Professional Documents
Culture Documents
net/publication/262686173
CITATIONS READS
22 914
3 authors:
Jorge Castro
Universidad de Cantabria
54 PUBLICATIONS 575 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Bryan A. McCabe on 16 January 2015.
RESEARCH PAPER
Jorge Castro
Received: 23 August 2012 / Accepted: 31 May 2013 / Published online: 26 September 2013
Springer-Verlag Berlin Heidelberg 2013
Abstract Numerous approaches exist for the prediction Keywords Analytical design methods Finite element
of the settlement improvement offered by the vibro- analyses Stone columns Settlement improvement factor
replacement technique in weak or marginal soil deposits.
The majority of the settlement prediction methods are
1 Introduction
based on the unit cell assumption, with a small number
based on plane strain or homogenisation techniques. In this
The potential of vibro-replacement stone columns to
paper, a comprehensive review and assessment of the more
reduce settlement [41], improve bearing capacity [5],
popular settlement prediction methods is carried out with a
accelerate consolidation [26], and reduce liquefaction
view to establishing which method(s) is/are in best agree-
potential [35] in soft soils is now widely appreciated in
ment with finite element predictions from a series of
geotechnical engineering practice. The vibro-replacement
PLAXIS 2D axisymmetric analyses on an end-bearing
technique and associated equipment have been described in
column. The Hardening Soil Model in PLAXIS 2D has
detail by Slocombe et al. [37] and Sondermann and Wehr
been used to model the behaviour of both the granular
[38]. Settlement performance tends to be the governing
column material and the treated soft clay soil. This study
design criterion in these soils, and most analytical design
has shown that purely elastic settlement prediction methods
methods provide a direct prediction of a settlement
overestimate the settlement improvement for large modular
improvement factor, n, defined as the settlement of
ratios, while the methods based on elastic–plastic theory
untreated ground (s0) divided by the settlement of the
are in better agreement with finite element predictions at
ground treated with granular columns (st), see Eq. 1.
higher modular ratios. In addition, a parameter sensitivity
study has been carried out to establish the influence of a s0
n¼ ð1Þ
range of different design parameters on predictions st
obtained using a selection of elastic–plastic methods. This settlement improvement factor can then be used to
predict the settlement of treated ground (st = s0/n). The
value of s0 (for a scenario in which the loading extends over a
wide area) is usually calculated from elastic theory (Eq. 2),
where pa is the applied pressure, H is the thickness of the
B. G. Sexton B. A. McCabe (&)
College of Engineering and Informatics, National University treated soil layer, and Eoed is the oedometric soil modulus.
of Ireland, Galway, Ireland pa H
e-mail: bryan.mccabe@nuigalway.ie s0 ¼ ð2Þ
Eoed
B. G. Sexton
e-mail: b.sexton1@nuigalway.ie Analytical design methods typically relate n to the area-
replacement ratio, Ac/A (where A is the cross-sectional area
J. Castro
of a unit cell treated with a single stone column of cross-
Department of Ground Engineering and Materials Science,
University of Cantabria, Santander, Spain sectional area, Ac, see Fig. 1). The area-replacement ratio is
e-mail: castrogj@unican.es a measure of the amount of in situ soil replaced with stone
123
994 Acta Geotechnica (2014) 9:993–1011
and is dependent on the column spacing, s and column 16, 29–32, 40]), with yielding of the column material
diameter, Dc (Eq. 3), where k is a constant depending on considered in the latter case. The elastic–plastic methods
the column arrangement. are typically based on the Mohr–Coulomb failure criterion,
2 with some assuming that the granular material deforms at
A s
¼k ð3Þ constant volume as it yields (dilatancy angle, w = 0),
Ac Dc
while others have accounted for dilation of the granular
A number of other influential variables have been con- column material at yield using a constant dilatancy angle.
sidered in analytical formulations, and these include the A selection of settlement design methods and their inherent
effect of installation, load level, modular ratio and the fric- assumptions have been summarised in Table 1.
tion and dilatancy angles of the column material. The pub- Balaam and Booker [4] and Pulko and Majes [31] have
lished solutions account for these variables in different ways, highlighted that elastic–plastic methods are preferable to
although few capture all of them. The aim of this study is to purely elastic methods because the elastic methods tend to
provide a systematic review of these methods before using a overpredict the settlement improvement offered by column
2D/axisymmetric finite element (FE) parametric study to installation, especially for high modular ratios (Ec/Es,
appraise the ability of these methods to cater for the variables where Ec is the modulus of the column and Es is the
identified above. The axisymmetric analyses are carried out modulus of the soil). This over-prediction is as a result of
using the PLAXIS 2D (Brinkgreve et al. [9]) Hardening Soil the fact that elastic methods overpredict the stress con-
(HS) Model to model an infinite grid of columns installed in centration factor (SCF = rc/rs, where rc is the stress in the
a soft soil profile. It should be noted that the majority of the column and rs is the stress in the soil).
analytical methods discussed here only consider potential Approaches to modelling the behaviour of the column–soil
improvement to primary (consolidation) settlements. Sexton system vary; some, such as Han and Ye [19] have accounted
and McCabe [36] have used finite element analyses to con- only for vertical deformation, while others have accounted for
sider the improvement to creep settlements, which are highly both radial and vertical deformation. For elastic methods that
relevant in organic soils. consider vertical deformation only, the SCF is equal to the
ratio of the oedometric moduli of the column and soil mate-
rials. Elastic solutions that consider both radial and vertical
2 Vibro-replacement settlement prediction methods deformation result in slightly lower SCFs (lateral deformation
reduces SCFs, e.g., Balaam and Booker [3]). However, these
2.1 Theoretical considerations in vibro-replacement SCFs will still be overpredicted because yielding of the col-
design umn material is not considered (column yielding and plastic
strains will reduce SCFs). Barksdale and Bachus [5] have
Analytical settlement design approaches tend to be either suggested that SCFs in practice range from 3 to 10 depending
elastic (e.g. [1, 3, 7, 19]) or elastic–plastic (e.g. [4, 8, 10, on the column spacing adopted in the field.
123
Table 1 Settlement design methods and assumptions
Settlement Elastic Unit cell (UC)/ Drained (D)/ Equal Dilatancy End- Shear Radial Installation? Incom- Immediate Mohr– Iterative Notes
prediction (E)/ Plane strain undrained vertical of granular bearing stresses at deformation pressible settlement Coulomb (I)/
method elastic– (PS)/ ? consolidation strain? material columns? column– considered? column? given by (MC) closed-
plastic homogenisation (U ? C) considered? soil method? failure form
(EP) (H) interface criterion? (CF)?
(sint)?
Bauer [7]
Han and Ye [19] E UC U?C 4 – 4 sint = 0 No – 7 – CF
Balaam and EP UC D 4 Constant w 4 sint = 0 Yes Input K 4 4 I* Immediate
Booker [4] settlement
negligible
compared to the
total final
settlement
*Iterative approach
requiring
numerical
implementation to
obtain a solution
Pulko and Majes EP UC D 4 Constant w 4 sint = 0 Yes Input K 7 4 CF
[31]
Pulko et al. [32] EP UC D 4 Constant w 4 sint = 0 Yes Input K 7 4 CF
Priebe [29] EP UC D 4 w = 0 4 sint = 0 Yes K=1 4 7 4 CF
Priebe [30] EP UC D 4 w = 0 4 sint = 0 Yes Input K 7 4 CF
Castro and EP UC U?C 4 Constant w 4 sint = 0 Yes Input K 4 4 CF
Sagaseta [10]
Goughnour and EP UC U?C 4 No 4 sint = 0 Yes K0 \ K \ 1/ 4 7 4 CF ‘Incremental
Bayuk [16, K0 method’
17]
Borges et al. [8] EP UC U?C* 4 No (FE 4 Perfect Yes K = 0.7 7 4 CF Finite element (FE)
Basis) bonding basis
from *Numerical analysis
soil to is based on an
column U ? C approach
but design
equation is
applicable for
drained
conditions also
Van Impe and EP PS D/U ? C 4 w = 0 4 sint = 0 Yes – 7 4 I
De Beer [39]
995
123
996 Acta Geotechnica (2014) 9:993–1011
volumetric strain
Empirical curves
due to dilation
strength of the
ssoil is the shear
lation and subsequent bulging has been accounted for in
different ways. Priebe [29] has assumed an increase in the
ev,d is the
coefficient of lateral earth pressure following column
Notes
soil
installation to the liquid earth pressure of the soil (K = 1).
Other methods allow for the input of different values
Iterative
closed-
(CF)?
–
I
failure
Yes
sint = 0
sint = 0
soil
4
7
eta [10] solution has been derived for the case of an elas-
Constant w
of granular
Dilatancy
Yes (ev,d)
material
D/U ? C
D/U ? C
(U ? C)
UC
(H)
plastic
(EP)
EP
EP
Schweiger and
Van Impe and
Pande [34]
Greenwood
empirical
[18]—
method
123
Acta Geotechnica (2014) 9:993–1011 997
developed in the interim, Greenwood’s [18] curves have can also be used to model the behaviour of floating col-
been proposed for end-bearing columns neglecting imme- umns (the plane strain and unit cell approaches are gen-
diate settlements and shear displacements (as noted in erally based on end-bearing stone columns). However, they
Greenwood’s original proposal). can be used to model floating columns in conjunction with
At present, the majority of the design methods have FE analyses (FE solutions generally assume that there is no
been derived for a unit cell representing an infinite grid of slip at the column–soil interface).
regularly spaced end-bearing columns, e.g. [1, 3, 4, 7, 8,
10, 16, 19, 29–32, 40]. Other solutions have been devel- 2.4 Unit cell approaches
oped based on plane strain (e.g. Van Impe and De Beer
[39]) or homogenisation techniques (e.g. Schweiger and The unit cell approach is based on the assumption of a large
Pande [34], Lee and Pande [22]). For all three approaches, grid of regularly-spaced columns subjected to a uniform
simplifying assumptions are usually considered, e.g. the load. Therefore, all of the columns will exhibit similar
column and the surrounding soil undergo equal vertical behaviour and an analysis of one such column, and its
settlement (referred to as the ‘equal vertical strain’ tributary soil area is sufficient. Owing to the symmetry of
assumption) and the shear stresses at the column–soil the problem, the shear stresses along the perimeter of the
interface are assumed to be zero. unit cell are assumed to be zero. The unit cell approach is
valid except for columns near the edges of the loaded area
[3, 25], which are assumed to be in the minority for large
2.3 Plane strain/homogenisation techniques
groups.
A flowchart detailing the origin and development of the
The plane strain approach involves replacing the stone
majority of the design methods based on the unit cell
columns with stone walls (trenches) having an ‘equivalent’
approach is presented in Fig. 2. The simplest analytical
overall plan area. The homogenisation technique involves
approach to stone column design is known as the ‘equi-
modelling the stone column and treated soil as a composite
librium method’. The approach is based on elastic theory
material with improved soil properties and is formulated
and has been described by Aboshi et al. [1]. It is based on
assuming that the influence of the columns is uniformly
vertical equilibrium between the soil and the columns with
and homogeneously distributed throughout the treated soil,
oedometric (i.e. elastic behaviour with full lateral con-
e.g. [34].
finement) conditions in the soil. From vertical equilibrium
The homogenisation technique can be used in conjunc-
(Eq. 4):
tion with flexible and rigid rafts (‘equal vertical stress’ and
‘equal vertical strain’ assumptions, respectively), which pa :A ¼ rc :Ac þ rs :ðA Ac Þ ð4Þ
makes it possible to isolate different behavioural aspects The settlement (assuming oedometric conditions) is then
associated with columns near the edge of a loaded area. It calculated as shown in Eq. 5.
123
998 Acta Geotechnica (2014) 9:993–1011
123
Acta Geotechnica (2014) 9:993–1011 999
(constant dilatancy angle, w) at yield, whereas Priebe’s problem is studied under drained or undrained (plus con-
[29, 30] method assumes the granular column material to solidation) conditions. For the case of an elastic column,
deform at constant volume (w = 0). Pulko and Majes [31] both approaches produce the same result. For a yielding
and Castro and Sagaseta [10] are elastic–plastic extensions column (elastic–plastic case), although the stress paths are
of the earlier elastic solution developed by Balaam and different, the final settlements are very similar (provided
Booker [3] for drained conditions. Castro and Sagaseta [10] that the drained solutions account for elastic strains of the
have considered an undrained loading situation followed column during its plastic deformation), as shown by Castro
by a consolidation process to allow for the dissipation of and Sagaseta [11] using finite element calculations. For
excess pore pressures, whereas Pulko and Majes [31] and drained analyses that neglect the elastic strains of the
Pulko et al. [32] have studied the unit cell problem under column during its plastic deformation (e.g. Pulko and
drained conditions. As noted by Castro and Sagaseta [11], Majes [31]), the final settlement will be underpredicted.
both approaches are considered to be limiting cases of the For undrained plus consolidation solutions (e.g. Castro and
real situation because load application is not rapid enough Sagaseta [10]), neglecting the elastic strains of the column
to be considered as undrained nor slow enough to be during its plastic deformation leads to negligible error in
considered as a drained process. the solution. The newer drained solution by Pulko et al.
The method developed by Pulko et al. [32], which deals [32] accounts for the elastic strains of the column during its
with encased stone columns, is an extension of the previous plastic deformation. Under such conditions, the differences
solution derived by Pulko and Majes [31]. The new method between the drained and undrained plus consolidation
by Pulko et al. [32] can also be applied to non-encased analyses will effectively vanish (i.e. Castro and Sagaseta
stone columns by setting the encasement stiffness to zero. [10] and Pulko et al. [32] will produce almost identical
The solutions derived by Castro and Sagaseta [10] and solutions for non-encased columns, as studied here).
Pulko and Majes [31] ignored the elastic strains in the The design methods derived by Castro and Sagaseta [10]
column during its plastic deformation, whereas the newer and Pulko et al. [32] have dealt with column yielding in
solution by Pulko et al. [32] has taken them into account. different ways. Castro and Sagaseta’s [10] undrained plus
Figure 3 (from Castro and Sagaseta, [11]) shows the consolidation formulation uses a factor Uey (elastic degree
different stress paths followed depending on whether the of consolidation at the moment of column yielding) to
Vertical stress, σ zc
Vertical stress, σ zc
U U U
I I I
0 0
Radial stress, σ rc Radial stress, σ rc Radial stress, σ rc
I: initial
Drained analysis U: undrained loading
Consolidation analysis Y: yield
F: final, drained
Fig. 3 Stress paths in the column; a elastic case b at yielding c elastic–plastic case (Castro and Sagaseta [11])
123
1000 Acta Geotechnica (2014) 9:993–1011
work out whether or not the column is in a plastic state (if capabilities of several of the aforementioned analytical
Uey [ 1, no yielding takes place, otherwise yielding of the methods. A unit cell approach (Fig. 4) with a column radius,
granular material occurs). Pulko et al. [32] have worked out Rc = 0.3 m (typical for columns at soft soil sites, e.g., Watts
a final yield depth, zy (i.e. yielding starts at the surface and et al. [41]), and a column length = 5 m has been adopted to
progresses downward as the applied load increases), to represent the behaviour of a single end-bearing column
which plastic strains appear in the column. within an infinite grid. Similar modelling approaches have
Borges et al. [8] have proposed a design method (based on been adopted by Debats et al. [12] and Ambily and Gandhi
a numerical rather than an analytical approach) relating the [2]. Horizontal deformation has been restricted at the sides
settlement improvement factor (n) to the area-replacement (roller boundaries), and both vertical and horizontal defor-
ratio, Ac/A, and to the ratio of the deformability of the soft mations have been restricted at the base. The water table is
soil to the deformability of the column material (alterna- located at the surface. The columns are fully penetrating and
tively referred to as the modular ratio, Ec/Es). Their resulting have been wished in place (as is common practice), e.g. Gäb
design equation (and chart) is based on curve-fitting to the et al. [15] and Killeen and McCabe [21]. For the initial study,
results of a series of axisymmetric FE analyses of a unit cell the coefficient of lateral earth pressure, K, is assumed to be
with a program incorporating Biot consolidation theory with unaffected by column installation (K0 = 1 - sin u0 = 0.44).
the p–q–h model (extension of the Modified Cam-Clay A parameter sensitivity study considering different K values
(MCC) Model, based on the Drucker-Prager failure crite- has been described in Sect. 4.2.4, e.g., Priebe [29], Gough-
rion). In contrast to the MCC Model, the parameter nour and Bayuk [17] and Gäb et al. [15] have accounted for
M (defining the slope of the critical state line) is not constant, the densification as a result of column installation by using
e.g., Lewis and Schrefler [24], Domingues et al. [14]. an increased coefficient of lateral earth pressure, K = 1 (for
The authors have adopted a value of K = 0.7 for the the soil).
coefficient of lateral earth pressure at rest following col- The behaviour of the composite model has been studied
umn installation (in between K = 1 - sin u0 and K = 1). under a 100 kPa load (the sensitivity study described in
The settlement improvement factor (Eq. 8) has been Sect. 4.2.1 has also examined the behaviour of the system
derived based on statistical analysis techniques and has under 50 and 75 kPa loads) applied through a plate element
been related to the two factors that the authors found had (normal stiffness, EA = 5 9 106 kN/m, flexural rigidity,
the most significant influence on the results. A design chart EI = 8.5 9 103 kNm2/m, Poisson’s ratio, m = 0). The
has been developed based on this design equation, which is plate element is intended to represent a rigid loading
applicable for 10 B Ec/Es B 100 and 3 B A/Ac B 10, with platform to prevent differential settlements. Different series
calculated improvement factors greater than 1.5. of analyses have been carried out for different modular
ð0:0038EEcs 0:3423Þ ratios, Ec/Es, of 5, 10, 20, and 40 (note that good com-
Ec
n ¼ 0:125 þ 0:7742 ðA=Ac Þ ð8Þ parison with elastic methods necessitates the use of unre-
Es alistically low Ec/Es ratios). These values of Ec/Es are in
the same range as those adopted by Balaam and Booker
3 Axisymmetric modelling (PLAXIS 2D) [3], Castro and Sagaseta [10], and Poorooshasb and Mey-
erhof [28]. In all cases, the properties of the column
Axisymmetric FE analyses using PLAXIS 2D (Brinkgreve material have been fixed, while the soil properties have
et al. [9]) have been carried out as a means of appraising the been varied to generate the necessary Ec/Es ratios. The
diameter of the unit cell has been altered to study the effect
of different area-replacement ratios, e.g., Domingues et al.
[14]. The column diameter has been fixed at 0.6 m (argu-
ably the column diameter in the field will be a function of
Ec/Es, but a fixed diameter has been considered here for
numerical purposes).
Load settlement behaviour (primary settlement) has
been analysed using the HS Model to model both the clay
and the stone. Both have been modelled as fully drained
materials. Similar results would be achieved modelling the
clay as an undrained material with a follow-up consolida-
tion period (analyses have been carried out in verification,
e.g., Fig. 5).
Fig. 4 Axisymmetric unit cell model (100 kPa load)
The HS Model is a hyperbolic elasto-plastic model that
accounts for increasing soil layer stiffness with stress-level
123
Acta Geotechnica (2014) 9:993–1011 1001
Settlement (m)
-0.10 -0.10
Max δ ≈ 0.235m Max δ ≈ 0.227m
-0.15 -0.15
Drained Undrained +
-0.20 -0.20 Consolidation
-0.25 -0.25
Settlement (m)
-0.10 -0.10
-0.15 -0.15 Max δ ≈ 0.270m
Max δ ≈ 0.277m
-0.20 -0.20
Drained Undrained +
-0.25 -0.25 Consolidation
-0.30 -0.30
Table 2 FE model parameters approximately equal to Eref 50 (secant modulus), and Eur
ref
ref
Clay (drained) Stone backfill (drained) (unload–reload modulus) was taken as 3E50 , as recom-
mended by Brinkgreve et al. [9]. The values of Eref ref
oed, E50 ,
c (kN/m3) 16.5 19.0 ref
and Eur for the stone quoted in Table 2 are based on Gäb
kx (m/day) 1 9 10-4 1.7 et al. [15]. The properties have been altered using Eq. 9 to
ky (m/day) 6.9 9 10-5 1.7 correspond to a confining pressure of 50 kPa (closer to the
einit 2.0 0.5 confining pressure in the subsequent numerical simula-
/0 () 34 45 tions). Gäb et al. [15] have defined the stiffness moduli at a
w () 0 15 reference pressure, pref, of 100 kPa. The stress dependency
Knc
0 0.441 0.296 of soil stiffness is dictated by the power, m (m = 1 is
C0 (kPa) 1.0 1.0 typical for soft soils [9]). For the granular column material,
Eref
oed (kPa) 2,840 56,800 a value of m = 0.3 has been used [15].
Eref
50 (kPa) 3,550 56,800 m
ref p
Eref
ur (kPa) 17,900 170,400 E¼E ð9Þ
pref
m (power) 1.0 0.3
A complete list of the parameters used in the FE model
pref (kPa) 50 50
for the case when Ec/Es = 20 is given in Table 2. The
mur 0.2 0.2
Ec/Es ratio has been defined as the ratio of the constrained/
K0 0.441 –
oedometric moduli at a reference pressure of 50 kPa, i.e., at
OCR 1.0 –
pref = 50 kPa, Eoed,c/Eoed,s = 56,800/2,840 = 20. The soil
properties represent a simplified single layer profile loosely
(no viscous effects). Its formulation has been described in based on parameters for the Bothkennar soft clay test site
detail by Schanz et al. [33]. A friction angle (u0 ) of 45 has (e.g. Leroueil et al. [23], Nash et al. [27]) proposed by
been selected for the stone, representative of bottom feed Killeen and McCabe [21]. The stiff crust has been excluded
columns, while the dilatancy angle (w) was calculated as from the soil profile. The values of Eref ref ref
oed, E50 , and Eur for
w = u0 - 30. Erefoed (oedometric modulus) was assumed the soil have been doubled and quadrupled for modular
123
1002 Acta Geotechnica (2014) 9:993–1011
ratios of Ec/Es = 10 and 5, respectively, while they have (a) 1.4 Balaam & Booker (1981)
1.3
been halved for Ec/Es = 40 (with all remaining soil prop- 1.2
Priebe's n0 (1976)
Priebe's n1 (1995)
erties remaining fixed), e.g., for a modular ratio of 40, 1.1 Priebe's n2 (1995)
n/nFE
Eoed,c/Eoed,s = 56,800/1,420 = 40 at pref = 50 kPa. 1.0
0.9
Pulko & Majes (2005)
Pulko et al. (2011)
It should be noted that the Ec/Es values quoted here are 0.8 Castro & Sagaseta (2009)
Borges et al. (2009)
0.7
just approximate indicators of the values that are actually Aboshi et al. (1979)
0.6
modelled in the numerical simulations (such values can 3 4 5 6 7 8 9 10
A/Ac
only be quoted as exact for a linear elastic soil model). In
this case (for the HS Model), the soil stiffness depends on (b) 1.4 Balaam & Booker (1981)
1.3
Priebe's n0 (1976)
stress-level and over-consolidation ratio, so the values of 1.2 Priebe's n1 (1995)
1.1 Priebe's n2 (1995)
n/nFE
Ec/Es will only be exact for a normally consolidated soil for 1.0 Pulko & Majes (2005)
which the reference pressures in the soil and column 0.9 Pulko et al. (2011)
0.8 Castro & Sagaseta (2009)
materials are identical (in this case, at pref = 50 kPa). 0.7
Borges et al. (2009)
Aboshi et al. (1979)
Nash et al. [27], among others, have carried out exten- 0.6
3 4 5 6 7 8 9 10
sive site characterisation at the Bothkennar site for which A/Ac
an over-consolidation ratio of between 1.5 and 1.6 has been (c) 1.4 Balaam & Booker (1981)
1.3
Priebe's n0 (1976)
reported for the lower Carse clay. However, since the 1.2 Priebe's n1 (1995)
analytical formulations are unable to consider an over- 1.1 Priebe's n2 (1995)
n/nFE
1.0 Pulko & Majes (2005)
consolidation effect, it was deemed more appropriate to use 0.9 Pulko et al. (2011)
OCR = 1.0 for defining the initial stress state for the 0.8 Castro & Sagaseta (2009)
0.7 Borges et al. (2009)
subsequent numerical analyses. It is acknowledged that all 0.6
Aboshi et al. (1979)
little bearing on calculated settlement improvement factors 1.0 Pulko & Majes (2005)
Pulko et al. (2011)
in this case, which are virtually the same for OCR = 1.0 0.9
Castro & Sagaseta (2009)
0.8
and OCR = 1.5. 0.7 Borges et al. (2009)
Aboshi et al. (1979)
0.6
3 4 5 6 7 8 9 10
A/Ac
4 Results
Fig. 7 n/nFE versus A/Ac (pa = 100 kPa); a Ec/Es = 5 b Ec/Es = 10
c Ec/Es = 20 d Ec/Es = 40
4.1 Design method predictions versus FE results (base
case) modular ratio is increasing, i.e., the influence of the mod-
ular ratio becomes negligible (again this is to be expec-
Settlement improvement factors for a ‘base case’ ted—only elastic design methods will show dependence on
0
(pa = 100 kPa, uc = 45, wc = 15o, K0 = 0.44) are plot- the modular ratio once the column has yielded and this is
ted in Fig. 6 for the four different modular ratios consid- why elastic methods overpredict n values for high modular
ered in this study. The results in Fig. 6 indicate that ratios). Parameters with a more dominant influence on the
improvement factors predicted using the FE method settlement behaviour include the friction angle of the col-
0
increase as the modular ratio increases, which is to be umn material, /c, and the coefficient of lateral earth
expected. The FE n values appear to be converging as the pressure, K.
4.0 8
Ec/Es = 5 7 Ec/Es = 5
3.5
6
3.0
Ec/Es = 10 5 Ec/Es = 10
SCF
nFE
2.5 4
Ec/Es = 20 3 Ec/Es = 20
2.0
2
1.5
Ec/Es = 40
Ec/Es = 40
1
1.0 0
3 4 5 6 7 8 9 10 3 4 5 6 7 8 9 10
A/Ac A/Ac
Fig. 6 nFE versus A/Ac (influence of Ec/Es) Fig. 8 PLAXIS-calculated SCFs versus A/Ac (base case)
123
Acta Geotechnica (2014) 9:993–1011 1003
Settlement improvement factors calculated using design • It appears that the newest methods (i.e. Castro and
methods based on the unit cell assumption are compared to Sagaseta [10], Pulko et al. [32]) offer the best agreement
the numerical results in Fig. 7a–d for the ‘base case’. The with the FE data (0.95 \ n/nFE \ 1.1) over the entire
data are presented as a ratio n/nFE (rather than n directly) range of modular ratios considered and are in almost
against A/Ac, e.g., n/nFE [ 1 indicates that the design perfect agreement with each other, despite the fact that the
method ‘overpredicts’ the settlement improvement factor former is based on an undrained loading situation with
(compared to the FE analyses), etc. Some of the analytical subsequent consolidation, while the latter is based on
predictions produce n/nFE values beyond the upper limit of drained conditions. However, as highlighted in Sect. 2.4,
1.4 depicted on Fig. 7 and hence not every solution is these methods (despite the different stress paths) are
represented on every plot. The predictions using Aboshi expected to give more or less identical results (the drained
et al. [1] have been obtained by deducing the SCFs at the solution which considers the elastic strains in the column
surface from the numerical output. While this is a non- during its plastic deformation will produce the same
standard approach, it is helpful in gauging the variation of results as the undrained plus consolidation solution).
n/nFE against A/Ac predicted by this method. These pre- • It is also worth noting that Balaam and Booker [4] (not
dictions are just used to establish whether the simple included in Fig. 7) will produce similar results. How-
equilibrium method can in fact be used to obtain reliable ever, this method requires both numerical implemen-
n values if sufficiently accurate input SCFs can be established. tation and an iterative solution technique and has not
been included in the graphs.
• In general, it appears that the agreement between the
4.1.1 Equilibrium approach
FE predictions (HS Model) and the elastic–plastic
analytical predictions improves with increasing modu-
The simple equilibrium method described by Aboshi et al.
lar ratio (1.0 \ n/nFE \ 1.3 in Fig. 7c, d).
[1], based on FE-calculated surface SCFs (see Fig. 8),
• For the analytical methods, the reason for the better
consistently predicts n/nFE & 0.9 irrespective of the
predictions at higher modular ratios is likely to be due to
modular ratio or area-replacement ratio. This indicates that
the variability of soil stiffness with stress-level. The
the method, despite its simple nature, could be safely
analytical formulations assume a constant stiffness mod-
applied in real-life design situations provided that the SCF
ulus for the soil and column. However, the HS Model in
is not overestimated.It is interesting to note that if the
PLAXIS accounts for the stress dependency of stiffness
average SCF over the complete soil profile was used
(i.e. the stiffness depends on the confining pressure, e.g.,
instead of the SCF at the surface, n/nFE would be mar-
Eq. 9). As a result of this, the modular ratio used in the
ginally lower for each modular ratio (n/nFE & 0.8).
analytical solutions will not be exactly the same as that in
4.1.2 Analytical approaches the FE calculations. For low modular ratios, the column
will not take as much of the load as it would take for higher
An appraisal of the analytical approaches can be summa- modular ratios, i.e., for a lower modular ratio, the
rised as follows: confining pressure in the column will be lower. Accord-
ingly, the confining pressure in the soil will be higher at
• Elastic methods, e.g., Balaam and Booker [3] over- lower modular ratios than at higher modular ratios. In
predict the settlement improvement for large modular general, the differences between the analytical and FE
ratios, i.e., n/nFE 1.4 for modular ratios of 20 and 40 predictions will be more evident in situations where
(Fig. 7c, d, respectively). For elastic methods, the SCF elastic strains are more prominent (e.g. low A/Ac values).
will be too high because yielding of the column
material is ignored (yielding/plastic strains reduces the 4.1.3 Semi-empirical approaches
SCF and hence the predicted settlement improvement).
• The Pulko and Majes [31] solution appears to predict An examination of the Priebe [29, 30] predictions in Fig. 7
n/nFE values consistently in the range 1.1–1.4 for modular yields the following:
ratios of 10, 20, and 40. This clearly shows how neglecting
the elastic strains in the column during its plastic • Priebe’s n0 [29] is independent of the modular ratio,
deformation for a drained solution influences the results Ec/Es (n0 predictions are closer to the FE results as the
(i.e. overpredicts settlement improvement factors because modular ratio increases because the FE n values rise
neglecting the elastic strains means lower ‘treated’ and thus n/nFE approaches 1).
settlements are predicted). As is clear from Fig. 7, the • Priebe’s n1 [30] predicts less of an improvement than n0
deviation from n/nFE = 1 is larger at low A/Ac values, i.e., in all cases, i.e., accounting for the compressibility of
in cases where the elastic strains are more important. the column material leads to lower n values. For lower
123
1004 Acta Geotechnica (2014) 9:993–1011
A/Ac values (i.e. more stone), there is more compress- 4.2 Parameter sensitivity study
ible column material to be accounted for, and hence n1
deviates further from n0 as the area-replacement ratio The comparisons carried out in the previous section clearly
increases (lower A/Ac values). indicate that the methods derived by Castro and Sagaseta
• The ratio n2/nFE (representing more lateral support) is [10] and Pulko et al. [32] offer the best agreement with
marginally greater than n1/nFE in all four graphs. The finite element predictions for the ‘base case’ considered.
difference between n2/nFE and n1/nFE would be more Based on this, a parameter sensitivity study is carried out to
pronounced for a higher at-rest coefficient of lateral establish the effect of altering selected key parameters (pa,
0
earth pressure, K, e.g., for K = 1, n2/nFE would be uc, wc, K0). In addition, the influence of these parameters
above n0/nFE in some cases. on Priebe’s n2 [30] has also been examined because of its
• In the case of Priebe [29, 30], the reason for the better popularity in European geotechnical practice.
agreement with the FE predictions at higher modular
ratios is due to the semi-empirical nature of the method. 4.2.1 Load level (pa)
The predictions appear to be better for the more
‘realistic’ higher modular ratios. This could be due to The behaviour of the composite soil–column system has
the assumption of a significant bulging mechanism also been studied under 50 and 75 kPa stresses (with all
which is more prevalent in softer soils, i.e., higher other parameters fixed). As before, design method predic-
modular ratios (e.g. CCE theory has been used by tions have been compared to FE results (Figs. 9, 10). The
Hughes and Withers [20] to model the lateral bulging elastic–plastic design methods predict larger improvement
failure of a single column and hence predict its ultimate
bearing capacity, while Priebe [29] has also used CCE (a) 1.4 Balaam & Booker (1981)
1.3
theory as the basis for the aforementioned design 1.2
Priebe's n0 (1976)
Priebe's n1 (1995)
method). 1.1 Priebe's n2 (1995)
n/nFE
much less of an improvement than the other design meth- 1.0 Pulko & Majes (2005)
0.9 Pulko et al. (2011)
ods for modular ratios of 5, 10 and 20 (n/nFE \ 0.8, pre- 0.8 Castro & Sagaseta (2009)
dictions are out of the range of plotted n/nFE values in 0.7
Borges et al. (2009)
Aboshi et al. (1979)
Fig. 7a, b), i.e., n values \1.5 (which do not appear on the 0.6
3 4 5 6 7 8 9 10
design chart). For Ec/Es = 40, Borges et al. [8] show better A/Ac
agreement with the other design methods. (c) 1.4 Balaam & Booker (1981)
1.3
Priebe's n0 (1976)
For modular ratios larger than Ec/Es = 40, the method 1.2 Priebe's n1 (1995)
proposed by Borges et al. [8] predicts even larger improve- 1.1 Priebe's n2 (1995)
n/nFE
123
Acta Geotechnica (2014) 9:993–1011 1005
(a) 1.4 Balaam & Booker (1981) reason for the discrepancy at Ec/Es = 5 is attributable to
1.3
1.2
Priebe's n0 (1976)
Priebe's n1 (1995)
the fact that the drained solution neglects the elastic strains
1.1 Priebe's n2 (1995) of the column during its plastic deformation. For low
n/nFE
1.0 Pulko & Majes (2005) tion by Pulko and Majes [31], as clarified in Sect. 2.4.
0.9 Pulko et al. (2011)
Castro & Sagaseta (2009)
Load level affects the depth to which plastic strains
0.8
0.7
Borges et al. (2009) appear in the column (yielding depends on the dimen-
0.6
3 4 5 6 7 8 9 10
Aboshi et al. (1979)
sionless load factor pa/(c0 .z) where c0 is the effective unit
A/Ac weight of the soil and z is the depth below ground level),
(c) 1.4
Balaam & Booker (1981)
i.e., yielding starts at the surface and progresses down-
1.3
1.2
Priebe's n0 (1976) wards with time (Castro and Sagaseta [10]); higher loads
Priebe's n1 (1995)
1.1 Priebe's n2 (1995) result in more and more column yielding. Yielding has
n/nFE
factors when columns are subjected to lower applied loads Priebe [29, 30], Pulko and Majes [31], Castro and Sagaseta
(as do the FE simulations), indicating that stone columns [10], and Pulko et al. [32] predict larger n values for higher
0
are more effective at lower load levels (less yielding). column friction angles, uc (with the exception of Pulko and
Elastic design methods have no dependency on load level Majes [31] at Ec/Es = 5, again illustrating that the method
(e.g. Balaam and Booker [3]), nor does Priebe’s n0 [29] or may not be applicable for Ec/Es B 5). The method by
the FE-based method derived by Borges et al. [8] which Borges et al. [8] is independent of the friction angle of the
depends only on Ac/A and Ec/Es. The SCFs used to obtain column material, while the elastic methods are over-sim-
n values for Aboshi et al. [1] have again been obtained plified in this respect. nFE is plotted against A/Ac in
from the FE output. Fig. 11a–d to show that the FE n values are also higher for
As was the case with pa = 100 kPa, it is worth noting higher friction angles. The influence of the friction angle
0
that the elastic–plastic method n values converge with (uc = 35, 40, 45) of the granular material is clearly
increasing modular ratio for both pa = 50 kPa (e.g. Fig. 9c, evident on the n/nFE values predicted by the favoured
d) and 75 kPa (e.g. Fig. 10c, d), i.e., 1.0 \ n/nFE \ 1.3 analytical settlement design methods in Fig. 12a–d for the
(despite some divergence for large quantities of stone, four different modular ratios considered (the other param-
e.g., A/Ac \ 4). For Ec/Es = 5, Pulko and Majes [31] pre- eters have been fixed at pa = 100 kPa, wc = 15 and
dicts lower n values at lower applied loads (this is in K0 = 0.44).
contrast with other methods, e.g., Priebe [30], Castro and • Priebe’s n2 [30] appears to consistently overpredict
Sagaseta [10], Pulko et al. [32]) and perhaps indicates that n values (i.e. n/nFE [ 1) for all friction angles consid-
the method may not be applicable for Ec/Es B 5. The ered in this study. It is interesting to note that the
123
1006 Acta Geotechnica (2014) 9:993–1011
3.0
nFE
2.0
1.0
3 4 5 6 7 8 9 10
A/Ac
(b)
(b) 4.0
3.5 HS Model (ϕ' = 35)
3.0
nFE
2.0
1.0
3 4 5 6 7 8 9 10
A/Ac
(c)
(c) 4.0
3.0
nFE
2.0
1.0
3 4 5 6 7 8 9 10 (d)
A/Ac
(d) 4.0
3.5 HS Model (ϕ' = 35)
3.0
nFE
2.0
0
Fig. 11 nFE versus A/Ac (influence of uc); a Ec/Es = 5 b Ec/Es = 10 • Their predictions appear to be in better agreement with
c Ec/Es = 20 d Ec/Es = 40 Priebe’s n2 [30] as the modular ratio increases (i.e.
softer soils with more associated bulging).
Examination of predicted SCFs (Fig. 13) illustrates part
influence of the friction angle is much more dominant
of the reason for the considerably different n value pre-
on Priebe’s n2 [30] values than it is on the n values
dictions for the design methods.
predicted by Castro and Sagaseta [10] and Pulko et al.
[32]. • Predicted SCFs are in excellent agreement for Castro
• The agreement between Priebe’s n2 [30] and the other and Sagaseta [10] and Pulko et al. [32] with ever so
analytical predictions is better for lower friction angles slight differences apparent for closely spaced columns
0
(e.g. uc = 35, e.g. n/nFE & 1.1) than it is for higher (A/Ac \ 4). When the columns are closely spaced, the
0
friction angles (uc = 45, e.g. n/nFE & 1.3). This is elastic strains of the column have a greater influence
generally why Priebe’s [30] method tends to be used and this is the reason for the slight differences in the
with conservative estimates for the friction angle of the SCFs.
granular column material. • Priebe’s [30] predicted SCFs are noticeably higher than
• Predicted n values from Castro and Sagaseta [10] and these predictions. However, it is not appropriate to use
Pulko et al. [32] are in almost perfect agreement with Priebe’s [30] method to estimate SCFs because the
one another for all modular ratios and friction angles method merely uses the SCF as a post-correction to
considered (and comparison with the FE output is again work out n2. The SCF is thus not considered an output
excellent, i.e., 0.9 \ n/nFE \ 1.1). of the method.
123
Acta Geotechnica (2014) 9:993–1011 1007
(a) 16 Priebe's n2 (ϕ' = 35) • The good agreement between FE-calculated n values
14 Priebe's n2 (ϕ' = 40)
12 Priebe's n2 (ϕ' = 45) and SCFs with those predicted by Castro and Sagaseta
Castro & Sagaseta (ϕ' = 35)
10 Castro & Sagaseta (ϕ' = 40) [10] and Pulko et al. [32] again affirms their greater
SCF
6
Pulko et al. (ϕt' = 35) applicability in design.
Pulko et al. (ϕ' = 40)
4 Pulko et al. (ϕ' = 45)
HS Model (ϕ' = 35)
2 HS Model (ϕ' = 40)
0
3 4 5 6 7 8 9 10
HS Model (ϕ' = 45) 4.2.3 Dilatancy angle of column material (wc)
A/Ac
(b) 16 Priebe's n2 (ϕ' = 35) Pulko and Majes [31], Castro and Sagaseta [10], and Pulko
14 Priebe's n2 (ϕ' = 40)
12 Priebe's n2 (ϕ' = 45) et al. [32] predict larger n values for higher dilatancy
Castro & Sagaseta (ϕ' = 35)
10 Castro & Sagaseta (ϕ' = 40) angles, wc. The n values predicted by elastic methods (e.g.
SCF
6
Pulko et al. (ϕ' = 35) Balaam and Booker [3]) and Borges et al. [8] are inde-
Pulko et al. (ϕ' = 40)
4 Pulko et al. (ϕ' = 45)
HS Model (ϕ' = 35)
pendent of the dilatancy angle. The nFE predictions have
2 HS Model (ϕ' = 40)
HS Model (ϕ' = 45)
been included in Fig. 14a–d in order to show the direct
0
3 4 5 6 7 8 9 10 influence of wc on n (higher wc values lead to higher
A/Ac
n values). The influence of the dilatancy angle (Fig. 15a–d)
(c) 16 Priebe's n2 (ϕ' = 35)
14 Priebe's n2 (ϕ' = 40) of the granular material has been examined in the range
Priebe's n2 (ϕ' = 45)
12
Castro & Sagaseta (ϕ' = 35) 0 \ wc \ 15 for Castro and Sagaseta [10] and Pulko
10 Castro & Sagaseta (ϕ' = 40)
et al. [32]. In this case, the remaining parameters have been
SCF
2.5
(d) 16 Priebe's n2 (ϕ' = 35)
HS Model (ψ = 10)
14 Priebe's n2 (ϕ' = 40) 2.0
12 Priebe's n2 (ϕ' = 45)
Castro & Sagaseta (ϕ' = 35) 1.5
10 Castro & Sagaseta (ϕ' = 40) HS Model (ψ = 15)
SCF
0 2.5
Fig. 13 SCF versus A/Ac (influence of uc); a Ec/Es = 5 b Ec/Es = 10 HS Model (ψ = 10)
2.0
c Ec/Es = 20 d Ec/Es = 40
1.5
HS Model (ψ = 15)
1.0
3 4 5 6 7 8 9 10
• The SCFs calculated using the PLAXIS 2D HS Model A/Ac
are in the range predicted by Castro and Sagaseta [10] (c) 4.0
HS Model (ψ = 0)
and Pulko et al. [32] which highlights why the 3.5
methods, but not for Priebe’s [30] method because of its 2.0 HS Model (ψ = 10)
Fig. 13a). The corresponding improvement factors also Fig. 14 nFE versus A/Ac (influence of wc); a Ec/Es = 5 b Ec/Es = 10
exhibit the largest differences for this case (Fig. 12a). c Ec/Es = 20 d Ec/Es = 40
123
1008 Acta Geotechnica (2014) 9:993–1011
Pulko et al. (ψ = 0)
SCF
1.2 Castro & Sagaseta (ψ = 15) 8 Pulko et al. (ψ = 5)
1.1 Pulko et al. (ψ = 0) 6 Pulko et al. (ψ = 10)
Pulko et al. (ψ = 5) Pulko et al. (ψ = 15)
1.0 4 HS Model (ψ = 0)
Pulko et al. (ψ = 10) HS Model (ψ = 5)
0.9 2
Pulko et al. (ψ = 15) HS Model (ψ = 10)
0.8 0 HS Model (ψ = 15)
3 4 5 6 7 8 9 10
3 4 5 6 7 8 9 10
A/Ac
A/Ac
(b) 1.6 Priebe's n2 (b) 16 Priebe's n2
1.5 Castro & Sagaseta (ψ = 0) 14 Castro & Sagaseta (ψ = 0)
1.4 Castro & Sagaseta (ψ = 5) Castro & Sagaseta (ψ = 5)
12 Castro & Sagaseta (ψ = 10)
1.3 Castro & Sagaseta (ψ = 10) Castro & Sagaseta (ψ = 15)
10
n/nFE
Pulko et al. (ψ = 0)
SCF
1.2 Castro & Sagaseta (ψ = 15)
8 Pulko et al. (ψ = 5)
1.1 Pulko et al. (ψ = 0) Pulko et al. (ψ = 10)
6
1.0 Pulko et al. (ψ = 5) Pulko et al. (ψ = 15)
Pulko et al. (ψ = 10) 4 HS Model (ψ = 0)
0.9 HS Model (ψ = 5)
Pulko et al. (ψ = 15) 2 HS Model (ψ = 10)
0.8 HS Model (ψ = 15)
3 4 5 6 7 8 9 10 0
A/Ac 3 4 5 6 7 8 9 10
A/Ac
(c) 1.6
Priebe's n2 (c) 16
1.5 Priebe's n2
Castro & Sagaseta (ψ = 0) 14 Castro & Sagaseta (ψ = 0)
1.4 Castro & Sagaseta (ψ = 5) Castro & Sagaseta (ψ = 5)
12 Castro & Sagaseta (ψ = 10)
1.3 Castro & Sagaseta (ψ = 10)
n/nFE
SCF
Pulko et al. (ψ = 0) 8 Pulko et al. (ψ = 5)
1.1
6 Pulko et al. (ψ = 10)
1.0 Pulko et al. (ψ = 5) Pulko et al. (ψ = 15)
Pulko et al. (ψ = 10) 4 HS Model (ψ = 0)
0.9 HS Model (ψ = 5)
Pulko et al. (ψ = 15) 2
0.8 HS Model (ψ = 10)
3 4 5 6 7 8 9 10 0 HS Model (ψ = 15)
A/Ac 3 4 5 6 7 8 9 10
A/Ac
(d) 1.6
Priebe's n2
1.5 Castro & Sagaseta (ψ = 0) (d) 16 Priebe's n2
1.4 Castro & Sagaseta (ψ = 5) 14 Castro & Sagaseta (ψ = 0)
Castro & Sagaseta (ψ = 5)
1.3 Castro & Sagaseta (ψ = 10) 12 Castro & Sagaseta (ψ = 10)
n/nFE
123
Acta Geotechnica (2014) 9:993–1011 1009
n/nFE
nFE
n/nFE
3.0
1.0 Castro & Sagaseta (K = 0.7)
nFE
n/nFE
3.0 1.0 Castro & Sagaseta (K = 0.7)
0.9 Castro & Sagaseta (K = 1.0)
nFE
HS Model (K = 0.44)
1.0 Castro & Sagaseta (K = 0.7)
3.0 Castro & Sagaseta (K = 1.0)
0.9
nFE
123
1010 Acta Geotechnica (2014) 9:993–1011
(a) 16 Priebe's n2
theoretically the method has been formulated assuming
14
12 Castro & Sagaseta
no dilatant behaviour of the column.
10
Pulko et al.
• Priebe’s [30] method should not be used to calculate
SCF
8
HS Model (K = 0.44)
SCFs (the SCF is merely used as a post-correction to
6
4 HS Model (K = 0.7)
work out n2).
2
HS Model (K = 1.0)
• Based on the results, it is suggested that the newest
0
3 4 5 6 7 8 9 10 methods (Castro and Sagaseta [10], Pulko et al. [32])
A/Ac offer the most reliable predictions which tend to be
(b) 16
Priebe's n2
consistently in excellent agreement with FE predictions
14
12 Castro & Sagaseta
for end-bearing columns. These design methods should
10
Pulko et al.
be used more often in geotechnical practice because
SCF
8
HS Model (K = 0.44) they give more realistic results and allow for the
6
4 HS Model (K = 0.7) consideration of significantly more input data.
2
HS Model (K = 1.0)
0
3 4 5 6 7 8 9 10 Acknowledgments The authors would like to acknowledge the
A/Ac
support provided by the Irish Research Council.
(c) 16 Priebe's n2
14
12 Castro & Sagaseta
10
Pulko et al. References
SCF
8
6 HS Model (K = 0.44)
1. Aboshi H, Ichimoto E, Enoki M, Harada K (1979) The ‘‘Com-
4 HS Model (K = 0.7)
2
pozer’’—a method to improve characteristics of soft clays by
HS Model (K = 1.0)
0 inclusion of large diameter sand columns. In: Paper presented at
3 4 5 6 7 8 9 10 the international conference on soil reinforcement: reinforced
A/Ac
earth and other techniques (Coll. Int. Renforcements des Sols.),
(d) 16 Priebe's n2
Paris, Mar 1979
14
2. Ambily AP, Gandhi SR (2007) Behavior of stone columns based
12 Castro & Sagaseta
10
on experimental and FEM analysis. J Geotech Geoenviron Eng
Pulko et al. 133(4):405–415
SCF
8
6 HS Model (K = 0.44) 3. Balaam NP, Booker JR (1981) Analysis of rigid rafts supported
4 HS Model (K = 0.7) by granular piles. Int J Numer Anal Meth Geomech 5(4):379–403
2
HS Model (K = 1.0)
4. Balaam NP, Booker JR (1985) Effect of stone column yield on
0 settlement of rigid foundations in stabilized clay. Int J Numer
3 4 5 6 7 8 9 10
A/Ac Anal Meth Geomech 9(4):331–351
5. Barksdale RD, Bachus RC (1983) Design and Construction of
Fig. 19 SCF versus A/Ac (influence of K0); a Ec/Es = 5 b Ec/Es = 10 Stone Columns Volume I. Federal Highway Administration
c Ec/Es = 20 d Ec/Es = 40 Report FHWA/RD-83/026, National Technical Information Ser-
vice, Springfield, Virginia
6. Barron RA (1948) Consolidation of fine-grained soils by drain
assumption). This may lead to some differences wells. Trans ASCE 113:718–742
between the analytical solutions and the numerical 7. Baumann V, Bauer GEA (1974) The performance of foundations
on various soils stabilized by the vibro-compaction method. Can
results, but as is evident from the results above, these
Geotech J 11(4):509–530
differences are small for Castro and Sagaseta [10] and 8. Borges JL, Domingues TS, Cardoso AS (2009) Embankments on
Pulko et al. [32]. n values and SCFs are consistently in soft soil reinforced with stone columns: numerical analysis and
agreement with the numerical output. proposal of a new design method. Geotech Geol Eng
27(6):667–679
• The parameter sensitivity study considering load level
0 9. Brinkgreve RBJ, Swolfs WM, Engin E (2011) PLAXIS 2D 2010
(pa), column friction angle (uc), column dilatancy angle Material Models Manual. PLAXIS B.V.
(wc), and the coefficient of lateral earth pressure (K) has 10. Castro J, Sagaseta C (2009) Consolidation around stone columns.
shown Priebe’s n2 [30] method to consistently over- Influence of column deformation. Int J Numer Anal Meth Geo-
mech 33(7):851–877
predict improvement factors. This explains the conser-
11. Castro J, Sagaseta C (2011) Consolidation and deformation
vative values for the column friction angle (e.g. around stone columns: numerical evaluation of analytical solu-
0
uc = 40) used in conjunction with this method in tions. Comput Geotech 38(3):354–362
practice. Additionally, the semi-empirical nature of the 12. Debats JM, Guetif Z, Bouassida M (2003) Soft soil improvement
due to vibro-compacted columns installation. In: Paper presented
method means it gives better predictions for more
at the P international workshop ‘‘geotechnics of soft soils. Theory
realistic higher modular ratios (e.g. Ec/Es = 40), K val- and practice’’, Noordwijkerhout, The Netherlands, 17–19 Sept
ues and dilatancy angles (e.g. wc = 10–15), although 2003
123
Acta Geotechnica (2014) 9:993–1011 1011
13. Degago SA (2011) On Creep during Primary Consolidation of 26. Munfakh GA, Sarkar SK, Castelli RJ (1983) Performance of a
Clays. PhD Thesis, Norwegian University of Science and Tech- test embankment founded on stone columns. In: Paper presented
nology (NTNU), Trondheim at the international conference on advances in piling and ground
14. Domingues TS, Borges JL, Cardoso AS (2007) Stone columns in treatment for foundations, London, 2–4 Mar 1983
embankments on soft soils. Analysis of the effects of the gravel 27. Nash DFT, Powell JJM, Lloyd IM (1992) Initial investigations of
deformability. In: Paper presented at the 14th European confer- the soft clay test site at Bothkennar. Géotechnique 42(2):163–181
ence on soil mechanics and geotechnical engineering, Madrid, 28. Poorooshasb HB, Meyerhof GG (1997) Analysis of behavior of
Spain, 24–27 Sept 2012 stone columns and lime columns. Comput Geotech 20(1):47–70
15. Gäb M, Schweiger HF, Kamrat-Pietraszewska D, Karstunen M 29. Priebe HJ (1976) Evaluation of the settlement reduction of a
(2008) Numerical analysis of a floating stone column foundation foundation improved by Vibro-replacement. Bautechnik
using different constitutive models. In: Paper presented at the 2nd 2:160–162 (in German)
international workshop on the geotechnics of soft soils—focus on 30. Priebe HJ (1995) The design of vibro replacement. Ground Eng
ground improvement, Glasgow, 3–5 Sept 2008 28(10):31–37
16. Goughnour RR, Bayuk AA (1979) Analysis of stone column-soil 31. Pulko B, Majes B (2005) Simple and accurate prediction of set-
matrix interaction under vertical load. In: Paper presented at the tlements of stone column reinforced soil. In: Paper presented at
international conference on soil reinforcement: reinforced earth the 16th international conference on soil mechanics and geo-
and other techniques (Coll. Int. Renforcements des Sols.), Paris, technical engineering, Osaka, Japan, 12–16 Sept 2005
Mar 1979 32. Pulko B, Majes B, Logar J (2011) Geosynthetic-encased stone
17. Goughnour RR, Bayuk AA (1979) A field study of long term columns: analytical calculation model. Geotext Geomembr
settlements of loads supported by stone columns in soft ground. 29(1):29–39
In: Paper presented at the international conference on soil rein- 33. Schanz T, Vermeer PA, Bonnier PG (1999) The hardening soil
forcement: reinforced earth and other techniques (Coll. Int. model: formulation and verification. In: Paper presented at
Renforcements des Sols.), Paris, Mar 1979 beyond 2000 in computational geotechnics. Ten years of PLAXIS
18. Greenwood DA (1970) Mechanical improvement of soils below international, Amsterdam, 18–20 Mar 1999
ground surface. In: Paper presented at the ground engineering 34. Schweiger HF, Pande GN (1986) Numerical analysis of stone
conference organised by the institution of civil engineers, Lon- column supported foundations. Comput Geotech 2(6):347–372
don, June 1970 35. Seed HB, Booker JR (1977) Stabilization of potentially liquefi-
19. Han J, Ye SL (2001) Simplified method for consolidation rate of able sand deposits using gravel drains. J Geotech Eng Div
stone column reinforced foundations. J Geotech Geoenviron Eng 103(7):757–768
127(7):597–603 36. Sexton BG, McCabe BA (2013) Numerical modelling of the
20. Hughes JMO, Withers NJ (1974) Reinforcing of soft cohesive improvements to primary and creep settlements offered by
soils with stone columns. Ground Eng 7(3):42–49 granular columns. Acta Geotechnica. doi:10.1007/s11440-012-
21. Killeen MM, McCabe BA (2010) A numerical study of factors 0205-4
affecting the performance of stone columns supporting rigid 37. Slocombe BC, Bell AL, Baez JI (2000) The densification of
footings on soft clay. In: Paper presented at the 7th European granular soils using vibro methods. Géotechnique 50(6):715–725
conference on numerical methods in geotechnical engineering, 38. Sondermann W, Wehr W (2004) Deep vibro techniques. Ground
Trondheim (Norway), 2–4 June 2010 improvement, 2nd edn. Spon Press, Abingdon
22. Lee JS, Pande GN (1998) Analysis of stone-column reinforced 39. Van Impe WF, De Beer E (1983) Improvement of settlement
foundations. Int J Numer Anal Meth Geomech 22(12):1001–1020 behaviour of soft layers by means of stone columns. In: Paper
23. Leroueil S, Lerat P, Hight DW, Powell JJM (1992) Hydraulic presented at the 8th European conference on soil mechanics and
conductivity of a recent estuarine silty clay at Bothkennar. foundation engineering, Helsinki, Finland, 23–26 May 1983
Géotechnique 42(2):275–288 40. Van Impe WF, Madhav MR (1992) Analysis and settlement of
24. Lewis RW, Schrefler BA (1987) The finite element method in the dilating stone column reinforced soil. Österreichische Ingenieur-
deformation and consolidation of porous media. Wiley, New und Architekten-Zeitschrift 137(3):114–121
York 41. Watts KS, Johnson D, Wood LA, Saadi A (2000) An instru-
25. McKelvey D, Sivakumar V, Bell A, Graham J (2004) Modelling mented trial of vibro ground treatment supporting strip founda-
vibrated stone columns in soft clay. Proc ICE Geotech Eng tions in a variable fill. Géotechnique 50(6):699–708
157(3):137–149
123