on soft soils by finite element method. In: Proceedings
of the 5th International Workshop on Applications of Computational Mechanics
in Geotechnical Engineering V, Guimaraes, pp. 281–291.
Duncan, J.M., Chang, C.Y., 1970. Nonlinear analysis of stress and strain in
soil. ASCE J. Soil Mech. Found. Div. 96 (5), 1629–1653.
Elshazly, H., Elkasabgy, M., Elleboudy, A., 2008a. Effect of inter-column
spacing on soil stresses due to vibro-installed stone columns: interesting
findings. J. Geotech. Geol. Eng. 26 (2), 225–236.
Elshazly, H.A., Hafez, D.H., Mossaad, M.E., 2008b. Reliability of conven-
tional settlement evaluation for circular foundations on stone columns. J.
Geotech. Geol. Eng. 26 (3), 323–334.
Gäb, M., Schweiger, H.F., Thurner, R., Adam, D., 2007. Field trial to
investigate the performance of floating stone columns. In: Proceedings of
the 14th European Conference on Soil Mechanics and Geotechnical
Engineering, Madrid, pp. 1311–1316.
Gäb, M., Schweiger, H.F., Kamrat-Pietraszewska, D., Karstunen, M., 2008.
Numerical analysis of a floating stone column foundation using different
constitutive models. In: Proceedings of the 2nd International Workshop on
Geotechnics of Soft Soils, Glasgow, pp. 137–142.
Hight, D.W., Bond, A.J., Legge, J.D., 1992. Characterization of the Bothken-
nar clay: an overview. Géotechnique 42 (2), 303–347.
Hughes, J.M., Withers, N.J., 1974. Reinforcing of soft cohesive soils with
stone columns. Ground Eng. 7 (3), 42–49.
Han, J., Ye, S.L., 2001. Simplified method for consolidation rate of stone
column reinforced foundations. J. Geotech. Geoenviron. Eng. ASCE
127 (7), 597–603.
Indraratna, B., Basack, S., Rujikiatkamjorn, C., 2013. Numerical solution of
stone column-improved soft soil considering arching, clogging, and smear
effects. J. Geotech. Geoenviron. Eng. 139 (3), 377–394.
Jardine, R.J., Lehane, B.M., Smith, P.R., Gildea, P.A., 1995. Vertical loading
experiments on rigid pad foundations at Bothkennar. Géotechnique 45 (4),
573–597.
Killeen, M.M., 2012. Numerical modelling of small groups of stone columns
(Ph.D. thesis). National University of Ireland, Galway.
Kirsch, F., 2004. Experimentelle und numerische Untersuchungen zum
Tragverthalten von Rüttelstopfsäulengruppen (Dissertation). Technische
Universität Braunschweig.
Kirsch, F., 2006. Vibro stone column installation and its effect on the ground
improvement. In: Proceedings of the International Conference on Numer-
ical Simulation of Construction Processes in Geotechnical Engineering for
Urban Environment, Bochum, pp. 115–124.
Kirsch, F. 2008. Evaluation of ground improvement by groups of vibro stone
columns using field measurements and numerical analysis. In: Proceedings
of the 2nd International Workshop on Geotechnics of Soft Soils, Glasgow,
pp. 241–248.
Karlsrud, K., Haugen, T., 1979. Lang-tids belastningsforsøk på kvikkleire.
Nordic Geotechnical Meeting, 609–619 (Helsinki).
Karstunen, M., Yin, Z.-Y., 2010. Modelling time-dependent behaviour of
Murro test embankment. Géotechnique 60 (10), 735–749.
Larsson, R., 1986.Consolidation of soft clays. Report No. 29, Swedish
Geotechnical Institute, Linköping, 174p.
McCabe, B.A., Nimmons, G.J., Egan, D., 2009. A review of field performance
of stone columns in soft soils. Proc. Inst. Civil Eng. – Geotech. Eng. 162
(6), 323–334.
McKelvey, D., Sivakumar, V., Bell, A.L., Graham, J., 2004. Modelling
vibrated stone columns in soft clay. Proc. Inst. Civil Eng. – Geotech.
Eng. 157 (3), 137–149.
Mitchell, J.K., Huber, T.R., 1985. Performance of a stone column foundation.
J. Geotech. Eng., ASCE 11 (2), 205–223.
Muir Wood, D., Hu, W., Nash, D.F.T., 2000. Group effects in stone column
foundations: model tests. Géotechnique 50 (6), 689–698.
Nash, D.F.T., Sills, G.C., Davison, L.R., 1992a. One-dimensional consolida-
tion testing of the soft clay from Bothkennar. Géotechnique 42 (2),
241–256.
776 M.M. Killeen, B.A. McCabe / Soils and Foundations 54 (2014) 760–776
Nash, D.F.T., Powell, J.J.M., Lloyd, I.M., 1992b. Initial investigations of the
soft clay test site at Bothkennar. Géotechnique 42 (2), 163–181.
Paul, M.A., Peacock, J.D., Wood, B.F., 1992. The engineering geology of the
Carse clay of the national soft clay research site, Bothkennar. Géotechnique
42 (2), 183–198.
Priebe, H.J., 1976. Evaluation of the settlement reduction of a foundation
improved by vibro-replacement. Bautechnik 5, 160–162.
Priebe, H.J., 1995. The design of vibro replacement. Ground Eng. 28 (10),
31–37.
Pulko, B., Majes, B., Logar, J., 2011. Geosynthetic-encased stone columns:
analytical calculation model. Geotext. Geomembr. 29 (1), 29–39.
Rowe, P.W., 1962. The stress-dilatancy relation for static equilibrium
of an assembly of particles in contact. Proc. R. Soc. A 269 (1339),
500–527.
Schanz, T., Vermeer, P.A., Bonnier, P.G., 1999. The hardening-soil model:
Formulation and verification. In: Brinkgreve, R.B.J. (Ed.), Beyond 2000 in
Computational Geotechnics – 10 Years of PLAXIS International.
A.A. Balkema, Rotterdam, Netherlands, pp. 281–290.
Serridge, C.J., Sarsby, R.W., 2008. A review of field trials investigating the
performance of partial depth vibro stone columns in a deep soft clay
deposit. In: Proceedings of the 2nd International Workshop on Geotechnics
of Soft Soils, Glasgow, pp. 293–298.
Sexton, B.G., McCabe, B.A., Castro, J., 2013. Appraising vibro settlement
prediction methods using the finite element method. Acta Geotech. (minor
corrections sought Jan 2013).
Shahu, J.T., Reddy, Y.R., 2011. Clayey soil reinforced with stone column
group: model tests and analyses. J. Geotech. Eng., ASCE 137 (12),
1265–1274.
Sivakumar, V., Jeludine, D.K.N.M., Bell, A., Glynn, D.T., Mackinnon, P.,
2011. The pressure distribution along stone columns in soft clay under
consolidation and foundation loading. Géotechnique 61 (7), 613–620.
Wehr, W., 2004. Stone columns – single columns and group behavior. In:
Proceedings of the 5th International Conference on Ground Improvement
Techniques, Malaysia, pp. 329–340.
Weber, T.M., Springman, S.M., Gäb, M., Racansky, V., Schweiger, H.F.,
2008.Numerical modeling of stone columns in soft clay under an
embankment. In: Proceedings of the 2nd International Workshop on
Geotechnics of Soft Soils, Glasgow, pp. 305–311.