Computers and Geotechnics 151 (2022) 104931

Contents lists available at ScienceDirect

Computers and Geotechnics

journal homepage: www.elsevier.com/locate/compgeo

Installation of groups of stone columns in clay: 3D Coupled Eulerian

Lagrangian analyses
Atefe Geramian a, Jorge Castro b, *, Mahmoud Ghazavi a, Marina Miranda b
a
Faculty of Civil Engineering, K. N. Toosi University of Technology, Tehran, Iran
b
Department of Ground Engineering and Materials Science, Universidad de Cantabria, Santander, Spain

A R T I C L E I N F O A B S T R A C T

Keywords: This paper describes the results of three-dimensional (3D) finite element analyses investigating the installation
Stone columns effects of groups of stone columns in purely cohesive soils. Installation of stone columns is simplified to the
Installation insertion of rigid cylindrical elements with conical tips in a single homogeneous soil layer (Tresca plasticity and a
Sequence
quasi-incompressible elastic law). Installation of a single column is simulated as the reference case and the in­
Radial stress
Finite element analyses
stallations of two columns and a group of nine columns are considered to study the interaction between the
Coupled Eulerian Lagrangian formulation installation of several columns and the influence of the installation sequence. The process is simulated using a
Coupled Eulerian Lagrangian formulation. Stone column installation alters the surrounding soil and the nu­
merical results show the increase in horizontal stresses and pore pressures. The installation effects of several
columns at common spacings overlap between each other and accumulate, producing higher horizontal stresses
and pore pressures in a larger area. The installation sequence is mainly visible around the last column installed,
where the radial stresses are lower.

1. Introduction of total and differential settlements, acceleration of consolidation,

improvement of the stability, and reduction of liquefaction potential (e.
The stone column technique is one of the most widely used soil g., Barksdale and Bachus 1983; Han 2015). Besides, column installation
improvement techniques in geotechnical engineering practise (e.g., alters the properties of the soil surrounding the column, for example,
Barksdale and Bachus 1983; Han 2015). Stone columns are commonly increasing horizontal stresses and pore pressures (e.g., Kirsch 2004).
employed to improve weak soils, such as soft clays. They are vertical Therefore, accounting for these installation effects is important to ach­
boreholes in the ground filled with crushed stone or gravel and are ieve safe and accurate designs (e.g., Egan et al. 2008; Indraratna et al.
normally constructed using a deep vibrator, either electric or hydraulic, 2013; Castro et al. 2014). In this paper, the analysis focuses on purely
that penetrates the ground and later compacts the gravel or crushed cohesive materials because stone columns are usually installed in soft
stone in stages from the base of the hole upwards. The two most common clays in a relatively short period of time (around 15–30 min per column)
construction methods are: vibro-replacement (also called “wet (e.g., Castro and Sagaseta 2012). Consequently, stone column installa­
method”), and vibro-displacement (also known as “dry method”) (e.g., tion in clays can be considered to be an undrained process. In soils with
McCabe et al. 2009; Kirsch and Kirsch 2010). In these methods, the deep higher permeabilities (e.g., some silt content or interbedded sandy
vibrator (poker) penetrates by vibration and its own weight, helped by layers), partial drainage and a faster dissipation of excess pore pressures
bottom jets of either water (“wet method”) or compressed air (“dry will take place.
method”). Field measurements (e.g., Watts et al. 2000; Watts et al. 2001; Kirsch
Stone columns increase the overall strength and stiffness of a foun­ 2004; Gäb et al. 2007; Castro and Sagaseta 2012; McCabe et al. 2013;
dation system because the added gravel or crushed stone has superior Amoroso et al. 2015) have shown some of the effects of column instal­
mechanical properties than those of the existing natural soft soil. In this lation, such as increases in pore pressures and horizontal stresses,
way, their main effects are: improvement of bearing capacity, reduction ground heave and soil remoulding. Some of these installation effects (e.

* Corresponding author at: Group of Geotechnical Engineering, Department of Ground Engineering and Materials Science, Universidad de Cantabria, Avda. de Los
Castros, s/n, 39005 Santander, Spain.
E-mail addresses: [email protected] (A. Geramian), [email protected] (J. Castro), [email protected] (M. Ghazavi), [email protected]
(M. Miranda).

https://doi.org/10.1016/j.compgeo.2022.104931
Received 27 April 2022; Received in revised form 5 July 2022; Accepted 20 July 2022
Available online 10 August 2022
0266-352X/© 2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
A. Geramian et al. Computers and Geotechnics 151 (2022) 104931

Fig. 1. General view of the numerical model and the finite element mesh.

Table 2
Summary of numerical simulations.
Number of columns Cases Results

1 R = 0.5 m Figs. 4-6

R = 0.3, 0.4, 0.5, 0.6 m Fig. 14
2 R = 0.5 m Figs. 7-9
R = 0.3, 0.4, 0.5, 0.6 m Fig. 15
9 Outside in Figs. 10-13
Inside out

Weber et al. 2010). In conclusion, experimental evidences of installation

effects are case specific, difficult to obtain, expensive, and consequently,
scarce.
On the other hand, numerical modelling may be used to study
installation effects of stone columns. In this regard, most studies (e.g.,
Guetif et al. 2007; Castro and Karstunen 2010; Sexton and McCabe 2015;
Fig. 2. Detail of the finite element mesh at the surface and the cylinder tip. Nagula et al. 2018) have considered the installation effects of a single
column in axial symmetry for simplicity and have modelled column
installation as a cylindrical cavity expansion. More advanced ap­
Table 1
Soil properties.
proaches, such as the press-replace method or specific large displace­
ment formulations, have also been used (e.g., Farias et al. 2005; Wang
Soil property Value Units
and Li 2019), particularly, for analogous problems, such as pile instal­
Mass bulk density (ρ) 2,000 kg/m3 lation (e.g., Pucker and Grabe 2012; Tehrani et al. 2016; De Chaunac
Undrained shear strength (cu) 30 kPa and Holeyman 2018), but they mainly consider the installation of only
Undrained Young’s modulus (Eu) 3,000 kPa
Undrained Poisson’s ratio (νu) 0.495
one column or pile. Besides, the installation effects are sometimes used
Earth pressure coefficient at rest (K0) 0.6 / 0.8 as input data of additional independent (uncoupled) numerical models
Henkel’s a pore pressure parameter 0.3 to study the stone column or pile performance (e.g., Castro et al. 2014;
Skempton’s A pore pressure parameter 0.75 Nagula et al. 2018; Al Ammari and Clarke 2018; Karlsson et al., 2019).
Stone columns are closely spaced (e.g., around 2 m with area
placement ratios in the range 15–30 %, Barksdale and Bachus 1983)
g., the increase in horizontal stresses) have also been estimated by nu­
and, consequently, their installation effects interact and overlap.
merical back-analysis of stone column performance in the field (e.g.,
Moreover, the installation sequence influences the soil disturbance as
settlement measurements) (e.g., Elshazly et al., 2008; Al Ammari and
known in practice by construction companies and numerically simulated
Clarke, 2018). The reliability of this approach depends on the appro­
for similar problems, such as for rigid inclusions or piles (e.g., Nguyen
priateness of the numerical predictions (e.g., parameter calibration and
et al. 2019; Soleimani and Weissenfels 2021). To the best of the authors’
constitutive models). Simulating the installation process of stone col­
knowledge, numerical modelling of the installation of several stone
umns in the laboratory generally requires using a centrifuge, and
columns have only been performed under highly oversimplified as­
consequently, only few cases have been published (e.g., Lee et al. 2004;
sumptions (e.g., Kirsch 2006; Ellouze et al. 2017; Al Ammari and Clarke

2
A. Geramian et al. Computers and Geotechnics 151 (2022) 104931

Fig. 3. Installation sequences, reference point and reference cross-section.

Fig. 4. Radial total stresses for the penetration of a single rigid cylinder (d = 4 m).

2018) or as a preliminary analysis using the Coupled Eulerian Castro and Karstunen 2010) normally use updated Lagrangian formu­
Lagrangian framework (CEL) (Nagula and Grabe 2018). lations (e.g., McMeeking and Rice 1975) to account for the large dis­
In this paper, numerical simulations of the installation effects of placements that occur during installation. For the present case, the mesh
several stone columns are carried out using CEL finite elements to study would be largely distorted before the installation of the next column. To
the interaction between the installation effects of adjacent columns and overcome this mesh distortion problem, the CEL formulation imple­
the influence of the installation sequence. The results show that the mented in the finite element code ABAQUS/Explicit (Dassault Systémes,
horizontal stresses and pore pressures are higher and spread over a 2020) was used to develop the numerical model (Fig. 1). A brief intro­
wider area when more columns are installed. In this way, Section 2 duction to the CEL method and the advantages of using an explicit time
presents the numerical model. Next, the numerical results are presented integration scheme in this type of problem may be found, for instance, in
and discussed (Section 3), and finally, some conclusions are drawn. Pucker and Grabe (2012).
The stone column installation process consists of vibrator penetra­
2. Numerical model tion, followed by the gravel compaction in stages, which further en­
larges the column diameter (e.g., Barksdale and Bachus 1983). For the
Finite element simulations of the installation of a single column (e.g., numerical model, stone column installations have been simplified to

3
A. Geramian et al. Computers and Geotechnics 151 (2022) 104931

Fig. 6. Comparison between numerical simulations (d = z = 4 m) and cylin­

drical cavity expansion solution: (a) radial stresses and (b) tangential stresses.

with a flat tip and the differences in any of the results, not very close to
the tip, were not notable in the surrounding soil as shown in the Sup­
plementary Material for the horizontal total stresses.
The cylindrical elements were inserted into a single homogeneous
Fig. 5. Total stresses with radial distance at several depths (d) for the pene­
soft soil layer for simplicity (Fig. 1). A common thickness of 8 m was
tration of a single rigid cylinder (z = 4 m): (a) radial stresses and (b) tangen­
assumed for the soil layer and a rigid bedrock appeared beneath it, the
tial stresses.
columns fully penetrating in the soft soil layer. The soil was discretised
into 8-node linear hexahedral Eulerian elements (with reduced inte­
quasi-static insertions of rigid cylindrical elements to reduce the gration and hourglass control). The bottom boundary of the soil was
computational cost in this novel study on the interaction between the fixed to reproduce the contact with the rigid bedrock and roller vertical
installation of several columns. Besides, the main effect of column boundaries were used for the soil sides. A model width of 32 m (Fig. 1)
installation in clays is the quasi-static lateral expansion (e.g., Castro and was enough to avoid the influence of the lateral boundaries.
Karstunen 2010). The diameter of the cylindrical elements has been set The soft soil layer was modelled using Tresca plasticity and a quasi-
as 1 m (i.e., a radius, R, of 0.5 m) and an axis-to-axis spacing of 2 m in a incompressible elastic law, because stone columns are usually installed
square pattern has been considered since they are round numbers within in soft cohesive soils and the installation process is fast enough (un­
the range of common stone column diameters and spacings. The drained conditions). Common properties were chosen for the soil,
resulting area replacement ratio (e.g., Barksdale and Bachus 1983) is namely an undrained shear strength (cu) of 30 kPa, an undrained
19.6 %. The tip of the cylindrical elements has been chosen as conical Young’s modulus (Eu) of 3,000 kPa, an undrained Poisson’s ratio (νu) of
with a cone angle of 90◦ (Fig. 2), which may be assumed as usual for 0.495 and a saturated unit weight (γ sat) of 19.62 kN/m3 (Table 1). The
deep vibrators (e.g., Kirsch and Kirsch 2010).The height of the conical latter one was chosen just as a round value of 2 kg/m3, which is some­
tip coincides with the height of the upper 0.5-thick void layer (Eulerian what high for a soft soil. The linear elastic perfectly plastic behaviour
mesh as detailed below), resulting in a simpler mesh; therefore, part of with a Tresca failure criterion was modelled in ABAQUS using the so-
this upper 0.5-thick layer is not shown in Fig. 2 for visualization pur­ called “Mohr-Coulomb” constitutive model with a negligible friction
poses of the conical tip. The tip of the cylinder is initially placed just at angle, namely 0.1. The specific position of the ground water table was
the soil surface level. 8-node brick Lagrangian elements were used for not modelled since it is not necessary for this case (undrained condi­
the cylinder mesh. Additional numerical simulations were performed tions), and the ratio of horizontal total stress to vertical total stress, at

4
A. Geramian et al. Computers and Geotechnics 151 (2022) 104931

SC1 SC2
0.5 m (void)

d=8 m
50 kPa

σr=100 kPa

150 kPa

Fig. 7. Radial total stresses for the full penetration of two rigid cylinders (d = 8 m).

initial state, was considered 0.8. This is equivalent to setting the ground well known (e.g., Pucker and Grabe 2012), the radial total stresses in­
water table at the ground surface and a lateral earth pressure coefficient crease around the penetrating tip. For a specific depth (z), the maximum
(K0) of 0.6, i.e., provides the same initial state of total stresses and radial stress is reached when the cylinder tip reaches that depth (d = z)
consequently, the same results. For example, at a depth of 4 m with (Fig. 5). When the rigid cylinder penetrates further, the radial stresses
water level at the ground surface, the vertical total and effective stresses decrease in the vicinity, even below the initial value, σr0 = 62.78 kPa
are 78.5 kPa and 38.5 kPa, respectively. By considering K0 = 0.6, the (Fig. 4 and Fig. 5a). The maximum (peak) value of the radial total stress
horizontal effective stress is 23 kPa, and consequently the horizontal is at the cylinder wall when d ≤ z and, when d > z, the peak moves away
total stress is 62.8 kPa, which means that the ratio of horizontal total from the cylinder wall (e.g., in an approximate range of 3-5R for the
stress to vertical one is 0.8. cases shown in Fig. 5a). Bond and Jardine (1991) already measured a
The numerical simulations started by generating the initial stress steady increase in the radial total stress at a specific depth during pile
state using γsat and K0. Next, the rigid cylinders were sequentially pushed penetration until the tip reached that depth, and a steady decrease as the
vertically down into the soil until their tip reached the bottom boundary pile tip advanced to greater depths. It is worth noting that for a pile, the
of the model (rigid substratum). Several installation configurations stresses at the pile wall are the most relevant ones, while for stone col­
(Table 2) were simulated to study the influence of the installation umns, the average ones between the columns are the important ones.
sequence and interaction effects. Firstly, a single column was installed as The tangential stresses at a specific depth also reach a maximum
the reference case and to compare with analytical estimations. Next, the when the tip is at this depth (Fig. 5b). In these results, the elastic region
installation of two columns was modelled and later, the installation of a (r > 6R), where the tangential stresses decrease, is clearly visible. When
group of 9 columns was studied. In this latter case, two installation se­ the cylinder tip further penetrates below the studied depth (d > z), the
quences were studied: “Outside in” and “Inside out” (Fig. 3). In fact, the tangential stresses decrease, but just in the close vicinity (e.g., r < 3R).
models with one and two columns are just the initial phases (i.e., the The more the cylinder penetrates, i.e. larger values of d, the larger the
installation of the first and second columns) of the model with 9 columns plastic annulus is at a specific depth (e.g., z = 4 m in Fig. 5b) and the
(“Outside in”) to avoid repetition and reduce the computational cost. more the peak of tangential stresses moves away from the cylinder wall.
Finally, a parametric study on the influence of the column radius, i.e., The maximum values of the radial and tangential stresses, i.e., when
penetrating cylinder radius (R), has been performed for the installation d = z, simulated numerically compare relatively well with the analytical
of 1 and 2 columns. The studied values have been chosen within the values obtained using the cylindrical cavity expansion solution (e.g.,
common range, namely 0.3, 0.4, 0.5 (reference case) and 0.6 m Baguelin et al. 1978) (Fig. 6).
(Table 2). ( r)
σr = p0 + cu 1 + 2ln p Plastic zone (r < rp )
r (1)
3. Results and discussion σθ = σ r − 2cu
3.1. Single column installation rp 2
σ r = p0 + c u
r2
The installation of a single isolated column serves as the reference Elastic zone (r ≥ rp ) (2)
rp 2
case and the first step when simulating the installation of several col­ σθ = p0 − cu
umns (Fig. 3). The radial total stress contours are plotted in Fig. 4, when r2
the depth of the cylinder tip (d) is 4 m (half of the soft soil layer thick­ where for this case, cu =30 kPa and p0 =σ r0 =σ θ0 = 62.78 kPa, and rp may
ness). The cylindrical coordinates are centered at the column axis. As
be analytically obtained as:

5
A. Geramian et al. Computers and Geotechnics 151 (2022) 104931

rigid cylinders (d = 8 m). The cylindrical coordinates are centered at the

first column axis, but the results are the same for any other position
within that plane, because the radial stresses are the horizontal in-plane
stresses, while the tangential stresses are the out of plane horizontal
stresses, i.e., normal to the plane. The results (Fig. 7) clearly show that
the installation order plays an important role and the effects are not
symmetric, i.e., the stresses around the column that is installed first
(SC1) are different from those around the second column (SC2).
Fig. 8 compares the radial and tangential stresses after the installa­
tion of the first and second columns (SC1 and SC2). For the sake of
comparison, the normalized radial distance is plotted from the axis of
the last column installed in each case, i.e., the axis is at the centre of the
first column for SC1 and for the installation of the second column (SC2),
the results are horizontally displaced, being the axis at the centre of the
second column and the first column is now on the left. These stresses are
also plotted centred at the axis of the first column in the Supplementary
Material. The stresses near the last column (r < 2-3R) are similar for both
installations. On the other hand, the installation of the second column
changes the radial total stresses near the first column and further in­
creases the radial total stresses beyond r > 3R (Fig. 8a). Besides, the peak
of radial total stress for the installation of the second column is slightly
larger and at a slightly larger distance from the column axis than those
for the installation of the first column. Similarly, the tangential total
stresses increase approximately in the range 3R < r < 8R with the
installation of the second column and the plastic annulus is enlarged
(Fig. 8b).
To highlight the non-symmetric and enlarged plastic zone (and
larger plastic strains) that occur when the second column is installed,
Fig. 9 compares the equivalent plastic strains (PEEQVAVG) during the
installation process of the first and second columns. The equivalent
√̅̅̅̅̅̅̅̅̅̅̅
plastic strain is defined as εpeq = 23εpij εpij , and in this case (Tresca mate­
√̅̅̅
rial, i.e., null plastic volumetric strains), it is equal to γp / 3. For visu­
alization purposes, the plastic bulbs are mainly in red (grey/red contact
is at 5 % of equivalent plastic strains).

3.3. Installation of a group of nine columns

To study the installation effects of a group of stone columns and the

influence of the installation sequence, a group of 9 columns was studied,
Fig. 8. Total stresses with radial distance for the penetration of two rigid cyl­ considering two installation sequences, namely “Outside in” and “Inside
inders (z = 4 m): (a) radial stresses and (b) tangential stresses.
out” (Fig. 3). The cylindrical coordinate system is centered at the central
column axis. Fig. 10 presents the radial (horizontal in-plane) and
rp √̅̅̅̅̅̅̅̅̅̅
= G/cu ≈ 5.8 (3) tangential (out-of-plane) stresses on the representative cross section
R marked in Fig. 3 at z = 4 m. The radial total stress further increases with
The differences between the numerical simulations and the analyt­ the installation of more columns (Fig. 10a). For instance, the peak value
ical cylindrical cavity expansion solution may be attributed to the is around 150 kPa for the installation sequence “Outside in” and 136 kPa
following two facts: (a) a free surface is numerically simulated; then, for “Inside out”, in comparison with the approximately 80 and 90 kPa
plane strain conditions in the vertical direction do not hold (for example, for one and two columns, respectively. The peak values are reached
the vertical stress from the numerical simulation is not constant in the outside the column group, but close to the outer columns of the group, at
elastic domain, see Supplementary Material); (b) the numerical simu­ a distance around 2-4R. Similar comments apply to the tangential total
lation accounts for the tip effects of the penetrating element. In stresses (Fig. 10b). In the Supplementary Material, a qualitative com­
conclusion, the analytical solution (cylindrical cavity expansion theory) parison with Kirsch (2006) data is presented.
reproduces reasonably well the situation when the column tip is at the The “Outside in” configuration is expected to create higher
studied depth (Fig. 6), but not when the column penetrates further confinement and radial stresses in the natural soil than the “Inside out”
causing soil remoulding and a decrease of radial stresses near the col­ configuration (as found, for example, for piles by Le Kouby et al. 2016).
umn (Fig. 5). However, from results in Fig. 10, this is not very clear. To further study
the influence of the installation sequence on the final radial total
3.2. Installation of two columns stresses, the stress contours of a plan view for z = 4 m are shown in
Fig. 11. The contours clearly show that the installation of the last column
The installation of a second stone column near an already installed (9) leaves a “remoulded” zone and lower radial stresses in its
column further alters the stress state of the natural soil. The installation surroundings.
effects of both columns interact because common column spacings (e.g., On the other hand, the overall differences do not seem important on
1.5–3 m, 3-6R) are smaller than the area of influence of the installation average, but the “Outside in” sequence generates a better improved area
of one column (>10R). Fig. 7 shows the total radial stress contours after surrounding the group of columns, i.e., larger radial stresses providing a
the installation of two columns, modelled as the full penetration of two better lateral confinement, in the line of the proposal used by Kirsch

6
A. Geramian et al. Computers and Geotechnics 151 (2022) 104931

Fig. 9. Equivalent plastic strains at different penetration depths: (a) First column penetration (SC1) and (b) Second column penetration (SC2).

(2006) to back fit field measurements. Besides, the installation effects of (Fig. 3) at z = 4 m. In the horizontal axis of Fig. 13, the number of the
the “Inside out” configuration are clearly not symmetric, which could installed column is indicated. For example, 1 indicates that the first
later result in non-uniform settlements of a footing on top of the group of column has reached the maximum depth (SC1, d = 8 m) and the
columns, even under centered loads. installation of the second column starts (SC2, d = 0 m), or 1.5 indicates
The numerical simulations are performed using total stresses, but the that the tip of the second column is at half of its maximum depth (SC2, d
excess pore pressures (Δu) may be estimated using Henkel’s equation = 4 m). From the figure and its comparison with field measurements, the
and the numerically simulated variations of total stresses: following comments may be made:
Δu = Δσoct + 3a|Δτoct | (4)
• The two largest peak values correspond to the installation of the two
where soil is assumed to be saturated, Δσoct and Δτoct are the variations contiguous columns (SC1 and SC2 for “Outside in” and SC2 and SC3
of octahedral normal and shear stresses, respectively, and a is the for “Inside out”, Fig. 3) and when the tip of the cylinder is at the same
Henkel’s a pore pressure parameter. The value of a is taken as 0.3 depth as the measuring point (z = d = 4 m).
(Table 1), which corresponds to a Skempton’s A pore pressure parameter • The numerical model does not simulate the gravel placement, and
of 0.75, which is a common value for normally consolidated or slightly consequently, those pore pressures variations are not reproduced in
overconsolidated clays. Fig. 12 illustrates the excess pore pressures in the numerical results.
the reference cross-section (Fig. 3) after the installation of the nine • Important drops in pore pressure occur when the rigid cylinder/
columns and, on average, the results of both installation sequences are vibrator penetrates beneath the measuring depth. Sometimes, these
not very different. drops may partly be attributed to a quick pore pressure dissipation
In the field, excess pore pressures are usually measured using a when interpreting field measurements, but the presented numerical
piezometer, either electric or hydraulic, located between columns at a simulations are fully undrained and show the important reductions
specific depth (e.g., Castro and Sagaseta 2012). To qualitatively caused by the deeper penetration of the cylinder (soil remoulding) (e.
compare with those field measurements, the excess pore pressures g., Fig. 5a).
during the installation of the 9 columns are plotted in Fig. 13, for both • Excess pore pressures are always larger at the end of the installation
installation sequences. The excess pore pressures are calculated using of each column than at the beginning, but for the installation of the
Eq. (4) and the values corresponds to a reference point between columns next column after the two adjacent to the measuring point, namely

7
A. Geramian et al. Computers and Geotechnics 151 (2022) 104931

Fig. 10. Total stresses with radial distance at the reference cross-section for the
penetration of nine rigid cylinders (z = 4 m): (a) radial stresses and (b)
tangential stresses.

SC3 for “Outside in” and SC4 for “Inside out”. For that column, there
is an important drop when the tip is between d = 1 and 2.5 m.
• Excess pore pressures do not change for the installation of the further Fig. 11. Radial total stress contours at z = 4 m after the installation of a group
column (SC5 for “Outside in” and SC6 for “Inside out”) because it is of nine columns.
outside the area of influence for pore pressures (plastic zone).
• The peak excess pore pressures (around 40 kPa) and the residual ones pile in a group due to neighboring installations were limited by the fact
(around 25 kPa) are very similar for both installation sequences. that the soil is at critical state conditions (constant mean effective stress)
• The similarity of final excess pore pressures for both construction and that the installations of additional piles result only in an accumu­
sequences agrees with the similarity of the final radial stresses shown lation of excess pore pressures beyond the “plastic zone.”.
in Fig. 11 for the chosen reference point (Fig. 3). The results for other The study of the dissipation of these excess pore pressures is beyond
measuring points will be different, but proportional to the radial the scope of this paper, but the presence of stone columns will notably
stress values shown in Fig. 11. accelerate their dissipation, as measured in the field (e.g., Castro and
Sagaseta 2012) and numerically simulated (e.g., Castro and Karstunen
McCabe and Lehane (2006) present field measurements of horizontal 2010).
total stress during pile installation, which show similar qualitative
trends as those observed in Fig. 13 because the major component in total
stress variations is the excess pore pressure. McCabe and Lehane (2006) 3.4. Parametric study
also found that excess pore pressures generated in the vicinity of a given
The installation of 1 and 2 columns with R = 0.3, 0.4, 0.5 (reference

8
A. Geramian et al. Computers and Geotechnics 151 (2022) 104931

80

60
Excess pore pressure, Δu (kPa)

40

20

-20

Outside in
-40
Inside out
-60
-8 -6 -4 -2 0 2 4 6 8
Distance to central column axis, r (m)

Fig. 12. Excess pore pressures with radial distance for the penetration of nine
rigid cylinders (z = 4 m).

45
40 z=4m Outside in
Excess pore pressure, Δu (kPa)

Inside out
35
30
25
20
15
10
5
0
0 1 2 3 4 5 6 7 8 9
Number of installed column

Fig. 13. Excess pore pressures at a reference point (z = 4 m) during the

penetration of the nine rigid cylinders.
Fig. 14. Radial total stresses at z = 4 m for the penetration of a single cylinder
with different radii, R: (a) d = 4 m; (b) d = 8 m.
case) and 0.6 m were numerically simulated. When the normalized
horizontal distance (r/R) is used (e.g., Fig. 14), the results are nearly the
same. The small differences in Fig. 14 are mainly attributed to the
different relative depths (z/R) of each case and slight numerical
deviations.
Fig. 15 shows the radial total stresses after the installation of the two
columns with different columns radii (R) along a line that follows both
column axes at z = 4 m. For the installation of the second column, the
axis-to-axis spacing of 2 m between columns has been kept constant. The
results show that a larger column radius leads to a higher remoulding
between columns (lower radial stresses) and a slightly higher peak of
radial stresses outside the columns and at a further distance (larger
plastic zone). Besides, based on the results for the installation of 1 col­
umn, the parametric study of the column radius (R) may also be inter­
preted as a parametric study on the column spacing (s), if the horizontal
distance is normalized with the column radius. Consequently, a larger
column radius would be equivalent to a shorter column spacing.

4. Conclusions

CEL finite element simulations of the installation of a single column,

Fig. 15. Radial total stresses at z = 4 m for the full penetration (d = 8 m) of two
two columns and a group of nine columns reproduce the surrounding
cylinders with different radii, R.
soil alteration, with increases in horizontal stresses and pore pressures.
Due to the complexity of the numerical analyses, the present study has

9
A. Geramian et al. Computers and Geotechnics 151 (2022) 104931

the following limitations: a single homogeneous soil layer of purely Bond, A.J., Jardine, R.J., 1991. Effects of installing displacement piles in a high OCR
clay. Geotechnique 41 (3), 341–363.
cohesive material with constant undrained shear strength is considered
Castro, J., Karstunen, M., 2010. Numerical simulations of stone column installation. Can.
and dynamic effects, gravel placement and dissipation of excess pore Geotech. J. 47 (10), 1127–1138.
pressures are not modelled. Castro, J., Karstunen, M., Sivasithamparam, N., 2014. Influence of stone column
The numerical results show that cylindrical cavity expansion ap­ installation on settlement reduction. Comput. Geotech. 59, 87–97.
Castro, J., Sagaseta, C., 2012. Pore pressure during stone column installation. Proc. ICE -
proaches reproduce reasonably well the situation when the column tip is Ground improvement 165 (2), 97–109.
at the studied depth, but are not able to simulate the decrease of radial De Chaunac, Holeyman, H., 2018. Numerical analysis of the set-up around the shaft of a
stresses and pore pressures near the column when the tip of the cylinder closed-ended pile driven in clay. Geotechnique 68, 332–344.
Egan, D., Scott, W., McCabe, B., 2008. Installation effects of vibro replacement stone
penetrates further causing soil remoulding. For a group of stone col­ columns in soft clay. Geotechnics of Soft Soils – Focus on Ground Improvement.
umns, the soil remoulding is visible for the latest column installed. Taylor and Francis, London, pp. 23–29.
When two columns are installed at common spacings, the installation Ellouze, S., Bouassida, M., Bensalem, Z., Znaidi, M.N., 2017. Numerical analysis of the
installation effects on the behaviour of soft clay improved by stone columns.
effects interact between each other. The plastic annuli around the col­ Geomech. Geoeng. 12, 73–85.
umns overlap and get larger than for a single column. Besides, the peak Elshazly, H.A., Elkasabgy, M., Elleboudy, A., 2008. Effect of inter-column spacing on soil
of horizontal stresses due to column installation is slightly higher than stresses due to vibro-installed stone columns: interesting findings. J. Geotech. Geol.
Eng. 26, 225–236.
that for the installation of a single column and is located a bit further Farias, M.M., Nakai, T., Shahin, H.M., Pedroso, D.M., Passos, P.G.O., Hinokio, M., 2005.
away from the column wall. This is more notable when the column radii Ground densification due to sand compaction piles. Soils Found. 45, 167–180.
are larger or the columns are closer spaced. Gäb, M., Schweiger, H.F., Thurner, R., Adam, D. 2007. Field trial to investigate the
performance of a floating stone column foundation. Proc. 14th Eur. Conf. Soil Mech.
For a group of nine stone columns, the radial stresses further increase
Geotech. Eng., Madrid, 24-27 September 2007. Millpress, Amsterdam, pp. 1311-
with the installation of more columns and the peak values are reached 1316.
outside the column group, but close to the outer columns of the group, at Guetif, Z., Bouassida, M., Debats, J.M., 2007. Improved soft clay characteristics due to
a distance around 2-4R. At the midpoint between two outer columns, stone column installation. Comput. Geotech. 34 (2), 104–111.
Han, J., 2015. Principles and practices of ground improvement. Wiley, Hoboken, New
excess pore pressures show similar trends as those measured in the field, Jersey, USA.
with peaks when the closest columns are being installed and the tip is at Indraratna, B., Basack, S., Rujikiatkamjorn, C., 2013. Numerical solution of stone
the studied depth, drops as the column penetrates further and gradual column-improved soft soil considering arching, clogging, and smear effects.
J. Geotech. Geoenviron. Eng. 139, 377–394.
accumulation of excess pore pressures after the installation of each Karlsson, M., Yannie, J., Dijkstra, J., 2019. Modeling aging of displacement piles in
column. Similar results have been found at this midpoint for the peak natural soft clay. J. Geotech. Geoenviron. Eng. 145, 04019070.
and final excess pore pressures with the two studied installation se­ Kirsch, F., 2004. Experimentelle und numerische Untersuchungen zum Tragverthalten
von Rüttelstopfsäulengruppen. Technische Universität Braunschweig. Dissertation.
quences, namely “Outside in” and “Inside out”. On the other hand, the Kirsch, K., Kirsch, F., 2010. Ground Improvement by Deep Vibratory Methods. Spon
“Outside in” sequence generates slightly larger radial stresses sur­ Press, London, UK.
rounding the group of columns (providing a better lateral confinement) Kirsch, F. 2006. Vibro stone column installation and its effect on ground improvement.
Proc. Num. Modelling Construction Processes Geotech. Eng. Urban Environment,
and the installation effects of the “Inside out” configuration are clearly Bochum, Germany, 23-24 March 2006. Taylor and Francis, London, pp. 115-124.
not symmetric. Le Kouby, A., Dupla, J.C., Canou, J., Francis, R., 2016. The effects of installation order on
the response of a pile group in silica sand. Soils Found. 56, 174–188.
Lee, F.H., Juneja, A., Tan, T.S., 2004. Stress and pore pressure changes due to sand
CRediT authorship contribution statement
compaction pile installation in soft clay. Géotechnique 54 (1), 1–16.
McCabe, B.A., Kamrat-Pietraszewska, D., Egan, D., 2013. Ground heave induced by
Atefe Geramian: Conceptualization, Methodology, Software, installing stone columns in clay soils. Proc. ICE – Geotech Eng. 166, 589–593.
Investigation, Writing – original draft, Visualization, Funding acquisi­ McCabe, B.A., Lehane, B.M., 2006. The behaviour of axially loaded pile groups driven in
clayey silt. J. Geotech. Geoenviron. Eng. ASCE 132 (3), 401–410.
tion. Jorge Castro: Conceptualization, Methodology, Resources, McCabe, B.A., Nimmons, G.J., Egan, D., 2009. A review of field performance of stone
Writing – original draft, Visualization, Supervision, Funding acquisition. columns in soft soils. Proc. ICE – Geotech Eng. 162, 323–334.
Mahmoud Ghazavi: Writing – review & editing, Supervision, Funding McMeeking, R.M., Rice, J.R., 1975. Finite-element formulations for problems of large
elastic-plastic deformation. Int. J. Solids Struct. 11, 601–616.
acquisition. Marina Miranda: Conceptualization, Methodology, Re­ Nagula, S.S., Grabe, J., 2018. First step to model the installation effect of geosynthetic
sources, Writing – review & editing, Supervision, Funding acquisition. encased sand columns on existing columns. 11th Int. Conf. Geosynth. (ICG 2018).
Nagula, S.S., Nguyen, D.M., Grabe, J., 2018. Numerical modelling and validation of
geosynthetic encased columns in soft soils with installation effect. Geotext.
Declaration of Competing Interest Geomembr. 46, 790–800.
Nguyen, H.H., Khabbaz, H., Fatahi, B., 2019. A numerical comparison of installation
The authors declare that they have no known competing financial sequences of plain concrete rigid inclusions. Comput. Geotech. 105, 1–26.
Pucker, T., Grabe, J., 2012. Numerical simulation of the installation process of full
interests or personal relationships that could have appeared to influence displacement piles. Comput. Geotech. 45, 93–106.
the work reported in this paper. Sexton, B.G., McCabe, B.A., 2015. Modeling stone column installation in an elasto-
viscoplastic soil. Int. J. Geotech. Eng. 9, 500–512.
Soleimani, M., Weissenfels, C., 2021. Numerical simulation of pile installations in a
Appendix A. Supplementary material
hypoplastic framework using an SPH based method. Comput. Geotech. 133, 104006.
Dassault Systémes. 2020. Abaqus User Manual 6.14.
Supplementary data to this article can be found online at https://doi. Tehrani, F.S., Nguyen, P., Brinkgreve, R.B.J., van Tol, A.F., 2016. Comparison of press-
org/10.1016/j.compgeo.2022.104931. replace method and material point method for analysis of jacked piles. Comput.
Geotech. 78, 38–53.
Wang, J., Li, Q., 2019. Coupled continuum-discrete modeling of rammed floating stone
References column installation. AIP Advances 9, 015222. doi: 10.1063/1.5083939.
Watts, K.S., Chown, R.C., Serridge, C.J., Crilly, M.S., 2001. Vibro stone columns in soft
Al Ammari, K.A., Clarke, B.G., 2018. Effect of vibro stone-column installation on the clay soil: a trial to study the influence of column installation on foundation
performance of reinforced soil. J. Geotech. Geoenviron. Eng. 144, 04018056. performance. Proc. 15th Int. Conf. Soil Mech. Found. Eng., Istanbul, pp. 1867–1870.
Amoroso, S., Rollins, K.M., Monaco, P., Thorp, A., 2015. Use of SDMT testing for Watts, K.S., Johnson, D., Wood, L.A., Saadi, A., 2000. An instrumented trial of vibro
measuring soil densification by ground improvement in Christchurch. New Zealand. ground treatment supporting strip foundations in a variable fill. Géotechnique 50
3rd Int. Conf. Flat Dilatometer DMT’15 177–184. (6), 699–708.
Baguelin, F., Jézéquel, J.-F., Shields, D.H., 1978. The pressuremeter and foundation Weber, T.M., Plötze, M., Laue, J., Peschke, G., Springman, S.M., 2010. Smear zone
engineering. Trans Tech Publications, Clausthal, Germany. identification and soil properties around stone columns constructed in-flight in
Barksdale, R.T., Bachus R.C., 1983. Design and construction of stone columns. Report centrifuge model tests. Géotechnique 60 (3), 197–206.
FHWA/RD-83/026. Springfield: Nat. Tech. Information Service.

10