materials

Review
Modeling Stone Columns
Jorge Castro
Department of Ground Engineering and Materials Science, University of Cantabria, Avda. de los Castros s/n,
39005 Santander, Spain; castrogj@unican.es; Tel.: +34-942-201813

Received: 4 May 2017; Accepted: 7 July 2017; Published: 11 July 2017

Abstract: This paper reviews the main modeling techniques for stone columns, both ordinary stone
columns and geosynthetic-encased stone columns. The paper tries to encompass the more recent
advances and recommendations in the topic. Regarding the geometrical model, the main options
are the unit cell, longitudinal gravel trenches in plane strain conditions, cylindrical rings of
gravel in axial symmetry conditions, equivalent homogeneous soil with improved properties and
three-dimensional models, either a full three-dimensional model or just a three-dimensional row or
slice of columns. Some guidelines for obtaining these simplified geometrical models are provided
and the particular case of groups of columns under footings is also analyzed. For the latter case,
there is a column critical length that is around twice the footing width for non-encased columns in a
homogeneous soft soil. In the literature, the column critical length is sometimes given as a function
of the column length, which leads to some disparities in its value. Here it is shown that the column
critical length mainly depends on the footing dimensions. Some other features related with column
modeling are also briefly presented, such as the influence of column installation. Finally, some
guidance and recommendations are provided on parameter selection for the study of stone columns.

Keywords: stone columns; encased stone columns; geosynthetic; numerical modeling; critical length;
parameter selection

1. Introduction
Ground improvement using stone columns, also known as granular piles or aggregate piers,
is one of the most popular techniques to improve soft soils for the foundation of embankments or
structures. These are vertical boreholes in the ground, filled upwards with gravel compacted by means
of a vibrator.
The idea of improving soft soils for foundation purposes using granular inclusions is relatively
old. It is documented [1] that in 1839 in Bayonne (France), the French colonel Burbach used for the
first time sand piles as deep foundations instead of the classical wood piles that rapidly degrade with
fluctuations of the ground water level. However, it was not until the 50 s of the last century when stone
columns started to be used. The ground improvement technique started as an extension of traditional
vibro-compaction (deep compaction) to non-granular soils, whose low permeability and cohesion do
not allow for a quick rearranging of soil particles in a denser configuration.
Stone columns act mainly as inclusions with a higher stiffness, shear strength and permeability
than the natural soil. Consequently, they improve the following aspects:

The bearing capacity

The stability of embankments and natural slopes
Final settlement
Degree of consolidation
Liquefaction potential

Materials 2017, 10, 782; doi:10.3390/ma10070782 www.mdpi.com/journal/materials

Materials 2017, 10, 782 2 of 23

The reduction of the liquefaction potential is beyond the scope of the present paper. Please refer
to [2,3] for further information on that topic. The present paper focuses on the other four improvements,
particularly, the settlement reduction. The modeling strategy for stone columns should be chosen
depending on which of the previous improvements is to be analyzed.
In extremely soft soils, stone columns are not suitable because their continuity, stability, geometric
shape, etc. cannot be guaranteed. The undrained shear strength (cu ) of natural soft soil is generally
used as the limiting parameter for stone column feasibility. A limiting value around 515 kPa [4] may
be adopted. To improve the lateral confinement of stone columns in those extremely soft soils, encasing
the columns with geotextiles or other geosynthetics, such as geogrids, has proven to be a successful
technique in recent times (e.g., [5]). Rigid or semi-rigid inclusions (e.g., adding lime or cement) is
another common alternative (e.g., [6,7]). Han [8] summarizes different ground column technologies.
Ground improvement using stone columns, either ordinary or encased stone columns, requires a
considerable amount of columns or, at least, a group of columns. This implies a complex modeling
process of the real geometry. This paper provides some guidelines for obtaining these simplified
geometrical models. Some other features related with column modeling are also briefly presented,
such as the critical length of the column and the influence of column installation. Additionally, some
guidance and recommendations are provided on parameter selection to study stone columns. Here,
the word modeling is understood in a broad sense, covering geometrical, mechanical, geotechnical
and installation features of stone columns.
The increase in computer power and the availability of finite element codes makes numerical
analyses very appealing in geotechnical design. They usually lead to more detailed studies but
require a clear conception of the modeling techniques. Ground improvement techniques, such as
stone columns, are more and more popular due to the increasing occupation of natural soft soils and
environmental concerns [9,10]. Within these current trends, the review of the modeling techniques
for stone columns seems interesting and useful. Besides, in some instances, there is some confusion
about geometrical models and their application. For example, the results for an isolated column
under concentrated load just on top of the column cannot be directly extrapolated for a large group of
columns under distributed uniform load.

2. Geometrical Models
To simplify the real geometry of the problem that usually involves a considerable amount of
columns (e.g., Figure 1a) and to be able to deal with the problem, the following simplified geometrical
models are usually adopted:

Unit cell in axial symmetry (Figure 1b). Only one column and its corresponding surrounding
soil are studied. It may be useful to study just a horizontal slice of the unit cell, rather than the
whole length.
Longitudinal gravel trenches (Figure 1c). The stone columns are transformed into longitudinal
gravel trenches to study the problem in plane strain conditions.
Cylindrical rings of gravel (Figure 1d). The columns are transformed into cylindrical rings of
gravel to study the problem in axial symmetry.
Homogenization or equivalent homogeneous soil (Figure 1e). The columns and the surrounding
soil are transformed into a homogeneous soil with equivalent improved properties.
Three-dimensional (3D) model of a row or slice of columns (Figure 1f).
Geometrical models for small groups of stone columns.
Unit cell with constant lateral pressure (triaxial conditions).
Isolated column.
Materials 2017, 10, 782 3 of 23
Materials 2017, 10, 782 3 of 23

(a) (b) (c)

(d) (e) (f)

Figure 1. Main geometrical models for stone column studies: (a) Full 3D model; (b) Unit cell;
Figure 1. Main geometrical models for stone column studies: (a) Full 3D model; (b) Unit cell; (c)
(c) Longitudinal gravel trenches; (d) Cylindrical gravel rings; (e) Equivalent homogenous soil; (f) 3D slice
Longitudinal gravel trenches; (d) Cylindrical gravel rings; (e) Equivalent homogenous soil; (f) 3D slice
of columns.
of columns.
The two latter cases, namely unit cell under triaxial conditions and isolated column, do not
The twoappear
usually latterincases,
real problems, but they
namely unit have
cell been used
under for conditions
triaxial laboratory tests
and(e.g., [1113]).
isolated The case
column, do not
of an isolated column is widely used for field tests (usually plate load tests) for the sake
usually appear in real problems, but they have been used for laboratory tests (e.g., [1113]). The case of simplicity
(e.g., [14]).
of an isolated column is widely used for field tests (usually plate load tests) for the sake of simplicity
As a simple introductory comparison between the different geometrical models, Table 1
(e.g., [14]).
summarizes the suitability of some of these models to study the improvements achieved with a stone
As a simple
column introductory
treatment comparison
for the foundation between the different geometrical models, Table 1
of an embankment.
summarizes the suitability of some of these models to study the improvements achieved with a
stone column
Tabletreatment for
1. Suitability of the foundation
simplified of anmodels
geometrical embankment.
to study different features of a stone column
treatment for the foundation of an embankment.
Table 1. Suitability of simplified geometrical models to study different features of a stone column
Geometrical Model Final Settlement Consolidation Stability
treatment for the foundation of an embankment.
Unit cell *** *** -
Gravel trenches ** ** **
Geometrical Model Final Settlement Consolidation Stability
Homogenization ** * *
Unit cell 3D slice *** *** *** *** *** -
Gravel trenches ** **
*** Completely suitable, ** Moderately suitable, * Slightly suitable, - Not suitable.
**
Homogenization ** * *
3D slice *** *** ***
All the geometrical models listed above are valid for non-encased stone columns. However, for
encased stone*** Completely
columns, suitable,
there is not**yet
Moderately suitable, * way
any satisfactory Slightly suitable, -the
to convert Notcylindrical
suitable. encasement
(geosynthetic) that surrounds the column for the cases of longitudinal gravel trenches and cylindrical
ringsthe
All of gravel.
geometrical models listed above are valid for non-encased stone columns. However,
for encased stone columns, there is not yet any satisfactory way to convert the cylindrical encasement
3. Unit Cell
(geosynthetic) that surrounds the column for the cases of longitudinal gravel trenches and cylindrical
The unit cell model is the most widely used for theoretical analyses and it is reviewed in detail,
rings of gravel.
for example, in [15]. The basis for the simplified model is the usage of a great number of columns,
3. Unit Cell distributed in a wide area under a uniform load. This is the case, for example, in the central
uniformly
part of an embankment on soft ground improved with stone columns. In these situations, the
The unit cell model is the most widely used for theoretical analyses and it is reviewed in detail,
behavior of all the columns is the same and then, it is enough to study the behavior of just one column
for example, in [15]. The basis
with the corresponding for thesoil
surrounding simplified
(tributarymodel is the
area). Due usage of a
to symmetry great number
conditions, of columns,
at the external
uniformly
lateraldistributed
boundary ofinthea wide
unit area under
cell, only a uniform
vertical load. Thisand
displacements is the
onlycase, for example,
vertical in theare
water seepage central
part of an embankment on soft ground improved with stone columns. In these situations, the behavior
allowed.
of all the columns is the same
Stone columns and then,
are generally it is enough
uniformly to study
distributed in the behavior
triangular or of just one
square column
grids. Thus, with
the the
tributary area of natural soil for each column is a hexagon or a square, respectively.
corresponding surrounding soil (tributary area). Due to symmetry conditions, at the external lateral To allow for axial
symmetry
boundary of theconditions,
unit cell, the
onlytributary
verticalarea is transformed
displacements andinto
only a circle
vertical(cylinder) of the same
water seepage are (cross-
allowed.
sectional) area. Therefore, the diameter of the unit cell is equal to
Stone columns are generally uniformly distributed in triangular or square grids. = 1.05 1.13 for triangular
Thus, the
and square grids, respectively, where is the centre-to-centre distance between columns (Figure 2).
tributary area of natural soil for each column is a hexagon or a square, respectively. To allow for
Figure 3 shows the unit cell model with axial symmetry that can be studied in two dimensions.
axial symmetry conditions, the tributary area is transformed into a circle (cylinder) of the same
(cross-sectional) area. Therefore, the diameter of the unit cell is equal to de = 1.05 1.13 s for triangular
and square grids, respectively, where s is the centre-to-centre distance between columns (Figure 2).
Figure 3 shows the unit cell model with axial symmetry that can be studied in two dimensions.
Materials 2017, 10, 782 4 of 23
Materials 2017, 10, 782 4 of 23
Materials 2017, 10, 782 4 of 23

s re
re rc
s rc re
re rc
rc s
s

Figure 2. Simplification of the unit cell to axial symmetry conditions.

Figure 2. Simplification of the unit cell to axial symmetry conditions.
Figure 2. Simplification of the unit cell to axial symmetry conditions.

Figure 3. Unit cell model for analytical solutions.

Figure3.3.Unit
Figure Unit cell
cell model
modelfor foranalytical
analyticalsolutions.
solutions.
The unit cell model is used for most of the existing analytical solutions (e.g., [16,17]). In the
analytical
The unit solutions,
cell model thereisisused usually for amostfurther
of thesimplifying assumption
existing analytical for the geometry,
solutions which
(e.g., [16,17]). In istheto
The unit cell model is used for most of the existing analytical solutions (e.g., [16,17]). In the
assume that
analytical the behavior
solutions, there of is each
usually horizontal
a further slice is independent,
simplifying i.e., the
assumption forshear stresses arewhich
the geometry, neglected.
is to
analytical
assume
solutions,
Numerical that analyses
there
the behavior
isofusually
[17,18] show
each
a further
that
horizontal shear simplifying
slicestresses assumption
are usually
is independent,
for the
i.e.,negligible for geometry,
the shear stresses distributed which
loads. is to
are neglected.
assume that the behavior
Balaam andanalyses
Numerical Booker [17,18] of each
[19] is ashow horizontal
notable thatexception slice
shear stresses is independent,
to theare simplifying i.e., the
assumption
usually negligible shear
for ofstresses are
neglectingloads.
distributed neglected.
shear
Numerical
Balaam analyses
stressesandbecause
Booker[17,18]
they [19] show
studyis a the that shear stresses
total exception
notable length of the are
to the usually
unitsimplifyingnegligible
cell as a whole. for distributed
Nevertheless,
assumption loads.
the solution
of neglecting shear Balaam
and stresses
Booker [19]
requiresbecause is athey
numerical notablestudyexception
integration, the which to theof
makes
total length simplifying
thethe solution assumption
complex
unit cell of neglecting
for practical
as a whole. purposes.
Nevertheless, theshear
solution stresses
because Most
requires
they analytical
numerical
study the solutions
integration,
total length focus
which on
ofthethesettlement
makes thecell
unit reduction
solution caused
as a complex
whole. forbypractical
stone columns
Nevertheless, the(e.g.,
purposes. [16,17,19]),
solution requires
but stone
numerical Most columns also
analytical
integration, which act as
solutions vertical
focus
makes onthe drains
the and, complex
settlement
solution therefore,
reductionforthey dissipate
caused by stone
practical excess
columns
purposes. pore(e.g.,
pressures.
[16,17,19]),The
consolidation
but stone columnsprocessalso may
act be
as studied
vertical independently
drains and, from
therefore,
Most analytical solutions focus on the settlement reduction caused by stone columns the
theysettlement
dissipate analysis
excess using
pore solutions
pressures. for
The
vertical drainsprocess
consolidation [20,21].may Hanbeand Ye [22]
studied and Castro and
independently from Sagaseta [23] showed
the settlement analysis that the consolidation
using solutions for
(e.g., [16,17,19]), but stone columns also act as vertical drains and, therefore, they dissipate excess pore
processdrains
vertical around stone Han
[20,21]. columnsand Ye is slightly
[22] and different
Castro and because
Sagasetaof the
[23]distribution
showed thatofthe vertical stresses
consolidation
pressures. The consolidation process may be studied independently from the settlement analysis using
betweenaround
process soft soil and columns
stone stone columns is slightly and different
they proposed because specific
of the analytical
distribution solutions to study
of vertical stresses the
solutions forsoft
vertical
consolidation
between process
soil
drains
and stonearound [20,21].
stoneHan
columns
and Yeproposed
columns.
and theyVery
[22] andspecific
recently, Castro andLogar
Pulkoanalytical
and Sagaseta [23] to
[24] have
solutions
showed
developed
study thea
that the
consolidation
consolidation process
fully coupled process around
solutionaround stone
for thestone columns
consolidation is
columns.process slightly different
assuming
Very recently, because
the and
Pulko soil Logar of the distribution
as a poroelastic
[24] have medium
developed of vertical
anda
stresses
the between
fully stone
coupled soft soil
column
solutionas an and
for thestone
elastoplastic columns
consolidationmaterial. and The
process they proposed
solution
assuming is the
very specific
soil as a analytical
accurate but requires
poroelastic solutions
numerical
medium to study
and
the consolidation
inversion
the of the
stone column process
as anaround
Laplace transform.
elastoplastic stonematerial.
columns. The Very recently,
solution Pulko
is very and Logar
accurate [24] have
but requires developed
numerical
Extending
inversion of the analytical
Laplace solutions
transform. for ordinary stone
a fully coupled solution for the consolidation process assuming the soil as a poroelastic mediumcolumns to encased stone columns is quite and
straightforward
Extending (e.g.,
analytical[25]). Equilibrium
solutions for and compatibility
ordinary stone conditions
columns
the stone column as an elastoplastic material. The solution is very accurate but requires numerical to of the
encasedgeosynthetic
stone encasement
columns is quiteare
those of a
straightforwardthin-walled
(e.g.,
inversion of the Laplace transform. tube,
[25]). or better
Equilibrium said,
and those of
compatibility a flexible membrane
conditions of the because
geosyntheticthe encasement
encasement does
are
not usually
those support compressive
of a thin-walled tube, or better stresses.
said, thoseThe of internal andmembrane
a flexible external pressures
because the areencasement
here denoted does as
Extending analytical solutions for ordinary stone columns to encased stone columns is quite
not usually support compressive stresses. The internal and external pressures are here denoted as
straightforward (e.g., [25]). Equilibrium and compatibility conditions of the geosynthetic encasement
are those of a thin-walled tube, or better said, those of a flexible membrane because the encasement
does not usually support compressive stresses. The internal and external pressures are here denoted
Materials 2017, 10, 782 5 of 23

as rcMaterials
and 2017,
rs , respectively
10, 782 (Figure 4). Thus, the equilibrium and compatibility conditions of the
5 of 23

geosynthetic encasement are the following, respectively:

and , respectively (Figure 4). Thus, the equilibrium and compatibility conditions of the
geosynthetic encasement are the following, respectively:
Tg
rc = + rs (1)
rc
= + (1)
sr
Tg = Jg (2)
rc
= (2)
where sr is the radial displacement of the column and the encasement, Jg is the hoop or circumferential
where
stiffness is the radial displacement
of the geosynthetic and Tg is the ofhoopthe force
column andgeosynthetic.
at the the encasement, is the
The units of Jhoop
g andor Tg are
force circumferential
per length (F/L) stiffness of thethe
because geosynthetic
thicknessand is the hoop force
of the geosynthetic is at the geosynthetic.
usually The
negligible. units
The of
equation
and the are
that relates forcestress
radial per length
of the(F/L) because
column (rcthe thickness
) with of the
that of thegeosynthetic
soft soil (rsis) usually negligible.
and allows The
the confining
equation that relates the radial stress of the column ( ) with that of the soft soil ( ) and allows the
effect of the encasement to be included is obtained by combining Equations (1) and (2):
confining effect of the encasement to be included is obtained by combining Equations (1) and (2):
Jg sr
rc == + rs
+ (3) (3)
rc2

rs
sr
rc

Tg rc Tg

Figure 4. Equilibrium and compatibility conditions for the geosynthetic surrounding an encased
Figure 4. Equilibrium and compatibility conditions for the geosynthetic surrounding an encased
stone column.
stone column.
The unit cell model is also very useful for numerical analyses because numerical simulations
allowunit
The more complex
cell modelfeatures
is alsotovery
be studied
usefulthanfor with the analytical
numerical analysessolutions,
because suchnumerical
as layered ground,
simulations
advance constitutive models or cyclic loads (e.g., [2629]). Nowadays,
allow more complex features to be studied than with the analytical solutions, such as layered 3D numerical codes are fairly
ground,
easy to access and to use; so, there are some authors that study the
advance constitutive models or cyclic loads (e.g., [2629]). Nowadays, 3D numerical codes areunit cell as a full 3D problem, i.e.,fairly
considering the hexagonal or square prism. Nevertheless, the differences between the full 3D prism and
easy to access and to use; so, there are some authors that study the unit cell as a full 3D problem, i.e.,
the two-dimensional (2D) model (cylinder) in axial symmetry are negligible [30,31]. The simplicity of
considering the hexagonal or square prism. Nevertheless, the differences between the full 3D prism
the unit cell model has recently led to highly advanced numerical models, such as that presented by
and the two-dimensional (2D) model (cylinder) in axial symmetry are negligible [30,31]. The simplicity
Indraratna et al. [32], where the column is modelled using 2D discrete elements to represent the gravel
of theparticles.
unit cellThis
modeltype has recentlyanalysis
of numerical led to highly advanced
can be regarded onlynumerical
as explorativemodels,
and forsuch as that
research presented
purposes.
by Indraratna
The unit cell etmodel
al. [32], where
or the model theofcolumn
a slice of is
themodelled
unit cell atusing 2Ddepth
a specific discrete elements
are also used fortosmall-scale
represent the
gravellaboratory
particles.studies
This type
(e.g., of numerical analysis can be regarded only as explorative and for research
[33,34]).
The unit cell also
purposes. The unit cell model or the allows floating
model columns to be
of a slice ofstudied,
the uniti.e.,
cellcolumns that do
at a specific not reach
depth a rigid
are also used for
substratum.
small-scale In thisstudies
laboratory case, the(e.g.,
settlement
[33,34]).due to column punching into the underlying layer and the
deformation
The unit cellofalsothe underlying
allows floatinglayer should
columns be considered [35]. The
to be studied, i.e.,unit cell model
columns thatisdo
only valid
not when
reach a rigid
the ratio between the width of the loaded area and the thickness of the soft soil layer ( / ) is high
substratum. In this case, the settlement due to column punching into the underlying layer and the
enough to ensure confined (oedometric) conditions. On the contrary, the applied load is distributed
deformation of the underlying layer should be considered [35]. The unit cell model is only valid when
with depth, in a roughly trapezoidal manner, its value decreasing with depth. Figure 5 presents some
the ratio between
specific proposalsthe of
width of the loaded
load distribution area
width andfor
depth theend-bearing
thickness and of the soft soil
floating layer[3638].
columns (B/H)The is high
enough to ensure confined (oedometric) conditions. On the contrary, the applied
unit cell model for these cases usually overestimates the settlement because the reduction of the applied load is distributed
with depth,
verticalin a roughly
stress with depthtrapezoidal manner, its value decreasing with depth. Figure 5 presents some
is not reproduced.
specific proposals
Finally, theofunitload celldistribution
is an appropriatewidth depth for
simplified end-bearing
geometrical modeland floating
to study columnsand
the settlement [3638].
its evolution
The unit cell model with time
for at the
these center
cases of an embankment,
usually overestimatesbut theobviously
settlement it is because
not valid the
for studying
reduction theof the
stability
applied verticalof the lateral
stress withslopes
depth(Table 1). For
is not similar reasons, it does not allow the lateral spreading and
reproduced.
the contribution of a geosynthetic reinforcement
Finally, the unit cell is an appropriate simplified that acts as a blanket
geometrical model or bridging
to study the layer over the and
settlement
columns and soft soil foundation (geosynthetic reinforced and column supported embankments,
its evolution with time at the center of an embankment, but obviously it is not valid for studying the
stability of the lateral slopes (Table 1). For similar reasons, it does not allow the lateral spreading
and the contribution of a geosynthetic reinforcement that acts as a blanket or bridging layer over
Materials 2017, 10, 782 6 of 23

Materials 2017, 10, 782 6 of 23

the columns and soft soil foundation (geosynthetic reinforced and column supported embankments,
GRCSE) GRCSE)
to be studied [31]. To
to be studied study
[31]. the lateral
To study spreading
the lateral spreadingof of
thetheembankment,
embankment, the unitcell
the unit cellmodel
model may
be improved by
Materials
may be substituting
2017, the horizontally-fixed
10, 782 by substituting
improved external
the horizontally-fixed lateral
external lateral boundary
boundary byby elastic
elastic springs
6 of
springs 23 [39].
[39].

GRCSE) to be studied [31]. To study the lateral spreading of the embankment, the unit cell model
may be improved by substituting the horizontally-fixed external lateral boundary by elastic springs [39].

(a) (b)
Figure 5. Stress distribution beneath footings: (a) End-bearing columns; (b) Floating columns.
Figure 5. Stress distribution beneath footings: (a) End-bearing columns; (b) Floating columns.
(a) (b)
4. Longitudinal Gravel Trenches
Figure 5. Stress distribution beneath footings: (a) End-bearing columns; (b) Floating columns.
4. Longitudinal
ManyGravel Trenches
geotechnical problems fulfill plane strain conditions. Thus, when stone columns are used
4.inLongitudinal
Many those problems,
geotechnical for example,
Gravel Trenches
problems for the foundation of an embankment for a linear infrastructure, it is
fulfill plane strain conditions. Thus, when stone columns are used
useful to study the problem under 2D plane strain conditions, transforming the columns into
in thoselongitudinal
problems, for
Many geotechnicalexample, for
problems the
fulfillfoundation
gravel trenches (Figure 6). plane
of an embankment
strainofconditions.
The width Thus,
the trenches and when
for acolumns
stone
the spacing
linear infrastructure,
between are used
them are
in
part of the unknown parameters that should be properly estimated to be equivalent. Besides, itit isis into
it is useful those
to problems,
study the for example,
problem for
under the2Dfoundation
plane of
strain an embankment
conditions, for a linear infrastructure,
transforming the columns
useful
longitudinal
usually tonecessary
gravelstudy the
to problem
trenchesalter(Figureunder 2D plane
6). The
the gravel widthstrain
parameters of these
for conditions,
the trenches transforming
and
equivalent the columns
the trenches.
gravel spacing between into
Furthermore,them are
longitudinal
to appropriatelygravel trenches (Figure
model the consolidation6). The width
process, of the trenches and
it is also estimated the
necessary totospacing
modify between them
the permeability are
of
part of the unknown parameters that should be properly be equivalent. Besides, it is
part
the of the unknown
natural soft soil. parameters that should be properly estimated to be equivalent. Besides, it is
usually usually
necessary to alter
necessary the the
to alter gravel
gravelparameters
parameters for theseequivalent
for these equivalent gravel
gravel trenches.
trenches. Furthermore,
Furthermore,
to appropriately modelmodel
to appropriately the consolidation
the consolidation process, it itis isalso
process, alsonecessary to modify
necessary to modifythe the permeability
permeability of of the
natural the
softnatural
soil. soft soil.

Figure 6. Longitudinal gravel trenches to model a stone column treatment in plane strain.

4.1. Settlement
Figure Figure 6. Longitudinal
6. Longitudinal gravel
gravel trenchesto
trenches tomodel
model aastone
stonecolumn treatment
column in plane
treatment strain. strain.
in plane
The classical proposal for this simplified model is by Van Impe and De Beer [40], who developed
4.1. Settlementsolution to study the settlement for this case. Their proposal [40] is to keep the (drained)
an analytical
4.1. Settlement
properties of the soil and column and transform each row of columns in a longitudinal gravel trench
The
of the classical
same area. proposal for this simplified model is by Van Impe and De Beer [40], who developed
Theanclassical
analytical
The most
proposal
solution tofor
important
thisthe
study simplified
parametersettlement model thisis
forcolumn
in a stone
by
case. Van Impe
Their
treatment
and
proposal De
[40]
is the area is Beer [40],
to keep
replacement
who
thefactor developed
(drained)
( ),
properties
an analytical
whichsolutionof theto
represents soil and
areacolumn
study
the the and transform
settlement
of soft each
fororthis
soil replaced row ofTheir
case.
displaced columns
by in acolumns:
theproposal
stone longitudinal
[40] is togravel
keep trench
the (drained)
of the
properties of same area.and column and transform each row of columns in a longitudinal gravel trench
the soil
The most important parameter in a stone column = =treatment is the area replacement factor ( (4)),
of the same area.
which represents the area of soft soil replaced or displaced by the stone columns:
The most important parameter in a stone column treatment is the area replacement factor (ar ),
The proposal by [40] seems reasonable because it allows the area replacement factor ( ) and the
which represents the area
soil and gravel of softtosoil
parameters replaced
be kept. or=displaced
However, = two major
it has by thedisadvantages:
stone columns: (4)

The proposal by [40] seems reasonable because rc2 the area replacement factor ( ) and the
Ac it allows
a r =
soil and gravel parameters to be kept. However,A =
it has two (4)
r2 major disadvantages:
e e

The proposal by [40] seems reasonable because it allows the area replacement factor (ar ) and the
soil and gravel parameters to be kept. However, it has two major disadvantages:
Materials 2017, 10, 782 7 of 23

Materials 2017, 10, 782 7 of 23

The resulting thickness of the gravel trenches is usually small, and consequently, the gravel
The resulting thickness of the gravel trenches is usually small, and consequently, the gravel
trenches are very slender.
trenches are very slender.
The confining conditions
The confining of the
conditions columns
of the columnsare
arenot thesame
not the sameasas
thethe lateral
lateral confinement
confinement of the gravel
of the gravel
trenches [30] (Figure 7).
trenches [30] (Figure 7).

Materials 2017, 10, 782 7 of 23

The resulting thickness of the gravel trenches is usually small, and consequently, the gravel
trenches are very slender.
The confining conditions of the columns are not the same as the lateral confinement of the gravel
trenches [30] (Figure 7).

Figure 7. Different confinement and seepage conditions for columns in axial symmetry and for
Figure 7. Different
longitudinal confinement
gravel and seepage
trenches in plane strain. conditions for columns in axial symmetry and for
longitudinal gravel trenches in plane strain.
Regarding the first bullet point, there is no restriction on increasing the thickness of the gravel
trenches, while proportionally increasing the spacing between the longitudinal trenches to keep
Regarding the first bullet point, there is no restriction on increasing the thickness of the gravel
constant. A reasonable value for the thickness of the gravel trenches may be the diameter of the
trenches, while proportionally increasing the spacing between the longitudinal trenches to keep
columns.
ar constant.The AFigure 7. Differentvalue
reasonable
second
confinement
bullet point causesfor the and thickness
seepage conditions
that the matching of the forgravel
of the
columns trenches
in axial symmetry
settlement between maythe be
and the
for diameter of
real situation
longitudinal gravel trenches in plane strain.
the columns.
and the plane strain model is not accurate enough, particularly when plastic strains appear in the
Thecolumns
second and
Regardingthen,
bullet thethe
point differences
causes
first thatinthe
bullet point, thematching
there column confinement
of theonsettlement
is no restriction are notable.
increasing thebetweenPlasticthe
thickness ofstrains
real
the in the
situation
gravel and
columns arewhile
trenches, common and, if theincreasing
proportionally column only the deforms
spacing elastically,
between the the designtrenches
longitudinal is overconservative.
to keep
the plane strain model is not accurate enough, particularly when plastic strains appear in the columns
Tanconstant.
et al. [30]A presented
reasonable two value analytical proposals
for the thickness of to
theobtain
gravelthe equivalent
trenches may be properties
the diameterof the of gravel
the
and then, the differences
trenches and the
in the column
surrounding soil, but
confinement
neither of them
are
is
notable.
totally
Plastic An
satisfactory.
strains in the
analytical
columns are
equation
columns.
common mayand, if the
be The
derived column
for bullet
second only
point deforms
the equivalent elastic
causes elastically,
thatmodulus
the matchingof thethe design
ofgravel
the is overconservative.
trenches
settlement [41], but the
between thatreal
is only Tan
validet
situation if al. [30]
presentedthe two
column
and analytical
the and the
plane proposals
soil
strain deform
model to obtain
elastically.
is not accurateForthe equivalent
general
enough, cases,properties
particularly comparing
when plastic of the
the unit gravel
cellappear
strains modeltrenches axialand the
ininthe
surrounding columns
symmetrysoil,andandthe
but then,
unitthecell
neither ofdifferences
of the is
them in the
gravel
totally column
trenches isconfinement
satisfactory.very useful are notable.
(Figure
An analytical Plasticstarting
8).equation
Thus, strains
mayin bethe
from an
derived for
columnsproposal
analytical are commonfor and,
the if the column
parameters of theonly deforms
gravel elastically,
trenches, they the design
may be is overconservative.
further adjusted or tuned
the equivalent elastic modulus of the gravel trenches [41], but that is only valid if the column and the
Tan et the
to match al. [30] presented
settlement twounit
of the analytical
cell inproposals to obtainTo
axial symmetry. thefitequivalent
the settlementproperties
in theofplastic
the gravel
range,
soil deformtrenches
elastically.
and For
the general cases,
surrounding soil, butcomparing
neither the unit
of them cellsatisfactory.
model in An axial symmetry and the unit
adjusting the friction angle of the gravel trenches seemsisthe totally
most analytical
appropriate alternative equation
[41]. Once
cell of the
the parameters have been calibrated using the unit cell case as an auxiliary problem, they mayif be for the
gravel
may be trenches
derived for is
the very useful
equivalent (Figure
elastic modulus8). Thus,
of the starting
gravel from
trenches an
[41], analytical
but that is only proposal
valid
parameters the
ofcolumn
the and the soil deform elastically.beFor generaladjusted
cases, comparing thetounit cell model in axial
used for thegravel
full 2Dtrenches,
plane they
strain mayHowever,
model. further the calibration oristuned
tailor-made match the settlement
or specific for each of the
symmetry and the unit cell of the gravel trenches is very useful (Figure 8). Thus, starting from an
unit cellcase.
in axial
Thus,symmetry.
it varies, forTo fit the with
example, settlement
the value inofthe
theplastic
appliedrange, load [42].adjusting the friction angle of the
analytical proposal for the parameters of the gravel trenches, they may be further adjusted or tuned
gravel trenches seems
to match the mostofappropriate
the settlement the unit cell inalternative [41].ToOnce
axial symmetry. fit thethe parameters
settlement have been
in the plastic range,calibrated
using the unit cell the
adjusting case as anangle
friction auxiliary problem,
of the gravel they
trenches may
seems thebe used
most for the alternative
appropriate full 2D plane strain model.
[41]. Once
the parameters
However, the calibration have been calibratedorusing
is tailor-made the unit
specific forcell
eachcase as anThus,
case. auxiliary problem,for
it varies, they may be with the
example,
used for the full 2D plane strain model. However, the calibration is tailor-made or specific for each
value of the applied load [42].
case. Thus, it varies, for example, with the value of the applied load [42].

Figure 8. Calibration of the parameters of longitudinal gravel trenches using the unit cell model as an
auxiliary problem.

4.2. Consolidation
TheFigure
coefficient of horizontal permeability (or the coefficient of consolidation ) of the
8. Calibration of the parameters of longitudinal gravel trenches using the unit cell model as an
Figure 8. Calibration
natural soil should
auxiliary of
bethe parameters
adjusted
problem. of longitudinal
to properly reproduce gravel trenches using
the consolidation the under
process unit cell model
plane as an
strain
conditions
auxiliary because, similarly to the column confinement, the seepage problem is different in axial
problem.
symmetry and plane strain conditions (Figure 7). Hird et al. [43] and Indraratna and Redana [44]
4.2. Consolidation

4.2. Consolidation
The coefficient of horizontal permeability (or the coefficient of consolidation ) of the
natural soil should be adjusted to properly reproduce the consolidation process under plane strain
The coefficient of horizontal
conditions because, similarly to the column kconfinement,
permeability h (or the coefficient
the seepageofproblem
consolidation cvhin) axial
is different of the natural
soil shouldsymmetry and plane
be adjusted strain conditions
to properly reproduce(Figure
the 7). Hird et al. [43]process
consolidation and Indraratna
under and Redana
plane strain [44]conditions
because, similarly to the column confinement, the seepage problem is different in axial symmetry and
plane strain conditions (Figure 7). Hird et al. [43] and Indraratna and Redana [44] proposed analytical
Materials 2017, 10, 782 8 of 23

expressions to modify k h , based on the comparison between analytical solutions for vertical drains for
both cases (axial symmetry and plane strain).
When using numerical analyses, the analytical values of k h [43,44] may be tuned using the unit
cell as an auxiliary problem (Figure 8) to have a better matching of the results. As it is difficult to
match the whole consolidation curve, it may be convenient to match the values of the degrees of
consolidation that are relevant for the study, usually around 8095%.

4.3. Stability
The plane strain model is useful for studying the global stability, or, for example, the stability
of the lateral slopes of an embankment (Table 1). In this case, the plane strain model should match
the average shear strength along the slip line of the real situation. Here, it is important to distinguish
between limit equilibrium analyses (e.g., method of slices) or full stress-strain analyses (e.g., finite
differences or finite element method).
Stress-strain analyses are able to capture the stress concentration on the columns. Thus, if the
stress concentration on the columns is properly reproduced (depending on the rigidity of soil and
gravel), the results are satisfactory just keeping constant the area replacement ratio ar and the resistance
properties (shear strength) of soil and gravel (e.g., [45]).
For limit equilibrium analyses, the stress concentration on the longitudinal gravel trenches
should be artificially generated because the vertical stress is directly the weight above the studied
point. Priebe [35] proposed altering the real profile of the embankment, so the modeled height of the
embankment is the real one times the stress concentration either on the columns or on the soft soil.
Another alternative is to modify the friction angle of the longitudinal gravel trenches to match the
average shear strength along the slip line [15].
A very difficult task is to evaluate the stress concentration on the columns because it depends
on many factors, such as the drainage conditions (undrained, partial drainage or fully drained), the
depth and inclination of the slip line, the specific position of the columns beneath the slope and the
applied load. Recent studies [45] show that in situations close to failure, there is no stress concentration
on the columns, i.e., the vertical stress on soil and columns is very similar. Thus, assuming no stress
concentration may be generally advisable and in any case, it is an assumption on the safe side.
Finally, rigid inclusions, such as deep mixing columns, may fail not only by shearing as stone
columns (shear failure along the slip line), but they may also fail by tilting or bending [46]. For widely
spaced columns, soil may also extrude between columns [47]. When columns break by bending,
it is necessary to consider their moment of inertia [48] or the measured bending failure for encased
stone columns.

5. Homogenization
The homogenization method consists in replacing the stone columns and the soft soil by an
equivalent homogeneous soil with improved properties. This equivalent homogeneous soil occupies
the zone treated with stone columns. This model simplifies enormously the geometry of the problem.
For example, when the problem itself has a highly complex geometry and the zone treated with
columns is just a part of the problem, this method is highly advisable (e.g., [49]). For the design stage,
this method allows the area replacement ratio (ar ) to be varied without changing the geometry of the
model, just the material parameters.

5.1. Equivalent Parameters

The most straightforward proposal for obtaining the improved parameters of the equivalent
homogeneous soil is just to average the soil and column parameters weighted by their corresponding
areas though ar . Thus, for the elastic modulus, the weighted average is:

Em = Es (1 ar ) + Ec ar (5)
Materials 2017, 10, 782 9 of 23

Materials 2017, 10, 782 9 of 23

However, this can only be regarded as a first approximation. A more detailed analysis shows
that the equivalent
orientation [50]. Some homogenous
authors (e.g., soil[5153])
shoulduse be anisotropic
theories for by naturemedia
periodic to account for the
to propose column
analytical
Materials 2017, 10, 782 9 of 23
orientation [50]. from
transformations Some authors
the columns (e.g.,
and[5153])
the soft usesoil theories for periodic
(heterogeneous media
periodic to propose
composite analytical
material at the
transformations
orientation [50]. Some authors (e.g., [5153]) use theories for periodic media to propose analyticaluse at
microscale) to a from the
homogeneous columns and
equivalent the soft
soil soil (heterogeneous
(homogenous material periodic
at a composite
macroscale). material
They a
macroscopic
the microscale) strengthfrom
transformationsto a criterion
homogeneous of the and
the columns anisotropic
equivalent soil
the soft soilhomogeneous
(homogenous
(heterogeneous equivalent
material atsoil to evaluate
a macroscale).
periodic composite the
material at bearing
They theuse a
microscale)
capacity.
macroscopic to a homogeneous
Thestrength
practical application
criterion equivalent
of the of thesesoiltechniques
anisotropic (homogenous
homogeneous material
is complex atand
equivalent a macroscale).
they
soil Theythe
are only
to evaluate use
valida for
bearing
macroscopic
sufficiently
capacity. Thesmall strength
practicalvalues criterion
of the ofof
application the
scale anisotropic
thesefactor, homogeneous
i.e.,
techniques the equivalent
ratio between
is complex and theythesoil
areto evaluate
micro
only andthe
valid bearing
macroscales,
for sufficiently
for
small capacity.
example,
values of The
thethe practical
ratio
scale between application
factor, the the
i.e., of these
column
ratio techniques
spacing
between and is characteristic
the
the micro complex and they
and macroscales, size of are only
forthe validproblem
whole
example, for ratio
the
sufficiently small values of the scale factor, i.e., the ratio between the micro and macroscales,
to be studied.
between the column spacing and the characteristic size of the whole problem to be studied.
for example, the ratio between the column spacing and the characteristic size of the whole problem
In aasimilar
similarmanner
manner as the
as for for longitudinal
the longitudinal gravelgravel
trenches,trenches,
a simplea method
simple when method when
using using
numerical
to be studied.
numerical
analyses is analyses
tuning or is adjusting
tuning or adjusting
the improved the improved
parameters parameters of the equivalent
of the equivalent
In a similar manner as for the longitudinal gravel trenches, a simple method when using
soil usingsoil theusing the
unit cell
unit
as ancell as an auxiliary problem (Figure 9). Ng and Tan [54] have calibrated
numerical analyses is tuning or adjusting the improved parameters of the equivalent soil using the for
auxiliary problem (Figure 9). Ng and Tan [54] have calibrated the the improved
improved parameters
parameters
for several
several cases
unit cases
cell as anand
and tabulated
tabulated
auxiliary their
their
problem results
results
(Figure asas
9). aand
reference.
a reference.
Ng Tan [54] The
The accuracy
accuracy
have ofthe
of
calibrated the
the matching
matching
improved may be high
may
parameters
and, yet the specific
for several cases values of the their
and tabulated excess poreaspressure
results a reference.between columns
The accuracy may
of the not bemay
matching modeled
be high using
and, yet the specific
the homogenization
homogenization values of the
technique,
technique, the
the excess
average
average pore pressure
degree
degree of between
of columns
consolidation
consolidation ormay
or thenot
the be modeled
settlement
settlement using
rate
rate may
may be
the homogenization
correctly capture after a proper technique,
proper the average
calibration
calibration of
of the degree
the of consolidation
parameters
parameters of
of the or the settlement
theequivalent
equivalent homogeneous
homogeneousrate maysoilbe [54].
soil [54].
correctly
For most cases, capture after
cases, it may be enough a proper calibration of the parameters of the equivalent
enough to match just the values of the degree of consolidation homogeneous soil [54].
consolidation that are
For most cases, it may be enough to match just the values of the degree of consolidation that are
relevant for the study, usually around 8095%. Figure 10 shows a simple example of application for
relevant for the study, usually around 8095%. Figure 10 shows a simple example of application for
the extension of an an airport
airport runway,
runway, whose embankment had to be founded founded on ground ground improved
improved with
the extension of an airport runway, whose embankment had to be founded on ground improved with
floating stone columns.
floating stone columns.

Figure 9. Calibration of the parameters of the equivalent homogeneous soil using the unit cell model
9. Calibration of
Figure 9.
Figure of the
the parameters
parameters of
of the
the equivalent
equivalent homogeneous
homogeneous soil
soil using
using the
the unit
unit cell
cell model
model
as anCalibration
auxiliary problem.
as an auxiliary problem.
as an auxiliary problem.

Figure 10. Example of calibration of the parameters of the equivalent homogeneous soil using the unit
cell model as an auxiliary problem.

Figure 10. Example of calibration of the parameters of the equivalent homogeneous soil using the unit
FigureThe
10. homogenization method
Example of calibration of cannot reproduce
the parameters physical
of the phenomena
equivalent at the local
homogeneous level, the
soil using such as
unit
cell model as an auxiliary problem.
stress concentration on the column
cell model as an auxiliary problem. or radial drainage towards the column, but the overall response

The homogenization method cannot reproduce physical phenomena at the local level, such as
stress concentration on the column or radial drainage towards the column, but the overall response
Materials 2017, 10, 782 10 of 23

The homogenization method cannot reproduce physical phenomena at the local level, such as
stress concentration on the column or radial drainage towards the column, but the overall response
of the system on a larger scale may be correctly reproduced after calibrating the parameters of the
equivalent homogeneous soil.
Materials 2017, 10, 782 10 of 23
5.2. Specific Constitutive Model
of the system on a larger scale may be correctly reproduced after calibrating the parameters of the
The idea of developing an advanced, specific constitutive model to represent the behavior of
equivalent homogeneous soil.
the periodic material formed by the columns and the corresponding surrounding soil was initially
formulated by Schweiger
5.2. Specific and
Constitutive Pande [55]. The constitutive model assumed that both soil and column
Model
undergo theThe sameidea of developing is
strains, there anno slippage
advanced, between
specific soil and
constitutive column
model and the
to represent themodel
behaviorinternally
of the uses
independent and existing constitutive models for the soil and the column and
periodic material formed by the columns and the corresponding surrounding soil was initially ensures that there is
equilibrium of horizontal
formulated stresses
by Schweiger at the
and Pande soil-column
[55]. interface
The constitutive [55]. The
model assumed model
that was
both soil later
and improved,
column
undergo
particularly its the same strains,
numerical there is no slippage
implementation between
[56]. soil andthe
Recently, column
ideaandhasthebeen
modelapplied
internallytouses
deep soil
mixing independent
columns using and existing
advanced constitutive modelsmodels
constitutive for the soil
and and the column
with the aimand of ensures
using itthat
forthere is
complex 3D
equilibrium of horizontal stresses at the soil-column interface [55]. The model was later improved,
geometries [57]. However, the numerical stability of its formulation and its range of application are
particularly its numerical implementation [56]. Recently, the idea has been applied to deep soil
still limited.
mixing columns using advanced constitutive models and with the aim of using it for complex 3D
geometries [57]. However, the numerical stability of its formulation and its range of application are
6. Gravel
stillRings
limited.
The idea of transforming the stone columns into gravel rings, tubes or cylindrical trenches
6. Gravel Rings
follows the same concept as for the longitudinal gravel trenches (Section 4) but for problems with
The idea of transforming the stone columns into gravel rings, tubes or cylindrical trenches
axial symmetry rather than plane strain conditions. Thus, the model is used for problems with
follows the same concept as for the longitudinal gravel trenches (Section 4) but for problems with
axial symmetry, suchrather
axial symmetry as circular
than planeembankments or Thus,
strain conditions. circular storage
the model tanks.
is used The model
for problems with implies
axial the
transformation of each group of columns into an equivalent gravel ring
symmetry, such as circular embankments or circular storage tanks. The model implies the with the same area to keep
ar constant (Figure 11).
transformation In the
of each caseofofcolumns
group Figureinto
11, an
theequivalent
8 neighboring columns
gravel ring of same
with the the central column are
area to keep
transformed constant
into an(Figure 11). In the
equivalent case
ring of Figure
with 11, the 8(tneighboring
a thickness columns
r ) that gives of thearea.
the same central column arethat tr is
Assuming
transformed
small enough, into an
its value is:equivalent ring with a thickness ( ) that gives the same area. Assuming that
is small enough, its value is:
tr = d2c /rr (6)
= / (6)
The distance of the gravel ring to the central point (rr ) is the weighted average of the distance of
The distance of the gravel ring to the central point ( ) is the weighted average of the distance of
the 8 columns to the center.
the 8 columns to the center.  
rr = 1 + 2 s/2 (7)
= 1 + 2 s/2 (7)
For a group of onlyof4only
For a group columns, the weighted
4 columns, average
the weighted directly rr =
averageisisdirectly = s s [58].
[58]. The proposalofof other
The proposal
authors other
[59] is slightly different and implies that the gravel ring surrounds the same area
authors [59] is slightly different and implies that the gravel ring surrounds the same area as that
as of the
square formed bysquare
that of the the 8formed
columns:by the 8 columns:

rr ==2s/
2s/ (8) (8)
For practical
For practical purposes,
purposes, the the differencesare
differences arenegligible.
negligible.

Figure 11. Transformation of stone columns in equivalent gravel rings.

Figure 11. Transformation of stone columns in equivalent gravel rings.
Materials 2017, 10, 782 11 of 23

Mitchel and Huber [58] seem to be the first authors to use this model. They used it in combination
with theMaterials
finite2017,
element
10, 782 method to study a field case (the foundation of a water treatment 11 of 23 plant).
ContraryMaterials
to the2017,
model of the longitudinal gravel trenches, here the confining conditions of11the
10, 782 of 23 gravel

rings seem toMitchel and Huber

be somehow [58] seem
similar to be the
to those of first authorscolumns
the stone to use thisand
model.
it isThey usedto
enough it in combination
maintain the values
Mitchel
with the finiteand Hubermethod
element [58] seem totostudy
be theafirst
fieldauthors to use
case (the this model.
foundation of They usedtreatment
a water it in combination
plant).
of ar and of the drained properties of the soil and the columns to obtain satisfactory results [60,61].
with the to
Contrary finite element
the model method
of the to studygravel
longitudinal a field case (the
trenches, foundation
here of a conditions
the confining water treatment plant).
of the gravel
As for the longitudinal
Contrary
rings seemtotothe gravel
bemodel
somehowof trenches,
thesimilar tothere
longitudinal isofnot
gravel
those yet any
trenches,
the stone satisfactory
here
columnstheand it way
confining to convert
conditions
is enough thegravel
of the
to maintain cylindrical
the
encasement (geosynthetic)
rings seem
values of to
andbe of that
somehow surrounds
similar
the drained propertiesthe
to those columns
ofofthethe for
stone
soil this
and columns case.
and to
the columns it obtain
is enough to maintain
satisfactory the
results
values As
[60,61]. of for and of the drained
the longitudinal properties
gravel of the
trenches, soilisand
there notthe
yetcolumns to obtainway
any satisfactory satisfactory results
to convert the
7. Three-Dimensional
cylindrical for theSlice
[60,61]. Asencasement of Columns
longitudinal gravelthat
(geosynthetic) trenches, therethe
surrounds is columns
not yet any
for satisfactory
this case. way to convert the
cylindrical encasement (geosynthetic) that surrounds the columns for this case.
There are problems that
7. Three-Dimensional fulfill
Slice plane strain conditions but for the stone column treatment, such
of Columns
as an embankment for a linear
7. Three-Dimensional Slice of Columns infrastructure. In these cases, analyzing a slice or a row of columns
There are problems that fulfill plane strain conditions but for the stone column treatment, such
is very useful
as an There
when
embankment areusing
problems3D that
numerical
for a linear fulfill analyses
plane
infrastructure. strain (Figure
conditions
In these
12).
cases,but
Nowadays,
for the astone
analyzing
the highofcomputer
slicecolumn
or a rowtreatment,
columns such
is
power
and the very
availability
as anuseful
embankment of
when using3D numerical
for 3D
a linear codes makes
infrastructure.
numerical this simplified
In these12).
analyses (Figure cases, geometrical
analyzingthe
Nowadays, a slice
highor model very
a row ofpower
computer columnsappealing
andis
because the
itvery
isavailability
relatively
useful when simple
of using and it does
3D numerical
3D numerical codesnot require
analyses
makes any
(Figure
this transformation
12). Nowadays,
simplified theorhigh
geometrical determination
computer
model of equivalent
power
very appealing and
the
becauseavailability
it is of
relatively3D numerical
simple andcodes
it makes
does not this simplified
require
parameters [48,62]. Furthermore, it allows any of the features or improvements achieved with any geometrical
transformation model
or very appealing
determination of a stone
column treatment to be studied, such as consolidation, deformations and stability (Table 1), of
because
equivalent it is relatively
parameters simple
[48,62]. and it
Furthermore, does it not
allowsrequire
any ofany
the transformation
features or or determination
improvements achieved and it is
equivalent
with a stone parameters
column treatment [48,62].toFurthermore,
be studied, such it allows any of the deformations
as consolidation, features or improvements achieved
and stability (Table 1),
valid for both encased and non-encased columns.
withitaisstone
and validcolumn
for both treatment
encased to benon-encased
and studied, suchcolumns.
as consolidation, deformations and stability (Table 1),
For and
theFor
itcase
is the ofcase
valid aforsquare
both
of gridgrid
encased
a square or mesh
andor meshofofcolumns,
non-encased columns.
columns, ititisisenough
enough to model
to model justofhalf
just half of of
a row a row of
columnscolumns
dueFor toduesymmetry
the tocase conditions
of a square
symmetry (Figure
grid or (Figure
conditions 13).
mesh of13). On
columns, the contrary,
On the itcontrary,
is enough for
for to a triangular
model justgrid
a triangular grid of
halfofofcolumns,columns,
a row of
it is necessary
itcolumns to due
is necessary study at least
to study
symmetry two different
conditions
at least two rows
(Figure
different rows of
13). columns
ofOn (a(ahalf
the contrary,
columns half
of of
for each is enough)
a triangular
each is enough) grid (Figure
13). 13).
of columns,
(Figure
it is necessary to study at least two different rows of columns (a half of each is enough) (Figure 13).

Figure
Figure 12. Finite
12. Finite elementmodel
element model of
ofaa3D
3Dslice of of
slice columns.
columns.
Figure 12. Finite element model of a 3D slice of columns.

Figure 13. Numerical model of a slice of columns for triangular and square grids.
Figure 13. Numerical model of a slice of columns for triangular and square grids.
Figure
8. Isolated 13. Numerical model of a slice of columns for triangular and square grids.
Column
8. Isolated Column
For the sake of simplicity, some experimental tests, either in the field or in the laboratory, are
8. Isolated Column
For the
performed onsake of simplicity,
an isolated some
or single experimental
stone column (e.g.,tests, eitherThe
[63,64]). in the
loadfield or in the
is usually laboratory,
applied are
only on
performed
the on an isolated
column surface (Figure or single
14a) or onstone column
an area (e.g.,
slightly [63,64]).
higher than The load is usually
the column. In theseapplied only
cases, the on
area
For the sake of simplicity, some experimental tests, either in the field or in the laboratory, are
the column surface
replacement ratio ( (Figure
) may 14a) or on an
be defined asarea
the slightly higher the
ratio between thanarea
the of
column. In these
the column cases,
and the the area
loaded
performed on an isolated
replacement
or single
ratio footprint
( ) may area
stone column
be defined as the
(e.g., [63,64]). The load is usually applied only on
area or the footing [61]. Thus, forratio between
isolated the areaisofusually
columns, the column
around and the which
100%, loaded
the column
isarea surface
notor (Figure
the footing
a realistic 14a)
footprint
situation or on an
area [61].
because area
it is Thus,slightly
more for higher
isolated
efficient than the
columns,the loaded
to increase column.
is usually In these
areaaround
( << 100%, cases,
100%)which the
[61]. area
replacement ratio (a )
is not a realistic may be defined as the ratio between the area of
r situation because it is more efficient to increase the loaded area (the column and
<< 100%) [61]. area
the loaded
or the footing footprint area [61]. Thus, for isolated columns, ar is usually around 100%, which is not a
realistic situation because it is more efficient to increase the loaded area (ar << 100%) [61]. Loading
Materials 2017, 10, 782 12 of 23

the surroundings of the columns is beneficial because it increases the horizontal stresses at the lateral
Materials 2017, 10, 782 12 of 23
boundaries of the columns and increases their confinement.
TheLoading
resultsthe ofsurroundings
research performed on isolated
of the columns columns
is beneficial becauseare sometimes
it increases directly stresses
the horizontal extrapolated
at to
the lateral
other cases, leading boundaries
to some ofconfusion
the columnsand and non-accurate
increases their confinement.
predictions. Some authors (e.g., [15] (p. 28))
show that the The results bulges
column of research(or performed on isolated
notably expands columns
radially) atare sometimes
their directly
upper part, extrapolated
specifically at to
an upper
other cases, leading to some confusion and non-accurate predictions. Some authors (e.g., [15] (p. 28))
zone with a length of two or three times the column diameter (23 dc ). Hughes and Withers [63]
show that the column bulges (or notably expands radially) at their upper part, specifically at an upper
measured a length
zone of 4 column
with a length diameters
of two or three forcolumn
times the the upper bulging
diameter (23 zone for an
). Hughes andisolated column
Withers [63] (ar = 100%)
measured
throughalaboratory tests. That
length of 4 column is valid
diameters forupper
for the casesbulging
with load
zone only
for anon the single
isolated column column,
( = 100%) dc = B. In the
i.e.,through
next section, it will
laboratory beThat
tests. shown that
is valid forfor other
cases withcases, it isonmore
load only meaningful
the single = .the
to define
column, i.e., bulging
In the next zone
section,
using the footing it will be shown
width that for other
(B) instead of thecases, it is more
column meaningful
diameter to define the
[61] because bulgingcases,
in those zone using
the column
bulgingthe
mayfooting width ( ) instead of the column diameter [61] because in those cases, the column bulging
be deeper [3] (p. 114) [15] (p. 28) (Figure 14). Besides, the failure mechanism of the
may be deeper [3] (p. 114) [15] (p. 28) (Figure 14). Besides, the failure mechanism of the columns may
columns may not be bulging, for example, it may be shearing [65].
not be bulging, for example, it may be shearing [65].

(a) (b) (c) (d)

Figure 14. Column deformation for different configurations: (a) Isolated column with = 100%;
Figure 14. Column
(b) Isolated deformation
column with < for
100%;different configurations:
(c) Groups (a) Isolated
of columns beneath footing; (d)column
Column with ar = 100%;
treatment
(b) Isolated column with a
beneath an embankment. r < 100%; (c) Groups of columns beneath footing; (d) Column treatment
beneath an embankment.
Isolated columns are generally used to study their bearing capacity. The maximum vertical stress
that a stone column may bear is usually given as:
Isolated columns are generally used to study their bearing capacity. The maximum vertical stress
that a stone column may bear is usually given as: 20 (9)
where is the undrained shear strength of the surrounding soft soil. Equation (9) is derived assuming
max
that the column is in an active state with an active
vc 20c
earth pressure
u coefficient of around = 1/3 and that (9)
the radial cavity expansion factor is around 68 (e.g., [63,66]). Equation (9) is strictly valid for isolated
where cucolumns with
is the undrained = 100%,
shear but when used
strength for larger
of the loaded areas,
surrounding softitsoil.
is conservative
Equation because it neglects
(9) is derived assuming
the increase in the radial stress due to vertical loading of the soil surrounding the
that the column is in an active state with an active earth pressure coefficient of around k ac = 1/3 and column. Besides,
Equation (9) assumes perfect undrained conditions for the soil surrounding the column and some
that the radial cavity expansion factor is around 68 (e.g., [63,66]). Equation (9) is strictly valid for
drainage could be expected near the granular column. For common cases, Equation (9) gives a vertical
isolatedload
columns withforareach
supported = 100%,
stone but when
column usedtons
of 2050 for [15]
larger loaded
(p. 6). areas,field
Sometimes, it istests
conservative
are performed because it
neglectstothe increase
check the exactinvertical
the radial
load stress duebytoone
supported vertical
isolatedloading
column withof the soil surrounding
= 100%. The approach theofcolumn.
Besides,considering
Equation (9) the assumes perfectofundrained
bearing capacity a footing asconditions
the sum of forthe the soil surrounding
contribution of each columnthe column
may and
usually be overconservative. For footings, an improvement for field tests
some drainage could be expected near the granular column. For common cases, Equation (9) gives (plate load tests) on an
isolated column may be achieved by using a larger loaded area, so
a vertical load supported for each stone column of 2050 tons [15] (p. is6).theSometimes, same as in the
field tests
footing [3] (p. 165).
are performed to check the exact vertical load supported by one isolated column with ar = 100%.
The approach of of
9. Groups considering
Columns the bearing capacity of a footing as the sum of the contribution of each
column mayStone usually
columnbetreatments
overconservative. Forused
are traditionally footings,
beneathan improvement
large for as
loaded areas, such field tests (plate load
embankments,
tests) onbut
anthey
isolated column may be achieved by using a larger loaded area, so a is the same
are also used beneath footings, when the applied loads are not high [67]. Forr these cases, theas in the
footing homogenization
[3] (p. 165). technique is valid [68]. Besides, the plane strain model using gravel trenches

9. Groups of Columns
Stone column treatments are traditionally used beneath large loaded areas, such as embankments,
but they are also used beneath footings, when the applied loads are not high [67]. For these cases,
Materials 2017, 10, 782 13 of 23

the homogenization technique is valid [68]. Besides, the plane strain model using gravel trenches
(Figure 1c)Materials
is also2017,
valid for strip footings. On the other hand, when the footing could be13converted
10, 782 of 23 into
a circular footing (e.g., a square footing), the simplified geometrical model in axial symmetry that uses
(Figure 1c) is also valid for strip footings. On the other hand, when the footing could be converted
gravel rings is also applicable [60,61] (Figure 15). For the latter case (i.e., axial symmetry conditions),
into a circular footing (e.g., a square footing), the simplified geometrical model in axial symmetry
the authorthat
hasuses
recently proposed
gravel rings an additional
is also applicable [60,61] simplified model
(Figure 15). For thatcase
the latter assumes that
(i.e., axial all the columns
symmetry
may be converted into
conditions), theaauthor
centralhascolumn
recently with the an
proposed same area (Figure
additional simplified15d),
model keeping
that assumesar constant
that all [61]. The
the columns
main advantages maymodel
of this be converted
are thatintoitamay
central
becolumn
used towith the same area
developed (Figure solutions
analytical 15d), keeping[38] and that it
constant [61]. The main advantages of this model are that it may be used to developed analytical
is also applicable to encased columns. For encased columns, it is also necessary to transform the elastic
solutions [38] and that it is also applicable to encased columns. For encased columns, it is also
propertiesnecessary
of the encasement,
to transform so the that the
elastic factor Jgof/rthe
properties c isencasement,
kept constant [69].
so that the The
factorvalidity
is of this central
kept
column modelconstant [69]. The validity of this central column model for small groups of columns is based on thesettlement
for small groups of columns is based on the small influence on the footing
small
(usually less influence
than 10%)onofthethe footing
numbersettlement (usually less
of columns andthan 10%)position
their of the number of columns
beneath and their [61]. This
the footing
position beneath the footing [61]. This simplified central column is only valid to study the footing
simplified central column is only valid to study the footing settlement and should not be used for
settlement and should not be used for other features, such as the consolidation process or the bending
other features, such
moments in as
the the consolidation process or the bending moments in the footing.
footing.

(a) (b) (c) (d)

Figure 15. Simplified geometrical model for groups of stone columns beneath a square footing:
Figure 15.(a)Simplified
3D model; (b)geometrical
Gravel rings; (c)model for groups
Homogenization; of stone
(d) One columns beneath a square footing:
central column.
(a) 3D model; (b) Gravel rings; (c) Homogenization; (d) One central column.
10. Critical Column Length
10. Critical Column Length
The concept of critical length is used for stone columns as in piles (e.g., [65]). Thus, for columns
longer than the critical length, the settlement reduction or the bearing capacity of the footing does
The concept
not notablyof change
criticalorlength
improve.isForused for stone
non-encased columns
stone columns,astheincritical
pileslength
(e.g., of
[65]). Thus, for
the columns for columns
longer than the critical
settlement length,
reduction is around 1.5 2 [61,70].
the settlement reduction or the
The critical bearing
length capacity
for encased columns ofisthe footing does
usually
not notably higher,
change around 2 3.5 for
or improve. Forcommon situations
non-encased [69]. The
stone main controlling
columns, parameter
the critical length of the critical
of the columns for
column length is the extension of the load, i.e., , but the specific value of the critical column length
settlementalsoreduction is around 1.5 2 B [61,70].
depends on other parameters, such as
The critical length for encased columns is usually
and the soil and column properties [61,65,69,70].
higher, aroundIt 2is 3.5 B for common situations
worth noting that many authors (e.g., [7174]) [69]. Thegive
main controlling
critical lengths as parameter
a function ofof thethe critical
column length is the extension of the load, i.e., B, but the specific value of the critical column length
column diameter, but the column length to diameter ratio has a minor influence by itself [61,69]. The
also depends originonofother
the misinterpretation
parameters, could such lie
asinartheandfactthe
thatsoil
the first
andproposals
column[63,64] were for[61,65,69,70].
properties the case of
an isolated column with load only on the column surface, i.e., = .
It is worthFor noting that many authors (e.g., [7174]) give critical lengths as a function of the column
settlement reduction, the critical column length is related to the pressure bulb that the
diameter, but thegenerates
footing column(Figure
length16a),to diameter ratio bearing
while for footing has a minor influence
capacity, the criticalby itself
length [61,69].
depends on The
the origin of
the misinterpretation
failure mechanismcould lie in16b).
(Figure theAsfact
thethat thelength
critical firstforproposals
settlement[63,64]
reduction were for the
is longer, thatcase of an isolated
is usually
the considered
load onlyvalue. Forcolumn
large loaded areas (high
i.e., dvalues of ), the critical column length is higher
column with on the surface, c = B.
than the soft layer thickness, and consequently, the concept of critical or optimal length does not apply.
For settlement reduction, the critical column length is related to the pressure bulb that the footing
The mechanisms in Figure 16 also explain the influence of additional columns outside the
generates footing.
(FigureColumns
16a), while
outside forthefooting
footing bearing
are crossed capacity,
by the sliptheline
critical length
or failure depends
mechanism; so, on
theythe failure
mechanism (Figure 16b). As the critical length for settlement reduction is longer, that is usually
contribute to the bearing capacity. On the contrary, they hardly influence the settlement reduction if the
considered value. For large loaded areas (high values of B), the critical column length is higher than
the soft layer thickness, and consequently, the concept of critical or optimal length does not apply.
The mechanisms in Figure 16 also explain the influence of additional columns outside the footing.
Columns outside the footing are crossed by the slip line or failure mechanism; so, they contribute to
the bearing capacity. On the contrary, they hardly influence the settlement reduction if they are outside
the pressure bulb. The only minor beneficial effect for the settlement reduction is that they distribute
Materials 2017, 10, 782
Materials 2017, 10, 782 14 of 23
14 of 23

they are outside the pressure bulb. The only minor beneficial effect for the settlement reduction is
the applied load
that theyover a slightly
distribute wider
the applied loadarea.
over Therefore, when
a slightly wider area.the bearing
Therefore, capacity
when is defined
the bearing capacityin terms of
is defined in the
a critical settlement, terms of a critical
effect of extrasettlement,
columns theoutside
effect of the
extrafooting
columnsisoutside the footing[65].
only marginal is only
marginal [65].
The presented analysis on critical column length is based on a homogeneous soft soil layer;
The presented analysis on critical column length is based on a homogeneous soft soil layer; for
for layered soils, this analysis
layered soils, this analysisisisnot directly
not directly applicable.
applicable.

Figure 16.Figure 16. Conceptual justification of critical column length in a homogenous soil layer: (a) for
Conceptual justification of critical column length in a homogenous soil layer: (a) for
settlement reduction; (b) for bearing capacity.
settlement reduction; (b) for bearing capacity.
11. Column Installation Effects
11. Column Installation Effects
Column installation alters the surrounding soil, especially when the columns are installed by
vibrodisplacement. However, it is commonly accepted that in a stone column treatment, installation
Column installation alters the surrounding soil, especially when the columns are installed by
effects are less significance and the main improvement is caused by the inclusion of gravel. The search
vibrodisplacement. However,
for more accurate designs it
hasisled
commonly accepted
to more detailed analyses that in installation
of the a stone column treatment,
effects [7578]. Notableinstallation
effects areexcess
less significance
pore pressuresand the main measured
are consistently improvement is caused
in the field by theinstallation
due to column inclusion ofHowever,
[79]. gravel. The search
for more inaccurate
most cases, these pore
designs has pressures
led to more quicklydetailed
dissipate analyses
[80]. Column of installation, especially
the installation by
effects [7578].
vibrodisplacement, also leads to an increase in the horizontal stresses [75]. This is usually quantified
Notable excess pore pressures are
through the earth pressure coefficient,
consistently measured in the field due to column
[3,75,76]. The increase in effective mean stress leads, in turn,
installation [79].
However,toinanmost cases,
increase these
in the pore
soft soil pressures
stiffness [3] (p. quickly
138). dissipate [80]. Column installation, especially by
vibrodisplacement,
The excessalsopore
leads to an increase
pressures and the in the horizontal
remolding caused by stresses
column [75]. This islead
installation usually
to an quantified
through theinstantaneous
earth pressurereduction of the undrained
coefficient, shear strength
K [3,75,76]. just afterin
The increase installation
effective[77,81].
meanAs the pore
stress leads, in turn,
pressures dissipates and the effective mean stress increases, the value of the undrained shear strength
to an increase in the soft soil stiffness [3] (p. 138).
recovers and usually increases above its initial value [77,81]. Special care must be taken with soil
The remolding
excess pore pressures
in sensitive and Poker
soils [67,77]. the vibration
remolding caused
is necessary forby column
proper column installation
compaction, butlead to an
instantaneous reduction
excessive use of the of thewith
poker undrained shear strength
many repenetrations just after
(overworking) mayinstallation
result in high[77,81].
excess pore As the pore
pressures and disturbance of sensitive or overconsolidated soil layers, such
pressures dissipates and the effective mean stress increases, the value of the undrained shear strength as a dry, stiff upper crust
that is common is soft soils and is beneficial for load distribution [79].
recovers and The usually increases above its initial value [77,81]. Special care must be taken with soil
remolding caused by column installation also alters the permeability of the soft soil in the
remoldingvicinity
in sensitive soils [67,77].
of the column. This zone,Pokerwhosevibration
permeability isisnecessary
lower than for propersoft
the natural column compaction, but
soil, is usually
excessive called
use ofthethe pokerzone
smear with many
[82,83]. Therepenetrations (overworking)
properties of the smear zone, specially itsmaysizeresult in high excess pore
and permeability,
pressures have
andbeen largely studied
disturbance for verticalor
of sensitive drains [8486]. Another related
overconsolidated problemsuch
soil layers, is column
as aclogging
dry, stiff [83].
upper crust
Gravel compaction during column installation and the high hydraulic gradients at the soil-column
that is common is soft soils and is beneficial for load distribution [79].
interface inevitably produce a migration of clay particles into the pores of the granular column,
The remolding
notably reducing caused by column
the permeability installation
of an external annulusalsoofalters the permeability
the column. This phenomenonofis the soft soil in the
normally
vicinity ofless
the column.
important This zone,encased
in geosynthetic whosecolumns
permeability
because the is geosynthetic
lower thanusually the natural
acts as asoft
filter.soil, is usually
The introduction
called the smear of installation
zone [82,83]. effects in the
The properties ofnumerical
the smear modeling
zone, of stone columns
specially is a complex
its size and permeability,
process. Some attempts have been made using field measurements and back-fitting the results [75]
have been largely studied for vertical drains [8486]. Another related problem is column clogging [83].
Gravel compaction during column installation and the high hydraulic gradients at the soil-column
interface inevitably produce a migration of clay particles into the pores of the granular column, notably
reducing the permeability of an external annulus of the column. This phenomenon is normally less
important in geosynthetic encased columns because the geosynthetic usually acts as a filter.
The introduction of installation effects in the numerical modeling of stone columns is a complex
process. Some attempts have been made using field measurements and back-fitting the results [75]
or using previous numerical analyses that simulate column installation as a cylindrical cavity
expansion [78]. From a practical point of view, column installation may be considered altering
the at-rest earth pressure coefficient of the natural soft soil. Priebes theoretical solution [16] considers,
Materials 2017, 10, 782 15 of 23

for example, a value of 1. Published values [59] vary between 0.4 and 2.5, with average values slightly
above 1. Obviously, the earth pressure coefficient after column installation should depend on the
initial value, the construction method and the type of soil. Finally, if the numerical model does
not consider a stress-dependent constitutive model for the soft soil, its stiffness should be increased
correspondingly [76].

12. Properties of the Columns

The modeling process of the soft soil is beyond the scope of this review and it is broadly
analyzed elsewhere (e.g., [87,88]). Some authors [79,89] warn about the importance of secondary
compression in some soils and the limited capacity of stone columns to reduce it. Gravel properties
are also well-analyzed (e.g., [9092]), but the properties of the gravel for stone columns are not so
readily available nor are they generally measured for each project. For stone columns, the gravel
should be clean, preferably crushed stone, hard, unweathered, free from organics or other deleterious
materials and its degradation using the Los Angeles testing machine should be less than a 45%
of loss [15] (p. 117). Its grain size should be between 12 and 75 mm. For bottom feed columns,
the maximum size is usually limited to 50 mm to avoid obstructions in the feeding tube.
The relative density of the gravel in the stone columns is not usually measured and it may vary
along the length of the column, in a similar manner as the column diameter. A proper stone column
construction should achieve relative densities of the gravel above 75% [15]. Herle et al. [93] measured
values close to 100%.
The friction angle of the columns (c ) has a notable influence on the results of a stone column
treatment. Its value decreases with the confining pressure (c ). Using samples in a dense state,
Herle et al. [93] measure values up to 60 for low confining pressures (c = 50 kPa) and values around
50 for medium confining pressures (c = 200 kPa) (Table 2). However, when choosing a value of the
friction angle to model stone columns, it is necessary to consider the following points:

Columns are usually under triaxial conditions and for granular soils, the friction angle measured
by direct shear tests is slightly higher than that measured using triaxial tests [94].
Values in Table 2 are peak values, but plastic strains in the columns may be important; so, it is
recommended to reduce peak friction values by 57% to obtain approximate residual values [3].
The validity of some theoretical methods and their success in predicting field measurements may
be linked to the use of conservative values of the friction angle.
Published numerical analyses consider values of 4045 (Table 3).

Table 2. Stress-dependent peak friction angles of dense gravel [93].

c,max c,max c,min c,min

Type of Gravel Remarks
( ) (kPa) ( ) (kPa)
Crushed lime stone 63.1 50 53.8 200 DS, Vibro SC
River gravel 58.8 50 51.9 200 DS, Vibro SC
River gravel, sub-round 57.1 50 50.9 200 DS, Vibro SC, d60 /d10 = 2.6
Rivel gravel, sub-round 59.2 50 53.2 200 DS, Vibro SC, d60 /d10 = 2.1
Rivel gravel, crushed 60.4 50 55.2 200 DS, Vibro SC
Dolomite 64.0 15 43 500 TX, = 17 kN/m3 , [90]
Dolomite 54.0 15 40 500 TX, = 15 kN/m3 , [90]
Sandstone 60.1 27 45.6 695 TX, [92]
Basalt 64.2 27 45.6 695 TX, [92]
Basalt 71.8 8 45.6 240 TX, d50 = 30 mm, [91]
Basalt 70.0 8 51.1 120 TX, d50 = 30 mm, [91]
DS: Direct shear test; TX: Triaxial test; Vibro SC: dense gravel for vibrated stone columns [93]; d60 /d10 :
uniformity coefficient.
Materials 2017, 10, 782 16 of 23

Table 3. Parameters used to model stone columns in numerical analyses.

Reference c ( ) c ( ) Ec (MPa) (-) m (-) d /sat (kN/m3 )

[59] 41 - 29.2 0.2 0.59 18.6/21.6
[68] 35 - 67.5 0.3 - -
[70] 4035 3 50 0.3 - -
[74] 45 0 100 0.3 - 19/19
[73] 48 26 2.5 0.3 - 16/-
[30] 40 - 30 0.3 - 15/15
[54] 40 0 30 0.3 - -/15
[49] 35 5 25 0.2 0.3 20/23.5
[78] 42 12 35 0.2 - 16/19
[95] 46 10 22 0.15 0.25 -
[96] 45 15 70 0.2 0.3 19/19
The reference pressure for the stiffness is 100 kPa; m: Exponent of the power law used to reproduce the stress
dependent stiffness.

The Youngs modulus of the columns is usually between 25 and 100 MPa (Table 3) and it also
varies with the confining pressure. A hyperbolic power law is sometimes used to reproduce the stress
dependent stiffness of the gravel of the columns (e.g., [58,95,96]). The common value of the exponent of
the power law is around 0.3 (Table 3). The unit weight of the gravel does not require further comment,
except for the cases where the columns do not reach a rigid substratum and they are installed by
vibrodisplacement (e.g., [49]). In those cases, the addition of gravel and its corresponding weight
may cause additional deformation in the underlying layer that is not improved with stone columns.
For example, [49] used non-realistic high values, namely sat = 23.5 kN/m3 to account for this effect.
Regarding encased stone columns, the material used to construct them is usually sand instead of
gravel. The available information is more limited, but it seems obvious to use lower values of both the
friction and dilatancy angles (Table 4).

Table 4. Parameters used to model encased stone columns in numerical analyses.

Reference c ( ) c ( ) Ec (MPa) (-) m (-) d /sat (kN/m3 )

[73] 48 4 9 0.3 - 16/-
[97] 40 10 40 0.3 - -/23
[98] 38 5 15.5 0.3 - -/18
[28] 40 - 80 - - -/20
[5] 32.5 0 - - - 19/20
[31] 35 0 30 0.2 - -
[48] 38 10 40 0.3 - -/22
[69] 40 5 30 0.33 - -/20
m: Exponent of the power law used to reproduce the stress dependent stiffness.

13. Modeling the Geosynthetic Encasement

As shown in Figure 4, the geosynthetic encasement may be modeled as a flexible membrane
that does not support compressive stresses, has a negligible thickness and behaves as a linear elastic
material with a modulus of Jg = 10005000 kN/m. Regarding its tensile strength, for modeling
purposes, it may usually be enough to verify that it is far from being reached. The tensile strength
is usually reached for circumferential strains of around 510%, which implies strength values of
100300 kN/m [5]. It is common that geosynthetic may be anisotropic and then different properties
should be input for each direction.
Some other geosynthetic features, such as creep and damage during installation, are usually
considered through reduction factors. Recent numerical analyses [99] model numerically the creep
behavior of geosynthetics.
Materials 2017, 10, 782 17 of 23

In those numerical analyses where the geosynthetic is modeled as a continuum element of small
thickness, it is necessary to ensure that it does not support compressive stresses, and hence, bending
moments. Its Youngs modulus should be Jg divided by its thickness.
Little attention is usually paid to the Poissons ratio of the geosynthetic (g ), but it may have
important consequences on the results [100]. When the encasement is modeled as a membrane, it is
common to study the two directions of the geosynthetic independently, without interaction between
them (Equation (3)). That means g = 0.
On the other hand, when the geosynthetic is modeled as a continuum element, it is necessary
to specify a value for g . In some cases (e.g., [31]), a common value of 0.3 is input. That implies
that, although the geosynthetic that surrounds the columns does not support compressive stresses
in the vertical direction, it compresses vertically (vertical strains) and then, that vertical compression
leads to a radial expansion of the geosynthetic (horizontal or circumferential strains) due to Poissons
effect. Thus, the geosynthetic encasement decreases its lateral confinement to the column due to this
radial expansion. Common geosynthetics for column encasement are woven geotextiles. For woven
geotextiles, the two directions work nearly independently; so, it seems logical to use values of the
Poissons ratio close to 0. Soderman and Giroud [101] propose values of g = 0.1 for woven geotextiles
and g = 0.35 for non-woven geotextiles.

14. Conclusions
This paper reviews the main modeling techniques for stone columns, both ordinary or
non-encased stone columns, and geosynthetic-encased stone columns. The paper tries to encompass
the more recent advances and recommendations in the topic.
There exist several simplified geometrical models. The suitability of each of them depends on
the type of process to be studied, e.g., bearing capacity or settlement, and the type of analyses, e.g.,
analytical or numerical in 2D or 3D. For numerical analyses of a problem that fulfils plane strain
conditions but for the stone column treatment, such as an embankment for a linear infrastructure,
the simplified geometrical model of a three-dimensional slice of columns is recommended because
it does not require any transformation of the problem parameters. For more simplified models for
the stone column treatment, such as gravel trenches or homogenization, calibrating or tuning the
parameters using the unit cell as an auxiliary problem is recommended. The model of an isolated
column with load just on top of it may be useful for field tests, but it is not a common situation
in real cases because loading the soil that surrounds the column is beneficial. The behavior of an
isolated column is different from that of a column within a large group under distributed load. Those
differences should be considered when extrapolating results.
For groups of columns, the critical column length depends mainly on the loading area.
For non-encased columns, its value is around twice the footing width and for encased columns
is slightly higher.
Some guidance is provided to consider installation effects and to model the column and the
geosynthetic encasement. Column installation usually increases the horizontal stresses and that is
often accounted for using a high value of the earth pressure coefficient, typically around 1. On the other
hand, the relevance of the Poissons ratio of the geosynthetic is taken into account and an appropriate
value for a woven geotextile could be close to 0, i.e., the two directions work nearly independently.
Finally, laboratory tests are not commonly performed to characterize stone column properties
and some column parameters published in the literature are tabulated as a reference.

Acknowledgments: The author thanks all the institutions that have allowed him to work on stone columns
and all the people that have collaborated in the research on stone columns. The author thanks the Spanish
Society for Soil Mechanics and Geotechnical Engineering (SEMSIG) for the invitation to give a lecture on
stone columns in the 10th National Symposium on Geotechnical Engineering, held in A Corua in October
2016. The present paper is based on that lecture and the corresponding conference paper [102] (In Spanish).
The author acknowledges the funding received through the following projects: Improvement of soft soils
with stone columns for foundation of embankments. Analysis of the process and design criteria for
Materials 2017, 10, 782 18 of 23

the Spanish Ministry of Public Works (03-A634), An integrated calculation procedure for stone columns,
considering the influence of the method of installation for the Spanish Ministry of Science and Innovation
(BIA2009-13602) and The critical distance in rock fracture for the Spanish Ministry of Economy, Industry and
Competitiveness (BIA2015-67479-R). The author is very thankful to the Group of Geotechnical Engineering
of the University of Cantabria, particularly to C. Sagaseta, J. Caizal, A. Da Costa, A. Cimentada and
M. Miranda. The research on stone columns has been contributed by the international cooperation of H.
Schweiger, B. McCabe, M. Karstunen and N. Sivasithamparam. The author acknowledges the participation
in the following projects supported by the European Commission: the Marie Cure Research Training Network
entitled Advanced Modelling of Ground Improvement on Soft Soils (AMGISS) (MRTN-CT-2004-512120) and
the Industry-Academia Partnerships and Pathways project on Modelling Installation Effects in Geotechnical
Engineering (GEO-INSTALL) (PIA-GA-2009-230638). The following companies have also contributed to the
research: Keller Group, mainly through E. Carvajal, G. Vukotic and J. Wehr; Huesker Inc., mainly through I. Diego;
and Geocisa, particularly for the field measurements carried out in Sueca, near Valencia (Spain).
Conflicts of Interest: The author declares no conflict of interest. The founding sponsors had no role in the design
of the review; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in the
decision to publish the results.

List of Symbols
ar Area replacement ratio: Ac /Ae
cu Undrained shear strength
dc Column diameter
d10 , d60 Diameters corresponding to 10% and 60% finer in the particle-size distribution, respectively
k Hydraulic conductivity/permeability
m Exponent of the power law used to reproduce the stress dependent stiffness
pa Uniform applied vertical pressure
pa (0) Uniform applied vertical pressure at the surface
pa (z) Reduced value using a trapezoidal distribution of the uniform applied vertical pressure at a depth z
r Radius
s Centre-to-centre column spacing
sr Radial displacement of the column
sz Settlement
sz0 Settlement without columns
t Thickness
u Pore water pressure
A Cross-sectional area
B Footing width
E Youngs modulus
H Soft soil layer thickness
Jg Encasement stiffness
K Coefficient of lateral earth pressure
K0 Coefficient of lateral earth pressure at rest
L (Column) length
Tg Tensile hoop force at the encasement
b Settlement reduction factor: = sz /sz0
g Unit weight
Poissons ratio
Stress
c Confining pressure
Friction angle
Dilatancy angle
Materials 2017, 10, 782 19 of 23

Subscripts
c, s, g, e column, soil, encasement or geosynthetic, loaded area or unit cell
d dry
m equivalent homogeneous soil
max maximum value
min minimum value
r equivalent gravel ring
r, , z Cylindrical coordinates
sat Saturated
x, y, z Cartesian coordinates
(upper bar) average value along the radius

Abbreviations
DS Direct shear test
G.W.T. Ground water table
SC Stone column
TX Triaxial test

References
1. Dhouib, A.; Magnan, J.-P.; Guilloux, A. Procds Damlioration des Sols: Historique, Mthodes de Reconnaissance,
Applications et Donnes conomiques; ASEP-GI; Presses de lENPC/LCPC: Paris, France, 2004; Volume 2,
pp. 577598. (In French)
2. Seed, H.B.; Booker, J.R. Stabilization of potentially liquefiable sand deposits. J. Geotech. Eng. ASCE 1977, 103,
757768.
3. Kirsch, K.; Kirsch, F. Ground Improvement by Deep Vibratory Methods; Spon Press: London, UK, 2010.
4. Wehr, J. The undrained cohesion of the soil as criterion for the column installation with a depth vibrator.
In International Symposium on Vibratory Pile Driving and Deep Soil VIBRATORY Compaction, TRANSVIB 2006,
Paris; Laboratoire Central des Ponts et Chausses: Paris, France, 2006.
5. Alexiew, D.; Raithel, M. Geotextile-Encased Columns: Case Studies over Twenty Years. In Embankments
with Special Reference to Consolidation and Other Physical Methods; Butterworth-Heinemann: Oxford, UK, 2015;
pp. 451477.
6. Simon, B. General Report S5: Rigid Inclusions and Stone Columns. In Proceedings of the ISSMGE-TC 211
International Symposium on Ground Improvement (IS-GI), Brussels, Belgium, 30 May1 June 2012; Volume
I, pp. 127168.
7. Golait, Y.S.; Padade, A.H. Analytical and experimental studies on cemented stone columns for soft clay
ground improvement. Int. J. Geomech. 2017, 17, 04016100. [CrossRef]
8. Han, J. Recent research and development of ground column technologies. Proc. ICE-Ground Improv. 2015,
168, 246264. [CrossRef]
9. Serridge, C.J. Achieving sustainability in vibro stone column techniques. Proc. ICE-Eng. Sust. 2005, 158,
211222. [CrossRef]
10. Egan, D.; Slocombe, B.C. Demonstrating environmental benefits of ground improvement. Proc. ICE-Ground
Improv. 2010, 163, 6369. [CrossRef]
11. Sivakumar, V.; McKelvey, D.; Graham, J.; Hughes, D. Triaxial tests on model sand columns in clay.
Can. Geotech. J. 2004, 41, 299312. [CrossRef]
12. Najjar, S.S.; Skeini, H. Triaxial response of clays reinforced with granular columns. Proc. ICE-Ground Improv.
2015, 168, 265281. [CrossRef]
13. Hong, Y.-S.; Wu, C.-S.; Yu, Y.-S. Model tests on geotextile-encased granular columns under 1-g and undrained
conditions. Geotext. Geomembr. 2016, 44, 1327. [CrossRef]
14. Stuedlein, A.W.; Holtz, R.D. Analysis of Footing Load Tests on Aggregate Pier Reinforced Clay. J. Geotech.
Geoenviron. Eng. ASCE 2012, 138, 10911103. [CrossRef]
Materials 2017, 10, 782 20 of 23

15. Barksdale, R.D.; Bachus, R.C. Design and Construction of Stone Columns; Report FHWA/RD-83/026; National
Technical Information Service: Springfield, VA, USA, 1983.
16. Priebe, H.J. Design of vibro replacement. Ground Eng. 1995, 28, 3137.
17. Pulko, B.; Majes, B.; Logar, J. Geosynthetic-encased stone columns: Analytical calculation model. Geotext.
Geomembr. 2011, 29, 2939. [CrossRef]
18. Castro, J.; Sagaseta, C. Consolidation and deformation around stone columns: Numerical evaluation of
analytical solutions. Comput. Geotech. 2011, 38, 354362. [CrossRef]
19. Balaam, N.P.; Booker, J.R. Effect of stone column yield on settlement of rigid foundations in stabilized clay.
Int. J. Numer. Anal. Methods Geomech. 1985, 9, 331351. [CrossRef]
20. Barron, R.A. Consolidation of fine-grained soils by drain wells. Trans. ASCE 1948, 113, 718742.
21. Hansbo, S. Consolidation of fine-grained soils by prefabricated drains. In Proceedings of the 10th
International Conference Soil Mechanics and Foundation Engineering, Stockholm, Sweden, 1519 June 1981;
Volume 3, pp. 677682.
22. Han, J.; Ye, S.L. Simplified method of consolidation rate of stone column reinforced foundation. J. Geotech.
Geoenviron. Eng. ASCE 2001, 127, 597603. [CrossRef]
23. Castro, J.; Sagaseta, C. Consolidation around stone columns. Influence of column deformation. Int. J. Numer.
Anal. Methods Geomech. 2009, 33, 851877. [CrossRef]
24. Pulko, B.; Logar, J. Fully coupled solution for the consolidation of poroelastic soil around elastoplastic stone
column. Acta Geotech. 2016. [CrossRef]
25. Castro, J.; Sagaseta, C. Influence of elastic strains during plastic deformation of encased stone columns.
Geotext. Geomembr. 2013, 37, 4553. [CrossRef]
26. Lo, S.R.; Zhang, R.; Mak, J. Geosynthetic-encased stone columns in soft clay: A numerical study. Geotext.
Geomembr. 2010, 28, 292302. [CrossRef]
27. Hosseinpour, I.; Riccio, M.; Almeida, M.S.S. Numerical evaluation of a granular column reinforced by
geosynthetics using encasement and laminated disks. Geotext. Geomembr. 2014, 42, 363373. [CrossRef]
28. Hosseinpour, I.; Almeida, M.S.S.; Riccio, M. Full-scale load test and finite-element analysis of soft ground
improved by geotextile-encased granular columns. Geosynth. Int. 2015, 22, 428438. [CrossRef]
29. Basack, S.; Indraratna, B.; Rujikiatkamjorn, C. Modelling the performance of stone column reinforced soft
ground under static and cyclic loads. J. Geotech. Geoenviron. Eng. ASCE 2016, 142, 04015067. [CrossRef]
30. Tan, S.A.; Tjahyono, S.; Oo, K.K. Simplified plane-strain modeling of stone-column reinforced ground.
J. Geotech. Geoenviron. Eng. ASCE 2008, 134, 185194. [CrossRef]
31. Khabbazian, M.; Kaliakin, V.N.; Meehan, C.L. Column Supported Embankments with Geosynthetic Encased
Columns: Validity of the Unit Cell Concept. Geotech. Geol. Eng. 2015, 33, 425442. [CrossRef]
32. Indraratna, B.; Ngo, N.T.; Rujikiatkamjorn, C.; Sloan, S.W. Coupled discrete elementfinite difference method
for analysing the load-deformation behaviour of a single stone column in soft soil. Comput. Geotech. 2015, 63,
267278. [CrossRef]
33. Cimentada, A.; da Costa, A.; Caizal, J.; Sagaseta, C. Laboratory study on radial consolidation and
deformation in clay reinforced with stone columns. Can. Geotech. J. 2011, 48, 3652. [CrossRef]
34. Miranda, M.; Da Costa, A.; Castro, J.; Sagaseta, C. Influence of geotextile encasement on the behaviour of
stone columns: Laboratory study. Geotext. Geomembr. 2017, 45, 1422. [CrossRef]
35. Priebe, H.J. Design of vibro replacement. Ground Eng. 2005, 38, 2527.
36. Sehn, A.L.; Blackburn, J.T. Predicting performance of aggregate piers. In Proceedings of the 23rd Central
Pennsylvania Geotechnical Conference, Hershey, PA, USA, 2830 May 2008.
37. Stuedlein, A.W.; Holtz, R.D. Displacement of Spread Footings on Aggregate Pier Reinforced Clay. J. Geotech.
Geoenviron. Eng. 2014, 140, 3645. [CrossRef]
38. Castro, J. An analytical solution for the settlement of stone columns beneath rigid footings. Acta Geotech.
2016, 11, 309324. [CrossRef]
39. Yu, Y.; Bathurst, R.J.; Damians, I.P. Modified unit cell approach for modelling geosynthetic-reinforced
column-supported embankments. Geotext. Geomembr. 2016, 44, 332343. [CrossRef]
40. Van Impe, W.; De Beer, E. Improvement of settlement behaviour of soft layers by means of Stone columns.
In Proceedings of the 8th European Conference on Soil Mechanics and Foundation Engineering, Helsinki,
Finland, 2326 May 1983; Volume 1, pp. 309312.
Materials 2017, 10, 782 21 of 23

41. Castro, J.; Sagaseta, C. Discussion of Simplified plane-strain modeling of stone-column reinforced ground.
J. Geotech. Geoenviron. Eng. ASCE 2010, 136, 892894. [CrossRef]
42. Weber, T.M.; Springman, S.M.; Gb, M.; Racansky, V.; Schweiger, H.F. Numerical modelling of stone columns
in soft clay under an embankment. In Geotechnics of Soft SoilsFocus on Ground Improvement; Taylor & Francis:
London, UK, 2009; pp. 305311.
43. Hird, C.C.; Pyrah, I.C.; Russell, D. Finite element modelling of vertical drains beneath embankments on soft
ground. Gotechnique 1992, 42, 499511. [CrossRef]
44. Indraratna, B.; Redana, I.W. Plane-strain modeling of smear effects associated with vertical drains. J. Geotech.
Geoenviron. Eng. ASCE 1997, 123, 474478. [CrossRef]
45. Zhang, Z.; Han, J.; Ye, G. Numerical investigation on factors for deep-seated slope stability of stone
column-supported embankments over soft clay. Eng. Geol. 2014, 168, 104113. [CrossRef]
46. Kitazume, M.; Okano, K.; Miyajima, S. Centrifuge model tests on failure envelope of column type deep
mixing method improved ground. Soils Found. 2000, 40, 4355. [CrossRef]
47. Navin, M.P.; Filz, G.M. Reliability of Deep Mixing Method Columns for Embankment Support. In GeoCongress
2006; ASCE: Atlanta, GA, USA, 2006.
48. Chen, J.-F.; Li, L.-Y.; Xue, J.-F.; Feng, S.-Z. Failure mechanism of geosynthetic-encased stone columns in soft
soils under embankment. Geotext. Geomembr. 2015, 43, 424431. [CrossRef]
49. Adam, D.; Schweiger, H.F.; Markiewicz, R.; Knabe, T. Euro 2008 Stadium KlagenfurtPrediction, Monitoring
and back calculation of settlement behavior. In Proceedings of the From Research to Design in European
Practice, Bratislava, Slovak Republic, 24 June 2010.
50. Hassen, G.; de Buhan, P.; Abdelkrim, M. Finite element implementation of a homogenized constitutive law
for stone column-reinforced foundation soils, with application to the design of structures. Comput. Geotech.
2010, 37, 4049. [CrossRef]
51. Gueguin, M.; Hassen, G.; de Buhan, P. Stability analysis of homogenized stone column reinforced foundations
using a numerical yield design approach. Comput. Geotech. 2015, 64, 1019. [CrossRef]
52. Hassen, G.; Gueguin, M.; de Buhan, P. A homogenization approach for assessing the yield strength properties
of stone column reinforced soils. Eur. J Mech. A/Solids 2013, 37, 266280. [CrossRef]
53. Gueguin, M.; Hassen, G.; de Buhan, P. Numerical assessment of the macroscopic strength criterion of
reinforced soils using semidefinite programming. Int. J. Numer. Methods Eng. 2014, 99, 522541. [CrossRef]
54. Ng, K.S.; Tan, S.A. Simplified homogenization method in stone column designs. Soils Found. 2015, 55,
154165. [CrossRef]
55. Schweiger, H.F.; Pande, G.N. Numerical analysis of stone column supported foundations. Comput. Geotech.
1986, 2, 347372. [CrossRef]
56. Lee, J.S.; Pande, G.N. Analysis of stone-column reinforced foundations. Int. J. Numer. Anal. Methods Geomech.
1998, 22, 10011020. [CrossRef]
57. Becker, P.; Karstunen, M. Volume averaging technique in numerical modelling of floating deep mixed
columns in soft soils. In Installation Effects in Geotechnical Engineering; CRC Press: Boca Raton, FL, USA, 2013;
pp. 198204.
58. Mitchell, J.K.; Huber, T.R. Performance of a stone column foundation. J. Geotech. Eng. ASCE 1985, 111,
205223. [CrossRef]
59. Elshazly, H.; Elkasabgy, M.; Elleboudy, A. Effect of Inter-Column Spacing on Soil Stresses due to
Vibro-Installed Stone Columns: Interesting Findings. Geotech. Geol. Eng. 2008, 26, 225236. [CrossRef]
60. Ng, K.S. Numerical Study of Floating Stone Columns. Ph.D. Thesis, National University of Singapore,
Singapore, 2013.
61. Castro, J. Numerical modelling of stone columns beneath a rigid footing. Comput. Geotech. 2014, 60, 7787.
[CrossRef]
62. Kirsch, F.; Sondermann, W. Field measurements and numerical analysis of the stress distribution below
stone column supported embankments and their stability. In Proceedings of the International Workshop on
Geotechnics of Soft Soils-Theory and Practice, Noordwijkerhout, The Netherlands, 1719 September 2003.
63. Hughes, J.M.O.; Withers, N.J. Reinforcing of soft soils with stone columns. Ground Eng. 1974, 7, 4249.
[CrossRef]
64. Hughes, J.M.O.; Withers, N.J.; Greenwood, D.A. A field trial of the reinforcing effect a stone column in soil.
Gotechnique 1975, 25, 3144. [CrossRef]
Materials 2017, 10, 782 22 of 23

65. Wood, D.M.; Hu, W.; Nash, D.F.T. Group effects in stone column foundations: Model tests. Gotechnique 2000,
50, 689698. [CrossRef]
66. Mitchell, J.K. Soil improvement: State-of-the-art report. In Proceedings of the 10th International Conference
on Soil Mechanics and Foundation Engineering, Stockholm, Sweden, 1519 June 1981; Volume 4, pp. 509565.
67. Watts, K.S.; Johnson, D.; Wood, L.A.; Saadi, A. An instrumented trial of vibro ground treatment supporting
strip foundations in a variable fill. Gotechnique 2000, 50, 699708. [CrossRef]
68. Schweiger, H.F. Finite Element Analysis of Stone Column Reinforced Foundations. Ph.D. Thesis, TU Graz,
Graz, Austria, 1989.
69. Castro, J. Groups of encased stone columns: Influence of column length and arrangement. Geotext. Geomembr.
2017, 45, 6880. [CrossRef]
70. Wehr, J. Stone columnsSingle columns and group behaviour. In Proceedings of the 5th International
Conference on Ground Improvement Technologies, Kuala Lumpur, Malaysia, 2223 March 2004; pp. 329340.
71. Babu, M.R.D.; Nayak, S.; Shivashankar, R. A critical review of construction, analysis and behaviour of stone
columns. Geotech. Geol. Eng. 2013, 31, 122. [CrossRef]
72. Black, J.A.; Sivakumar, V.; Bell, A. The settlement performance of stone column foundations. Gotechnique
2011, 61, 909922. [CrossRef]
73. Malarvizhi, S.N.; Ilamparuthi, K. Comparative study on the behaviour of encased stone column and
conventional stone column. Soils Found. 2007, 47, 873885. [CrossRef]
74. Ali, K.; Shahu, J.T.; Sharma, K.G. Model tests on geosynthetic-reinforced stone columns: A comparative
study. Geosynth. Int. 2012, 19, 433451. [CrossRef]
75. Kirsch, F. Vibro stone column installation and its effect on ground improvement. In Numerical Modelling of
Construction Processes in Geotechnical Engineering for Urban Environment; Taylor & Francis: London, UK, 2006;
pp. 115124.
76. Guetif, Z.; Bouassida, M.; Debats, J.M. Improved soft clay characteristics due to stone column installation.
Comput. Geotech. 2007, 34, 104111. [CrossRef]
77. Castro, J.; Karstunen, M. Numerical simulations of stone column installation. Can. Geotech. J. 2010, 47,
11271138. [CrossRef]
78. Castro, J.; Karstunen, M.; Sivasithamparam, N. Influence of stone column installation on settlement reduction.
Comput. Geotech. 2014, 59, 8797. [CrossRef]
79. McCabe, B.A.; Nimmons, G.J.; Egan, D. A review of field performance of stone columns in soft soils.
Proc. ICE-Geotech. Eng. 2009, 162, 323334. [CrossRef]
80. Castro, J.; Sagaseta, C. Pore pressure during stone column installation. Proc. ICE-Ground Improv. 2012, 165,
97109. [CrossRef]
81. Aboshi, H.; Ichimoto, E.; Enoki, M.; Harada, K. The CompozerA method to improve characteristics of
soft clays by inclusion of large diameter sand columns. In Proceedings of the International Conference on
Soil Reinforcement: Reinforced Earth and Other Techniques, Paris, France, 2022 March 1979; Volume 1,
pp. 211216.
82. Weber, T.M.; Pltze, M.; Laue, J.; Peschke, G.; Springman, S.M. Smear zone indentification and soil properties
around stone columns constructed in-flight in centrifuge model tests. Gotechnique 2010, 60, 197206.
[CrossRef]
83. Indraratna, B.; Basack, S.; Rujikiatkamjorn, C. Numerical solution to stone column reinforced soft ground
considering arching, clogging and smear effects. J. Geotech. Geoenv. Eng. ASCE 2013, 139, 377394. [CrossRef]
84. Bergado, D.T.; Asakami, H.; Alfaro, M.C.; Balasubramaniam, A.S. Smear effects of vertical drains on soft
Bangkok clay. J. Geotech. Eng. ASCE 1991, 117, 15091530. [CrossRef]
85. Indraratna, B.; Redana, I.W. Laboratory determination of smear zone due to vertical drain installation.
J. Geotech. Geoenv. Eng. ASCE 1998, 124, 180184. [CrossRef]
86. Rujikiatkamjorn, C.; Indraratna, B. Design procedure for vertical drains considering a linear variation of
lateral permeability within the smear zone. Can. Geotech. J. 2009, 46, 270280. [CrossRef]
87. Wood, D.M. Soil Behaviour and Critical State Soil Mechanics; Cambridge University Press: Cambridge, UK,
1990; p. 462.
88. Brand, E.W.; Brenner, R.P. Soft Clay Engineeering; Elsevier: Amsterdam, The Netherlands, 1981; p. 779.
89. Pugh, R.S. Settlement of floor slabs on stone columns in very soft clays. Proc. ICE-Geotech. Eng. 2017, 170,
1626. [CrossRef]
Materials 2017, 10, 782 23 of 23

90. Raymond, G.P.; Davies, J.R. Triaxial tests on dolomite railroad ballast. J. Geotech. Eng. ASCE 1978, 104,
737751.
91. Indraratna, B.; Ionescu, D.; Christie, H.D. Shear behavior of railway ballast based on large-scale triaxial tests.
J. Geotech. Geoenviron. Eng. ASCE 1998, 124, 439449. [CrossRef]
92. Charles, J.; Watts, K. The influence of confining pressure on the shear strength of compacted rockfill.
Gotechnique 1980, 30, 353367. [CrossRef]
93. Herle, I.; Wehr, J.; Arnold, M. Influence of pressure level and relative density on friction angle of gravel in
vibrated stone column. In Pfahl-Symposium 2007; Mitteilung des Instituts fr Grundbau und Bodenmechanik,
TU Braunschweig: Braunschweig, Germany, 2007; pp. 8193.
94. Rowe, P.W. The relation between the shear strength of sands in triaxial compression, plane strain and direct
shear. Gotechnique 1969, 19, 7586. [CrossRef]
95. Castro, J.; Cimentada, A.; da Costa, A.; Caizal, J.; Sagaseta, C. Consolidation and deformation around stone
columns: Comparison of theoretical and laboratory results. Comput. Geotech. 2013, 49, 326337. [CrossRef]
96. Killeen, M.; McCabe, B.A. Settlement performance of pad footings on soft clay supported by stone columns:
A numerical study. Soils Found. 2014, 54, 760776. [CrossRef]
97. Yoo, C. Performance of geosynthetic-encased stone columns in embankment construction: Numerical
investigation. J. Geotech. Geoenviron. Eng. ASCE 2010, 136, 11481160. [CrossRef]
98. Almeida, M.S.S.; Hosseinpour, I.; Riccio, M. Performance of a geosynthetic-encased column (GEC) in soft
ground: Numerical and analytical studies. Geosynth. Int. 2013, 20, 252262. [CrossRef]
99. Liu, K.-W.; Rowe, R.K.; Su, Q.; Liu, B.; Yang, Z. Long-term reinforcement strains for column supported
embankments with viscous reinforcement by FEM. Geotext. Geomembranes 2017, 45, 307319. [CrossRef]
100. Castro, J. Discussion of Column Supported Embankments with Geosynthetic Encased Columns: Validity of
the Unit Cell Concept. Geotech. Geol. Eng. 2016, 34, 419420. [CrossRef]
101. Soderman, K.L.; Giroud, J.P. Relationships between uniaxial and biaxial stresses and strains in geosynthetics.
Geosynth. Int. 1995, 2, 495504. [CrossRef]
102. Castro, J.; Sagaseta, C.; Caizal, J.; Da Costa, A. Modelizacin de columnas de grava. In Proceedings of the
10 Simposio Nacional de Ingeniera Geotcnica, La Corua, Spain, 1921 October 2016; Sociedad Espaola
de Mecnica del Suelo e Ingeniera Geotcnica: Madrid, Spain, 2016; Volume II, pp. 567588. (In Spanish)

2017 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).