Download as pdf or txt
Download as pdf or txt
You are on page 1of 20

Geosynthetics International, 2014, 21, No.

A parametric study of geosynthetic-reinforced


column-supported embankments
N. N. S. Yapage1 and D. S. Liyanapathirana2
1
PhD Candidate, School of Computing, Engineering and Mathematics, University of Western Sydney,
NSW 2751, Australia, Telephone: +61 2 4736 0450; Telefax: +61 2 4736 0833,
E-mail: n.yapage@uws.edu.au
2
Associate Professor in Geotechnical Engineering, School of Computing, Engineering and Mathematics,
University of Western Sydney, NSW 2751, Australia, Telephone: +61 2 4736 0653;
Telefax: +61 2 4736 0833, E-mail: s.liyanapathirana@uws.edu.au

Received 31 July 2013, Revised 24 November 2013, Accepted 7 March 2014

ABSTRACT: This paper presents an extensive parametric study for geosynthetic-reinforced column-
supported (GRCS) embankments. The analysis was carried out using the finite-element method
based on the coupled-hydraulic modelling, incorporating the full geometry of the embankment and
strain-softening behaviour of deep cement mixed (DCM) columns. The influencing factors
considered in the parametric study were elastic modulus of DCM columns, area replacement ratio
(ARR) based on spacing and diameter of DCM columns, elastic modulus and permeability of soft
foundation soil, stiffness of the geosynthetic reinforcement, thickness of the platform layer and
friction angle of the fill material. Embankment behaviour during the parametric study was
investigated by comparing maximum total and differential settlements, maximum tension in the
geosynthetic, maximum lateral deformation of columns, efficiency coefficient of columns and
arching ratio. The influencing factors were ranked based on their degree of influence on the
behaviour of the embankment. Results show that column spacing and diameter, which control the
ARR, and elastic modulus of soft soil were the most important design parameters to be considered
in GRCS design procedures. In comparison with an unreinforced column-supported embankment,
by using a geosynthetic layer, lateral deformations can be reduced by about 10 times and
differential settlements at the base of the embankment can be reduced by about three times. In
addition, numerical results show that a thin platform layer was more efficient in transferring the
embankment load to columns than a thick platform layer.

KEYWORDS: Geosynthetics, Column-supported embankment, Coupled consolidation analysis,


Reinforcement, Arching ratio, Area replacement ratio, Efficiency coefficient of columns

REFERENCE: Yapage, N. N. S. and Liyanapathirana, D. S. (2014). A parametric study of


geosynthetic-reinforced column-supported embankments. Geosynthetics International, 21, No. 3,
213–232. [http://dx.doi.org/10.1680/gein.14.00010]

runways and other civil engineering projects over soft


1. INTRODUCTION ground.
The rapid growth of infrastructure development in urban Investigation of GRCS embankment behaviour using
and metropolitan areas to meet the demand of increasing numerical modelling based on the finite element or finite
population has often been hindered by the scarcity of difference method is economical and less time consuming
suitable land. Therefore construction on marginal and soft compared to experimental methods employing centrifuge
ground with poor soil conditions has become a necessity. modelling or full-scale embankments in the field. There-
However, these soil deposits have low shear strength and fore, two and three-dimensional finite element models are
high compressibility, and their properties should be im- widely used to enhance the understanding of the behav-
proved before the construction begins. Deep cement mix- iour of GRCS embankments (Stewart et al. 2004; Navin
ing (DCM) is often used to improve the strength and 2005; Madhyannapu et al. 2006; Navin and Filz 2006;
stiffness of soft soils and consequently DCM column- Filz 2007; Huang et al. 2009; Huang and Han 2009;
supported embankments utilising geosynthetic reinforce- Ariyarathne et al. 2013a). However, modelling of DCM
ments are widely used to construct roads, railways, columns with an elastic perfectly plastic Mohr–Coulomb
1072-6349 # 2014 Thomas Telford Ltd 213

Downloaded by [ UNIVERSIDADE CRUZEIRO DO SUL] on [12/09/16]. Copyright © ICE Publishing, all rights reserved.
214 Yapage and Liyanapathirana

model is insufficient to investigate the overall behaviour ments are designed using composite foundation theories
of embankments beyond the yielding of cement-admixed which are based on equal strain concept between soft soil
soils. Cement-admixed soils show strain-softening behav- and DCM columns. Most of these design procedures use
iour beyond yielding. Therefore, if column yielding is the composite modulus of the improved ground, which is
likely to occur, numerical models with linear elastic based on the area replacement ratio, to calculate the
perfectly plastic column behaviour underestimate the total settlement of the ground at the base and the crest of the
and differential settlements, produce unrealistic failure embankment. Hence, the load distribution between col-
patterns and overestimate bearing capacities (Yapage et al. umns and soft foundation soil has not been considered
2012). Nevertheless previous studies have not considered within the stabilised area. However, many field measure-
the strain-softening behaviour of DCM columns. ments and numerical results have shown that the differ-
The majority of embankments are constructed over very ential settlement occurs between DCM columns and
soft clay with ground water table closer to the surface. surrounding soft soil (Yu and Gao 2002; Huang and Han
Therefore, consolidation is inevitable during and after 2009; Huang et al. 2009). Therefore, equal strain concept
completion of the embankment construction and important for calculating settlement of cement stabilised ground is
to understand the performance of embankments under not valid.
serviceability conditions. Huang and Han (2009) pointed Yapage et al. (2013) showed that currently available
out that most of the numerical modelling has considered design procedures yield inconsistent results. This is mainly
long-term or short-term behaviour with drained or un- due to various assumptions made in the process of develop-
drained strength parameters, without incorporating the ment of analytical solutions in these guidelines. Therefore,
time-dependent behaviour of these embankments due to further research in this area is necessary to improve the
consolidation of soft soils. There can be significant understanding of the behaviour of GRCS embankments.
differences in load transfer during and after construction, Consequently, in this study, a parametric study is carried
if the analysis is carried out as a drained analysis or a out to investigate the influence of: (i) elastic modulus, area
coupled consolidation analysis. Only a few numerical replacement ratio based on spacing and diameter of DCM
models are available in the literature coupling mechanical columns, (ii) elastic modulus and permeability of soft
and hydraulic modelling (Huang et al. 2009; Huang and foundation soil, (iii) tensile stiffness of geosynthetic, (iv)
Han 2009, 2010; Borges and Marques 2011). thickness of the platform layer and (v) friction angle of
Parametric studies are important to investigate the embankment fill material on the time-dependent behaviour
embankment behaviour with the variation of geometric of the embankment. A two-dimensional plane-strain finite-
properties of the embankment, and material properties for element model with coupled mechanical and hydraulic
the soft ground, DCM columns and geosynthetic. How- modelling has been used for the parametric study. In
ever, a limited number of parametric studies can be found addition, the strain-softening behaviour of DCM columns
in the literature combining both time-dependent behaviour has been incorporated into the analysis aiming to investi-
and the effect of lateral movement of GRCS embankments gate the GRCS embankment behaviour during column
considering the full cross-sectional geometry of the em- softening if the DCM columns yield during the variation of
bankment (Huang and Han 2010). The majority of parameters. However, strain softening is not significant for
previous studies used axisymmetric, two-dimensional or the parameter ranges selected for this study. Therefore
three-dimensional unit cells to conduct the parametric strain softening did not affect the final conclusion signifi-
studies. These unit cell models reduce the number of cantly.
elements in the finite element mesh, which in turn uses
less computational time compared to a model considering
the full cross-sectional geometry of the embankment.
2. DESCRIPTION OF THE PROBLEM
However, none of these unit cell models describe the real The geometry of the numerical model and mechanical and
embankment behaviour because they do not replicate the hydraulic boundary conditions for the baseline case are
actual load transfer mechanism of the GRCS embankment shown in Figure 1. Due to the symmetry of the embank-
system. Especially these unit cell models do not include ment, only half of the geometry was considered for the
the group effect of DCM columns, lateral spreading of fill analysis. Foundation soil below the embankment consisted
layers and subsequent lateral deformation of columns. of a 12 m thick soft clay layer underlain by a 3 m thick
A number of design procedures based on analytical silty sand layer. To improve the soft ground, DCM
solutions are available in the literature for the column- columns with 800 mm diameter at 2 m centre-to-centre
supported embankments (e.g. BSI (2010); German method spacing in a square array were used. The embankment
by Kempfert et al. (2004); Hewlett and Randolph (1988); height was 5 m and crest width was 14 m. The width at
Swedish method by Carlsson (1987)). However, some of the base was 24 m. The geosynthetic layer was placed
them, which are frequently used for designs, have not 0.25 m above the columns. The ground water table was
considered the contribution of soft foundation soil in the assumed to be at the existing ground level. The traffic
load transfer mechanism (e.g. BSI (2010); Hewlett and load was assumed to be 12 kPa. The finite-element mesh
Randolph (1988); Swedish method by Carlsson (1987)). for the existing ground was extended 50 m beyond the toe
The concept of arch formation above DCM columns, and of the embankment to minimise the boundary effects. The
their shape and evolution during loading are not consistent numerical procedure used in this parametric study has
in these guidelines (Yapage et al. 2013). GRCS embank- been validated by Yapage et al. (2012, 2014a) using field
Geosynthetics International, 2014, 21, No. 3

Downloaded by [ UNIVERSIDADE CRUZEIRO DO SUL] on [12/09/16]. Copyright © ICE Publishing, all rights reserved.
A parametric study of geosynthetic-reinforced column-supported embankments 215
10 m 14 m
Surcharge
Embankment fill Cell considered

0.25 m

5m
Geosynthetic 1 Node A
GWT 2 Node B
Node C Platform layer

0.25 m

DCM columns

12 m
Soft clay

3m
Silty sand

50 m 2m 2m 2m

Figure 1. Dimensions and boundary conditions used in the numerical model for the embankment

measurements given by Forsman et al. (1999) for a GRCS differential settlement over column heads and clay surface,
embankment constructed in Finland. Although the same maximum tension in the geosynthetic, lateral displacement
procedure was adopted here for embankment construction of columns, efficiency coefficient of columns and stress
simulation, constitutive behaviour for DCM columns, concentration ratio, incorporating consolidation and strain
foundation soil and geosynthetic reinforcement, and con- softening behaviour for DCM columns, for a range of
solidation behaviour of columns and foundation soil, the material and geometric properties for GRCS embank-
column configuration and embankment geometry used in ments. He confirmed that two-dimensional plane-strain
the parametric study were different to the case study model based on the equivalent area method can capture
presented by Forsman et al. (1999). three-dimensional behaviour with reasonable accuracy for
As the three-dimensional problem was simulated using the serviceability performance of GRCS embankments.
a two-dimensional plane strain model in this study, Furthermore, when analysing a large number of cases in a
isolated DCM columns were modelled as DCM walls. The parametric study, it is not practicable to adopt a three-
properties of isolated columns were kept similar to the dimensional modelling procedure, which consumes exten-
original values, but the widths of the walls were changed sive amount of time for a single analysis.
based on the equivalent area method initially proposed by When the column-soil interaction was included, analysis
Bergado and Long (1994) as shown below gave convergence problems for some parameter combina-
tions. Furthermore, the analysis time increased signifi-
d 2
D9 ¼ (1) cantly. As a large number of models were used for the
4s parametric study, the interaction between columns and soil
where D9 is the width of the equivalent DCM walls, d is was ignored assuming that it was not significant. On the
the diameter of isolated DCM columns and s is the centre- other hand, when a column is created by deep cement
to-centre spacing of columns. mixing it can be assumed that a perfect bond is created
However, there are several approaches available in the between the columns and surrounding soil. This is differ-
literature to transform three-dimensional problems into ent to the case when a timber or concrete pile is driven
two-dimensional plane strain models. Ariyarathne et al. into the soil or when cast-in-place concrete piles are used
(2013a) carried out a comparison of the different methods in a soft clay deposit, where slip between piles and soil
available for the conversion of three-dimensional problems can be significant. Therefore, this was a reasonable
into two-dimensional finite-element models considering a assumption for the present study.
pile-supported embankment. They showed that the geome- The mechanical boundary conditions used for the analy-
trical idealisation based on the equivalent area method has sis are illustrated in Figure 1. With regard to hydraulic
the ability to predict three-dimensional numerical model boundary conditions, a zero pore pressure boundary condi-
results and field measurements with a reasonable accuracy tion was applied at the top of the soft clay layer and hence
in comparison with a number of other methods available it was permeable. Due to the symmetrical boundary
in the literature. In addition, the case studies simulated by condition imposed at the centre of the embankment, the
Huang et al. (2009) and Yapage et al. (2014a, 2014b) also hydraulic boundary condition at the right-hand side was
confirmed that the equivalent area method was appropriate impervious as drainage cannot take place through that
to convert a three-dimensional column-supported embank- boundary. At the bottom of the mesh, an impervious
ment into a plane strain model. In addition, Yapage (2013) boundary condition was employed assuming the under-
carried out a detailed comparison of two- and three- lying bed rock had a very low permeability compared to
dimensional numerical modelling results for total and the silt layer. An impermeable boundary condition was
Geosynthetics International, 2014, 21, No. 3

Downloaded by [ UNIVERSIDADE CRUZEIRO DO SUL] on [12/09/16]. Copyright © ICE Publishing, all rights reserved.
216 Yapage and Liyanapathirana

applied to the left-hand side boundary because it was far a friction coefficient of 0.8. Shear stress at the interface
away from the loaded area. was limited using a limiting shear stress, which was equal
The embankment was constructed at a rate of 1 m per to the friction coefficient times the normal stress at the
month and service lifewa simulated for 30 years. After interface. Numerically simulating pullout tests, Ariyar-
placing each fill layer with a thickness of 1 m, a waiting athne et al. (2013b) showed that the above-mentioned
period of one month was imposed. interface procedure was appropriate to model the geosyn-
thetic-platform fill interface. Embankment fill, platform
fill, soft soil, and silty sand were modelled as linearly
3. PARAMETERS FOR THE MATERIAL elastic perfectly plastic materials using the Mohr–
MODEL Coulomb failure criteria. Eight-node plane strain elements
A comprehensive literature review was conducted to select with pore pressure degrees of freedom at corner nodes
the typical material properties and geometry for the were used for the soft soil, silty sand and DCM columns.
numerical model. The parameters selected for DCM The same element without pore pressure degrees of free-
columns, soft clay, silty sand, geosynthetic and embank- dom was used for fill and platform layers.
ment fill layers are given in Table 1 (EPRI 1990; The softening behaviour of the DCM columns was
GeoTesting Express 1996; Budhu 2000; Das 2001; Bruce modelled using peak and residual friction angles and
2001). The total unit weight of the cement-admixed clay cohesions, and corresponding plastic deviatoric strains
(by dry mixing method) increases about 3–15% of (Yapage et al. 2014a). Based on the comprehensive
unstabilised clay (CDIT 2002). Therefore, the density of literature survey carried out by authors for cement-
the DCM columns was considered to be 10% higher than stabilised soils, the residual softening index, which is
that of soft clay. Unconfined compressive strength (UCS) defined as the ratio between c9res =c9p and 9res =9p , vary
was assumed to be 750 kN/m2 (Baker 2000; Japanese between 0.4–0.7 and 0.4–0.8, respectively. Therefore in
Geotechnical Society 2000; Bruce 2001; Jacobson et al. this study, a residual softening index of 0.6 was assumed
2003). The elastic modulus of DCM columns was deter- when deriving residual values of cohesion and friction
mined using the correlation E ¼ 200qu (GeoTesting Ex- angle (9res ¼ 188 and c9res ¼ 130 kPa). The plastic deviato-
press 1996; Baker 2000; Broms 2003; Jacobson et al. ric strain terms at peak (pd,peak ) and residual (pd,res ) states
2005; Navin and Filz 2005). A failure envelope with both are assumed to be 2 and 12%, respectively (Yapage et al.
cohesion and friction angle considering effective stress 2014a).
parameters was used for the columns and the shear
strength was calculated using c9 ¼ 0.289qu (Broms 2003).
The tensile strength of DCM columns was assumed to be 4. PARAMETRIC STUDY
15% of qu (CDIT 2002; EuroSoilStab 2002; Navin and
Filz 2005). 4.1. Preface
Elasto-plastic Mohr–Coulomb constitutive model is In the parametric study, some of the design considerations
extended through the user-defined field subroutine and potential problems which can be expected during and
(USDFLD) in ABAQUS/Standard (ABAQUS 2010) finite- after construction of the embankment are discussed. As
element program to address the strain-softening behaviour given in Table 2, one parameter is varied at a time within
of cement stabilised soil. More details on material exten- the range of typical values from the baseline case (bold
sion, numerical model and calibration are described in font in Table 2) to assess the influence of each parameter
Yapage et al. (2014a). The geosynthetic reinforcement was on the embankment behaviour. Parameters used for the
modelled using three-node truss elements with a linear softening model are kept constant throughout the para-
elastic perfectly plastic material behaviour. The interface metric study.
between geosynthetic and platform fill was modelled using The effect of stiffness of the geosynthetic, column and

Table 1. Material properties used in the baseline case

Material Model and element type Parameters

Soft clay Elastic-perfectly plastic (Mohr–Coulomb) E ¼ 2 MPa, c9 ¼ 0 kPa, ct ¼ 0 kPa, k ¼ 109 m/s 9 ¼ 258, e ¼ 2.0,  ¼ 0.3,
ª ¼ 18 kN/m3
Silty sand Elastic-perfectly plastic (Mohr–Coulomb) E ¼ 50 MPa, c9 ¼ 0 kPa, ct ¼ 0 kPa, k ¼ 106 m/s 9 ¼ 338, e ¼ 2.5,  ¼ 0.3,
ª ¼ 20kN/m3
Embankment fill Elastic-perfectly plastic (Mohr–Coulomb) E ¼ 30 MPa, c9 ¼ 3 kPa, 9 ¼ 350 ,  ¼ 0.3, ª¼ 20 kN/m3
Platform layer Elastic-perfectly plastic (Mohr–Coulomb) E ¼ 20 MPa, c9 ¼ 5 kPa, 9 ¼ 320 ,  ¼ 0.33, ª ¼ 20 kN/m3
DCM columns Strain-softening (extended Mohr–Coulomb) E ¼ 150 MPa, c9 ¼ 216.75 kPa, ct ¼ 112.5 kPa, k ¼ 109 m/s 9 ¼ 300 ,
e ¼ 2.0,  ¼ 0.35, ª ¼ 20 kN/m3
Geosynthetic Elastic–perfectly plastic E ¼ 500 MPa, J ¼ Et, T ¼ 300 kN/m, ci ¼ 0.8, J ¼ 2000 kN/m,  ¼ 0.3,
t ¼ 4 mm

J, stiffness of the geosynthetic; t, thickness of the geosynthetic; ci , interaction coefficient between geosynthetic and platform fill; ª, unit weight
of soil; ct , tensile strength of soil; k, permeability; E, elastic modulus; c9, effective cohesion of soil; 9, effective friction angle; e, void ratio;
, Poisson’s ratio; T, yield strength of geosynthetic.

Geosynthetics International, 2014, 21, No. 3

Downloaded by [ UNIVERSIDADE CRUZEIRO DO SUL] on [12/09/16]. Copyright © ICE Publishing, all rights reserved.
A parametric study of geosynthetic-reinforced column-supported embankments 217

Table 2. Range of properties used for parametric study

Parameter Influencing factor Range of value

DCM columns Elastic modulus (MPa) 50, 100, 150, 200, 250
Spacing (m) 1, 1.5, 2, 2.5, 3, 3.5
Diameter (m) 0.6, 0.8, 1.0, 1.2
Soft clay Elastic modulus (MPa) 1, 2, 3, 4, 6, 8, 10, 12
Permeability (m/s) 106 , 107 , 1029 , 1010 , 1012
Geosynthetic Tensile stiffness (kN/m) 1000, 1500, 2000, 3000, 5000, 10000
Distance from the top of the DCM columns (m) 0.05, 0.15, 0.25, 0.35, 0.45
Embankment Friction angle of embankment fill material (o ) 25, 30, 33, 35, 37, 40

soft soil modulus, friction angle of embankment fill foundation soil. The arching mechanism transfers the
material, area replacement ratio based on column spacing major portion of the embankment load to the columns.
and diameter, and soft soil permeability on the perform- The soil arching ratio, r, is widely used to estimate the
ance of embankment were investigated. Results for the effect of soil arching in embankment fill (Han and Gabr
settlement, tension in geosynthetic, lateral deformation 2002; Borges and Marques 2011) and it is defined as
and efficiency coefficient of columns and arching ratio follows
were extracted from the numerical study at the end of p
construction (short-term behaviour) and after 30 years of r¼ (5)
(ªfill H þ q)
service (long-term behaviour).
For most of the cases considered in the parametric where p is the average vertical stress applied on the
study, the differential settlement at the crest level was geosynthetic over the soft soil within the unit cell. The
nearly zero. Therefore, only the maximum settlement at unit cell considered for calculating efficiency coefficient
the crest is presented. Settlement was measured at nodes of columns and arching ratio was the same cell as shown
A and B at the base, at the node directly above node B at in Figure 1.
the crest and lateral deformation was measured at node C The results were initially presented using dimensionless
as shown in Figure 1. The maximum tension in the parameters (e.g. stiffness ratio and permeability ratio).
geosynthetic generally occurs at the edge of column heads. However, some assessing parameters (e.g. maximum set-
Therefore maximum tension in the geosynthetic was tlement and lateral deformations) behaved differently with
recorded at the edge of the column at the centre line of dimensionless parameters depending on which parameter
the embankment. Borges and Marques (2011) used a was changed (e.g. varying Es or Ec when calculating
column efficiency coefficient to assess the proportion of Ec /Es ). This is due to the complex nature of the boundary
embankment load transferred to the columns, which takes value problem of GRCS embankments involving nonlinear
into account the soil arching within the embankment fill behaviour of materials. Hence it is not possible to come
layers, load transfer from geosynthetic to columns by up with a simple set of dimensionless parameters for this
membrane action and load transfer from soft soil to parametric study and results are presented using absolute
columns due to the shear stresses developed at the values.
column–soft soil interface due to soil settling more than
columns. This coefficient is defined as 4.2. Influence of elastic modulus of columns
Qc The influence of modulus of DCM columns, Ec , on
f ¼ (2) assessing factors was investigated by varying the column
Qt
modulus from 25 to 250 MPa. This range represents both
where Qc is the total load supported by a column and Qt upper and lower bounds for Ec of cement-stabilised soils
is the total load of the embankment fill and traffic load using the dry mixing method.
applied over a unit cell relevant for a column as shown in Figure 2 shows the maximum settlement at the clay
Figure 1. Qc and Qt can be calculated using Equations 3 surface and DCM column head at the base and at the crest
and 4, respectively. of the embankment for the short-term and long-term
Qc ¼  column Acolumn (3) conditions with Ec : The increasing Ec reduces the total
settlement at the base of the embankment at both DCM
Qt ¼ (ªfill H þ q)Aunit cell (4) column heads and surface of the foundation soil, and the
crest of the embankment. The post-construction settlement
where  column is the average vertical stress applied on a due to consolidation decreases with increasing Ec but this
column, Acolumn is the area of a column, ªfill is the unit settlement decrease is negligible beyond a limiting Ec ,
weight of the fill material, H is the embankment height, q which is about 200 MN/m2 :
is the traffic load and Aunit cell is the area of a unit cell. With increasing Ec , the stiffness of the DCM columns
Soil arching effect within embankment fill takes place increases and subsequently the coefficient of consolidation
due to the shear stresses mobilised within the fill due to of columns increases (due to lowering of compressibility).
an increase in differential settlement between columns and This will increase the degree of consolidation of the
Geosynthetics International, 2014, 21, No. 3

Downloaded by [ UNIVERSIDADE CRUZEIRO DO SUL] on [12/09/16]. Copyright © ICE Publishing, all rights reserved.
218 Yapage and Liyanapathirana
250
Maximum settlement (mm) Clay surface (long term) the total settlement at both crest and base of the embank-
Clay surface (short term)
200 Column head (long term)
ment.
Column head (short term) According to Figure 4, efficiency coefficient of col-
150 Crest (long term) umns for both long-term and short-term cases increases
Crest (short term)
with increasing Ec , while the arching ratio decreases.
100 During the consolidation, effective stress in both columns
and clay layer increases. However, total stress on the clay
50
decreases during the consolidation process. As a result, an
0
additional total stress is transferred to the columns from
0 50 100 150 200 250 300 the clay layer increasing the efficiency of columns at the
Column modulus (MN/m2) long term than that of short term. With increasing Ec , the
influence of differential settlement on embankment behav-
Figure 2. Maximum settlement at the base and the crest with iour dominates due to increased stiffness contrast between
column modulus, Ec columns and soft clay (Figure 3). Consequently this will
assist in developing more soil arching above the columns,
columns, increasing the rate of consolidation as shown in transferring more embankment load to the columns. This
Figure 2, irrespective of the constant permeability for results in an increase in efficiency coefficient of the
columns and foundation soil. Hence, the settlement differ- columns and a decrease in arching ratio. It is important to
ence between long-term and short-term cases decreases mention here that r ¼ 0 represents the full arching (i.e.
with increasing Ec : At the same time this increase in total embankment and traffic loads are transferred to
column stiffness causes stress concentration over columns columns) and r ¼ 1 represents no soil arching. In conclu-
and reduces the loads transferred to the clay layer. Hence sion, column modulus has a significant influence on the
settlement at the clay surface decreases with increasing loads transferred to the columns due to arching. However,
Ec : Although more loads are transferred to columns with BS8006 (BSI 2010), which is based on the Marston’s
increasing Ec , settlement of columns is reduced due to formula for positive projecting subsurface conduits, and
increased stiffness of columns. However, increasing Ec the research carried out by Hewlett and Randolph (1988),
beyond the limiting value will not change settlements and does not take into account the influence of column
hence produce overconservative and uneconomical em- stiffness (or modulus) in the evaluation of loads trans-
bankment designs. ferred to columns due to arching.
Maximum differential settlement at the base is impor- The interesting finding from this part of the parametric
tant as it transfers part of that settlement to the crest of study is that the maximum tension in the geosynthetic
the embankment, subsequently influencing the smoothness layer increases despite the decrease in arching ratio with
of the embankment crest, which is important for service- Ec for both short- and long-term cases as illustrated in
ability performance of the embankment. Figure 3 illus- Figure 5. This is mainly due to the increase in differential
trates the effect of Ec on this parameter. Differential settlement between columns and clay due to larger
settlement between columns and the clay at the base stiffness contrast between the columns and clay as shown
continuously increases with increasing column modulus. in Figure 3.
These results confirm that design procedures based on In addition, the increase in loads transferred to the
equal strain method, which assumes that both columns and columns increases the normal stress transferred to the
soft soil undergo the same total vertical strain do not geosynthetic layer over columns, which in turn increases
describe the settlements, and loads transferred to columns the degree of fixity applied to the geosynthetic over the
and soft soil in GRCS embankments appropriately. The column heads. Hence, the shearing resistance,
differential settlement at the base can be decreased by  ¼  9 tan , where  is the mobilised shearing resistance,
reducing the modulus difference between columns and soft
soil. However, the reduced column modulus will increase
0.8 0.8
Efficiency coefficient of columns, f

Long term (f ) Long term (r)


50
Long term Short term (f ) Short term (r)
0.7 0.7
Differential settlement (mm)

Short term
Arching ratio, r

40
0.6 0.6
30
0.5 0.5
20
0.4 0.4
10
0.3 0.3
0 0 50 100 150 200 250 300
0 50 100 150 200 250 300 Column modulus (MN/m2)
Column modulus (MN/m2)
Figure 4. Efficiency coefficient of columns and arching ratio
Figure 3. Maximum differential settlement at the base with with column modulus, Ec : f, column efficiency coefficient;
column modulus, Ec r , soil arching ratio
Geosynthetics International, 2014, 21, No. 3

Downloaded by [ UNIVERSIDADE CRUZEIRO DO SUL] on [12/09/16]. Copyright © ICE Publishing, all rights reserved.
A parametric study of geosynthetic-reinforced column-supported embankments 219
35
Long term during the consolidation process, due to the dissipation of
excess pore water pressure, they rebound laterally towards
Maximum tension kN/m)

Short term
the centre of the embankment reducing the lateral defor-
30
mation of soil and columns for the long-term case as
shown in Figure 6. This may cause some stress release in
the geosynthetic layer resulting in a decrease in tension.
25
As explained before, this may be the reason for lower
geosynthetic tension for the long-term case than that for
the short-term case when Ec is less than 180 kN/m2 in
20 Figure 5. When Ec is high, lateral deformation and subse-
0 50 100 150 200 250 300
quent lateral rebound for columns are less in comparison
Column modulus (MN/m2)
with that with a lower Ec as shown in Figure 6. This is
Figure 5. Maximum tension in the geosynthetic with column mainly due to the increase in bending stiffness, Ec I, of
modulus, Ec columns with an increasing Ec , where I is the area
moment of inertia of columns. Therefore increasing Ec
can decrease the lateral deformation of columns and soil
 9 is the effective contact pressure and is the friction in between columns and these results indirectly show that
angle, increases at the column–geosynthetic interface. the contribution of Ec on reduction in lateral spreading in
Hence the tension in the geosynthetic increases even fill layers.
though the contribution from lateral deformation (Figure
7) and arching ratio towards tension developed in the 4.3. Influence of column spacing and diameter
geosynthetic decreases with increasing Ec : Due to in- Column diameter, d, and column spacing, s, are two of
creased degree of fixity at the column heads, the maxi- most important design parameters in the design process of
mum tension is normally recorded adjacent to column GRCS embankments, because d and s control the area
heads. replacement ratio (ARR) of the improved ground. There-
Beyond an Ec value of 180 MN/m2 , maximum tension fore, in this section, the influence of ARR on the behav-
for the short-term case is lower than the tension for the iour of GRCS embankments was investigated by varying d
long-term case. DCM columns consolidate faster than the and s, as shown in Table 2. For the range of values
surrounding clay because they have a higher coefficient of selected for d, ARR varies from 7 to 28% and for the
consolidation with lower compressibility. Therefore a range of values selected for s, ARR varies between 4 and
major portion of column settlement occurs during the 50%. Results are presented for ARR obtained by changing
construction stage but for clay, settlement is not complete s and d in Figures 7–12. In these analyses, the embank-
at the end of construction. However, for the long-term ment base width has not been changed. Only the number
case, excess pore pressure developed in soft soil has of columns has been changed by keeping a column at the
dissipated and consolidation is almost complete. Conse- centre line of the embankment.
quently differential settlement for the long-term case is Figure 7 shows the variation of maximum settlement
higher than that for the short-term case as shown in Figure over columns and clay layer at the base and at the crest of
3 and also the tension in the geosynthetic for the long- the embankment for short-term and long-term conditions
term case is higher than that for the short-term case. For with ARR. When ARR increases, maximum settlement
Ec less than 180 MN/m2 , this trend has reversed due to the over the base and crest of the embankment decreases
combined effect of degree of consolidation as described significantly. As shown in Figure 8, the differential
before and the lateral rebound of columns as shown in settlement over the base also notably decreases with
Figure 6 and described in the following section. increasing ARR.
During loading, soil and columns laterally displace, and Figure 9 shows the variation of maximum tension in the
geosynthetic with ARR. According to these results, the
maximum tension in the geosynthetic reinforcement de-
180
Maximum lateral deformation (mm)

Long term creases with the increasing ARR due to reduction in loads
Short term transferring to the geosynthetic layer. This will be ex-
160
plained in the following paragraph and further justified
140 using the arching ratio later in this section. Additionally,
the lateral deformation increases when ARR decreases as
120 shown in Figure 11 and this also contributes to increase
the tension in the geosynthetic layer at lower values of
100 ARR.
When ARR decreases due to increasing s or decreasing
80
d, the loads transferred to clay and geosynthetic increases.
0 50 100 150 200 250 300
Column modulus (MN/m2)
In other words, the increased s or decreased d creates a
large soil wedge underneath the soil arch formed within
Figure 6. Maximum lateral deformation of the toe with the embankment fill between two columns. The weight of
column modulus, Ec this wedge is supported by the geosynthetic layer and the
Geosynthetics International, 2014, 21, No. 3

Downloaded by [ UNIVERSIDADE CRUZEIRO DO SUL] on [12/09/16]. Copyright © ICE Publishing, all rights reserved.
220 Yapage and Liyanapathirana
300 Distance to centre from crest edge of the embankment (m)
Clay surface (long term)
Maximum settlement (mm) 0 2 4 6 8 10 12 14
250 Clay surface (short term)
0
Column head (long term)
Column head (short term)
200
Crest (long term) 50
Crest (short term)

Settlement (mm)
150
100
100
150
50

0 200 50% 22%


0 10 20 30 40 50 60 13% 8%
Area replacement ratio (%) 6% 4%
(a) 250
300
Clay surface (long term) Figure 10. Vertical settlement at the crest with ARR
Maximum settlement (mm)

250 Clay surface (short term) calculated based on s for the long-term case
Column head (long term)
200 Column head (short term)
Crest (long term)
200

Maximum lateral deformation (mm)


150 Crest (short term) Long term (s)
Short term (s)
160 Long term (d)
100
Short term (d)
50 120

0
80
0 10 20 30 40 50 60
Area replacement ratio (%)
(b) 40

Figure 7. Maximum settlement at the base and the crest with 0


area replacement ratio, ARR: (a) s as the influencing 0 10 20 30 40 50 60
Area replacement ratio (%)
parameter (b) d as the influencing parameter
Figure 11. Maximum lateral deformation of the toe with area
replacement ratio, ARR
80
Long term (s)
Differential settlement (mm)

Short term (s) 1.0 1.0


Efficiency coefficient of columns, f

60 Long term (d)


Short term (d) Long term (f-s)
0.8 0.8
Short term (f-s)

Arching ratio, ρ
40 Long term (ρ-s)
0.6 Short term (ρ-s) 0.6
Long term (f-d)
20 Short term (f-d)
0.4 0.4
Long term (ρ-d)
Short term (ρ-d)
0 0.2 0.2
0 10 20 30 40 50 60
Area replacement ratio (%)
0 0
0 10 20 30 40 50 60
Figure 8. Maximum differential settlement at the base with Area replacement ratio (%)
area replacement ratio, ARR: d, column diameter; s, centre-
to-centre spacing of column Figure 12. Efficiency coefficient of columns and arching
ratio with area replacement ratio, ARR

50
Long term (s) soft soil. Furthermore, when column spacing increases,
Short term (s) total embankment load is shared by a lesser number of
Maximum tension (kN/m)

40
Long term (d) columns, which results in an increase in load supported by
Short term (d) a column. These facts contribute to an increase in
30
differential and total settlements at the crest and the base
20 of the embankment, and tension in the geosynthetic with
decreasing ARR.
10 In addition, post-construction settlement over clay and
columns increase rapidly when ARR is lower than 13%. It
0 was also observed that the columns started to yield when
0 10 20 30 40 50 60
Area replacement ratio (%)
ARR goes below 13%, which results in softening of the
DCM columns. During consolidation, yielding continues
Figure 9. Maximum tension in the geosynthetic with area due to increasing plastic deviatoric strain. Hence
replacement ratio, ARR strain-softening of DCM columns continues until the end
Geosynthetics International, 2014, 21, No. 3

Downloaded by [ UNIVERSIDADE CRUZEIRO DO SUL] on [12/09/16]. Copyright © ICE Publishing, all rights reserved.
A parametric study of geosynthetic-reinforced column-supported embankments 221

of the consolidation process. With the strain-softening with ARR, although differential settlement decreases with
behaviour, column bearing capacity continuously de- ARR. Therefore, it can be concluded that the contribution
creases, transferring part of the load, which was previously of differential settlement to arching is not prominent in
supported by columns, to the geosynthetic and soft clay. this case but other factors such as increased column
This entire phenomenon is due to the breakage of the diameter and reduced column spacing has more control
cement–soil structure beyond yielding and leads to a over soil arching.
sudden increase in post-construction maximum settlement Furthermore, it is important to mention here that,
when ARR is lower than 13%. However, this phenomenon although strain-softening occurred below 13% of ARR,
increases the tension for the long-term case slightly com- few elements of some columns yielded. Strain-softening
pared with that for the short-term case, when ARR is increases during consolidation and it should lower the
lower than 13% as shown in Figure 9. loads transferred to the columns (efficiency coefficient)
Figure 10 shows the long-term settlement profile over while increasing the loads transferred to the geosynthetic
the crest with ARR to investigate the effect of ARR on (arching ratio). However, in this case strain-softening is
differential settlement at the crest. This figure shows that not sufficient to lower the long-term efficiency coefficient
the degree of smoothness of the crest decreases with below that of the short-term case. Furthermore, the
decreasing ARR. Therefore ARR, which depends on both amount of strain-softening is not sufficient to increase the
column spacing and diameter, is an important factor long-term arching ratio compared to the short-term case.
governing the serviceability performance of DCM column- Instead, the long-term arching ratio decreased for the
supported embankments. long-term case compared to the short-term case, indicating
According to Figure 11, ARR significantly influences increased load transfer to DCM columns due to the
the maximum lateral deformation of columns, which oc- increased stiffness of the columns due to consolidation.
curs near the toe of the embankment (at node C). The However, the strain-softened elements increase settlements
maximum lateral deformation decreases with increasing over column heads and clay surface (due to column–clay
ARR. Short-term lateral deformation is reduced by 61%, interaction) significantly during consolidation.
when ARR increases from 4 to 50%. With increasing Results presented in this section investigated the influ-
ARR, bending stiffness of the composite ground increases, ence of ARR calculated based on two approaches: (i) ARR
reducing the lateral deformations for both short- and based on column diameter (varying column diameter) and
long-term conditions. Lateral deformation significantly (ii) ARR based on column spacing (varying column
increases when ARR is lower than 13% due to strain- spacing), on the embankment behaviour. When the ARR
softening of the columns. The lower long-term lateral was greater than 15%, maximum tension showed consider-
deformation in comparison with the short-term lateral able difference in results based on the two approaches.
deformation again indicates the lateral rebound of soft soil However, a significant difference in the embankment
and columns. However, when ARR is less than 13%, the behaviour could not be seen for the two approaches if
difference between geosynthetic tension for short-term and other results were considered. Therefore, ARR alone is not
long-term cases are insignificant because the increase in a suitable parameter to investigate tension developed in the
tension due to extra load transferred to the geosynthetic geosynthetic, when ARR is greater than 15%. Embank-
during strain softening resulting from bond breakage is ment behaviour should be investigated by changing the
higher than the tension release due to lateral rebound of column spacing and diameter separately.
columns and clay.
Figure 12 show the variation of efficiency coefficient of 4.4. Influence of elastic modulus of soft soil
columns and arching ratio, respectively, with ARR. With The influence of soft soil modulus, Es , on the maximum
increasing ARR, the efficiency coefficient of the columns total and differential settlements at the base, total settle-
increases considerably due to transferring more loads to ment at the crest, maximum tension in the geosynthetic
improved ground, which consists of DCM columns. How- layer, efficiency coefficient of columns, arching ratio and
ever, the arching ratio decreases when ARR increases. This lateral deformation of columns are investigated in this
implies that the portion of embankment load transferred section.
from the arching mechanism onto the columns increases Figure 13 illustrates the maximum settlement at the
with increasing ARR. This is also confirmed by Borges base of the embankment over columns and clay with Es :
and Marques (2011) using an axisymmetric finite-element Settlements over columns and clay decrease with increas-
model. Although the tension developed in the geosynthetic ing Es : Beyond a limiting modulus, Es , which is about
decreases, with increasing ARR, the contribution from the 8 MPa, the rate of settlement decrease is insignificant for
vertical component of tension to the vertical load applied both short-term and long-term cases. Figure 14 shows the
on the columns is not sufficient to reverse the increasing variation of maximum differential settlement at the base
overall efficiency coefficient of columns due to arching. of the embankment with Es : With increasing Es , the
The general agreement is that the shear stresses induced maximum differential settlement decreases. The differen-
due to differential settlement is the essential factor for soil tial settlements for both short- and long-term cases shown
arching. Therefore, increasing differential settlement is in Figure 14 are slightly different only when Es is less
expected to increase the degree of arching, transferring than the limiting value and they are the same beyond the
more embankment load to the columns. However, in this limiting Es : Therefore, beyond limiting Es , both soft clay
case, the efficiency coefficient of the columns increases and adjoining columns experience nearly the same vertical
Geosynthetics International, 2014, 21, No. 3

Downloaded by [ UNIVERSIDADE CRUZEIRO DO SUL] on [12/09/16]. Copyright © ICE Publishing, all rights reserved.
222 Yapage and Liyanapathirana
160
Clay surface (long term) acting underneath the geosynthetic reinforcement layer
140
Maximum settlement (mm)
Clay surface (short term) over soft clay increases while reducing the axial strain in
120 Column head (long term)
the geosynthetic. Consequently low tension is developed
Column head (short term)
100
Crest (long term) in the geosynthetic layer at higher values of Es : Further-
80 Crest (short term) more, the influence of Es on lateral spreading of the
60 embankment contributes to the tension developed in the
40 geosynthetic. The tendency for lateral spreading decreases
20
as Es increases, and therefore, the tension in the geosyn-
thetic reinforcement decreases as Es increases. This can be
0
0 2000 4000 6000 8000 10 000 12 000 14 000 further explained using Figure 16, which shows the lateral
Soft soil modulus (kN/m2) deformation of the column at the toe (node C). Increasing
Es decreases the lateral deformation of the columns and
Figure 13. Maximum settlement at the base and the crest the soil in between columns. This indirectly shows that the
with soft soil modulus, Es resistance provided by improved ground to lateral spread-
ing, which takes place in fill layers, increases with
80 increasing Es : Very soft soils with low modulus can cause
Long term
considerable lateral deformation of columns. This may
Differential settlement (mm)

70
Short term
60 result column failure due to bending because bending and
50
tensile strengths of DCM columns are low compared to
the compressive strength (Kitazume and Maruyama 2006).
40
Therefore bending stiffness of individual columns and
30
composite ground should be increased (by increasing
20
column diameter and decreasing column spacing or in-
10 creasing columns modulus) when soft soil has a very low
0 modulus.
0 2000 4000 6000 8000 10 000 12 000 14 000
Design of geosynthetic reinforcement in existing design
Soft soil modulus (kN/m2)
methods is completely based on the membrane force and
Figure 14. Maximum differential settlement with soft soil spreading force. The spreading force is assumed to be
modulus, Es equal to the horizontal force due to active earth pressure
through a section of the crest of the embankment and
ignore the influence of Ec and Es of the improved ground.
strain. Hence, the equal strain concept is valid beyond the According to results presented in this paper, this abandon-
limiting Es : ment of Ec and Es may result in inadequacies in those
Variation of maximum tension in the geosynthetic layer design procedures. Therefore, an improved analytical
with Es is shown in Figure 15. Tension in the geosynthetic solution, which takes into account both Ec and Es , is
reinforcement layer decreased continuously with increas- needed to calculate the spreading forces and thereby
ing Es : This is mainly due to the decrease in differential tension in the geosynthetic. Although soil and columns
settlement with increasing Es : The difference between the laterally displace with loading, during the consolidation
short- and long-term cases was nearly the same because process, due to the dissipation of excess pore water
there was not much difference between short- and long- pressure, they rebound laterally towards the centre of the
term differential settlements. embankment reducing the lateral deformation of soil and
The compressibility of soft soil reduces with increasing columns for the long-term case as shown in Figure 16.
Es and hence differential settlement becomes compara- With increasing Es , the efficiency coefficient of col-
tively low. In other words, the upward reaction force umns decreases while increasing the arching ratio. As

60 250
Maximum lateral deformation (mm)

Long term Long term


Maximum tension (kN/m)

50 Short term Short term


200
40
150
30
100
20

10 50

0 0
0 2000 4000 6000 8000 10 000 12 000 14 000 0 2000 4000 6000 8000 10 000 12 000 14 000
Soft soil modulus (kN/m2) Soft soil modulus (kN/m2)

Figure 15. Maximum tension in the geosynthetic with soft Figure 16. Maximum lateral deformation of the toe with soft
soil modulus, Es soil modulus, Es
Geosynthetics International, 2014, 21, No. 3

Downloaded by [ UNIVERSIDADE CRUZEIRO DO SUL] on [12/09/16]. Copyright © ICE Publishing, all rights reserved.
A parametric study of geosynthetic-reinforced column-supported embankments 223

explained before, the stiffness ratio (Ec /Es ), decreases with Forchheimer’s law, which considers permeability as a
increasing soft soil modulus giving a lower differential function of fluid flow velocity. According to the theory,
settlement and efficiency coefficient, and a higher arching the volumetric flow rate of wetting liquid/unit area of the
ratio. During consolidation, effective stresses in both porous medium is a function of fluid saturation, porosity
columns and clay layer increase. As a result, embankment and volume of the wetting fluid. Therefore, the void ratio
load transferred to columns increases during the consoli- is an input value even though linear elastic-perfectly
dation process reducing the total stress transferred to the plastic soil models are used in the numerical analysis
clay. Hence, the efficiency coefficient of columns for the (Table 1).
long-term case is higher than that for the short-term case. Maximum total settlement over columns (at node A)
At the same time the long-term arching ratio is lower than and clay layer (at node B), and the differential settlement
the short-term ratio (Figure 17). Some of the design at the base are shown in Figures 18 and 19, respectively.
guidelines available in the literature have not considered Maximum settlement over clay decreases with increasing
the contribution of Es on the load transfer mechanism ks for the long-term case, until kc /ks ¼ 1 but when k c =k s is
(e.g. BS8006 (BSI 2010), Hewlett and Randolph (1988), less than one, maximum settlement over clay does not
Swedish method by Carlsson (1987)) and these results change with ks : For the short-term case, maximum settle-
show that further amendments are necessary to incorporate ment over clay does not change with increasing ks until
the contribution of Es on the load transfer mechanism of kc /ks ¼ 1 but when the ratio is less than or equal to one,
column-supported embankments. maximum settlement over clay increases with ks : Conver-
sely, the maximum settlement over columns is nearly
4.5. Influence of soft soil permeability constant for the long-term and short-term cases until
Huang and Han (2009) and Huang et al. (2009) used the kc /ks ¼ 1, but for kc /ks less than one, short-term column
same permeability for the soft soil, ks , and cement mixed settlement increases. Differential settlements at the base
columns, kc : However, it has been experimentally shown shown in Figure 19 decreases with increasing ks until
that kc is higher than that for the same soil at the untreated kc /ks ¼ 1 but when kc /ks is less than one, ks does not have
state (Lorenzo and Bergado 2004; Chew et al. 2004) due a significant influence on both long- and short-term
to the higher void ratio sustained by artificially cemented differential settlements.
soil compared to the same soil in the untreated state. Post-construction settlement over columns and clay layer
Conversely, Baker (2000) and CDIT (2002) pointed out decreases with increasing soft soil permeability despite
that kc is lower than ks : This contradiction concludes that kc /ks , 1 or kc /ks . 1, because the degree of consolidation
kc depends on the properties of the parent soil and hence
there is no universal rule to determine kc : Both kc and ks
are important in determining the post-construction behav- 110
Maximum settlement at base (mm)

iour of the embankment because the generation and


95
dissipation of excess pore water pressures are governed by
kc and ks : In addition, the influence of permeability on the 80
Clay surface (long term)
performance of GRCS embankments is a topic not widely Clay surface (short term)
discussed in the literature. In this paper, the effect of 65 Column head (long term)
permeability on GRCS embankment behaviour is investi- Column head (short term)

gated by varying the ks while keeping the kc as a constant 50


at 1 3 109 m/s. Furthermore, the embankment behaviour
based on excess pore pressure generation also depends on 35
1 ⫻ 10⫺12 1 ⫻ 10⫺10 1 ⫻ 10⫺8 1 ⫻ 10⫺6
the construction time. However, the construction time is a Soil permeability (m/s)
constant throughout the parametric study. Hence, excess
pore pressure generation and dissipation is influenced Figure 18. Maximum settlement at the base with soft soil
only by the permeability. Permeability is defined using permeability, ks

0.8 0.8 50
Efficiency coefficient of columns, f

Long term
Differential settlement (mm)

0.7 0.7 45 Short term


0.6 0.6
Arching ratio, ρ

40
0.5 0.5
35
0.4 0.4
Long term (f ) 30
0.3 0.3
Short term (f )
0.2 Long term (ρ) 0.2 25
Short term (ρ)
0.1 0.1 20
0 2000 4000 6000 8000 10 000 12 000 14 000 1 ⫻ 10⫺12 1 ⫻ 10⫺10 1 ⫻ 10⫺8 1 ⫻ 10⫺6
Soft soil modulus (kN/m2) Soil permeability (m/s)

Figure 17. Efficiency coefficient of columns and arching Figure 19. Maximum differential settlement at the base with
ratio with soft soil modulus, Es soft soil permeability, ks
Geosynthetics International, 2014, 21, No. 3

Downloaded by [ UNIVERSIDADE CRUZEIRO DO SUL] on [12/09/16]. Copyright © ICE Publishing, all rights reserved.
224 Yapage and Liyanapathirana

achieved during the construction period is significant due were 261 mm and 37.5 kN/m, respectively. Therefore, it
to rapid dissipation of excess pore water pressures with can be concluded that ks has more influence on tension in
increasing ks : However, the long-term settlement over the geosynthetic and lateral deformation than kc :
columns does not change with ks : It is found that when the When ks is greater than 1 3 109 m/s, tension and
permeability is extremely low, i.e. 1 3 1012 m/s, soft clay lateral deformation for short- and long-term cases are
and embankment fill yielded considerably compared to the similar because nearly 100% of the degree of consolida-
cases with higher permeability (i.e. ks . 1 3 1012 m/s). tion is achieved at the end of construction.
When permeability is low, excess pore water pressures When ks increases, difference between short-term and
generate significantly due to near undrained nature of the long-term cases for the arching ratio and efficiency coef-
soft soil. This increase in excess pore pressures reduces ficient of columns decreases and finally they overlap with
the shear strength of soft soil causing significant yielding. each other (Figure 22). If a particular ks is considered,
This induces lateral deformations in soft soil. Furthermore, arching ratio decreases whereas the efficiency coefficient
excess pore water pressures exert lateral pressures on increases from the short-term to long-term case. This
columns resulting in higher lateral deformations at the explains the load transfer from foundation soil to columns
column heads as shown in Figure 20. This may lead to during the consolidation process as described before.
significant lateral spreading of the embankment fill. However, when ks increases the arching ratio for the long-
Therefore tension in the geosynthetic layer increases term case increases, whereas the arching ratio for the
significantly, as shown in Figure 21. Consequently, lower short-term case decreases until they reach the same value.
permeability may overstress the geosynthetic layer, yield The opposite effect can be seen for the efficiency coeffi-
the embankment fill and breakdown the columns due to cient of columns. These parameters for both short- and
excessive lateral deformations. When kc is varied by long-term cases are similar and constant when ks >
keeping ks as a constant at 1 3 109 m/s, all the para- 1 3 107 m/s.
meters used to assess embankment behaviour vary in the All these results show unfavourable effects of low
same range, except tension in the geosynthetic and lateral permeability of soft soil on the embankment performance.
deformation. Maximum lateral deformation and tension Therefore it is important to accurately measure the per-
were 132 mm and 30.6 kN/m, respectively, when kc is meability of soft soil and design the embankment taking
1 3 1012 m/s and ks is a constant at 1 3 109 m/s. How- into account the unfavourable characteristics relevant for
ever, when kc is constant at 1 3 109 m/s and ks is soils with low permeability. In addition, the geosynthetic
1 3 1012 m/s, maximum lateral deformation and tension layer with high stiffness or load transfer platform with
several geosynthetic layers would be helpful to reduce
300
excessive lateral deformation of columns or foundation
Maximum lateral deformation (mm)

Long term soil and lateral spreading of embankment fill layer when
Short term
250 ks is low for the soft ground.

200 4.6. Influence of friction angle of the fill material


The influence of friction angle of fill material, 9, on the
150
embankment behaviour is discussed in this part. Basically
granular materials are selected as embankment fill as they
100
are capable of providing adequate support to the embank-
50 ment and traffic load, without causing excessive deforma-
1 ⫻ 10⫺12 1 ⫻ 10⫺10 1 ⫻ 10⫺8 1 ⫻ 10⫺6 tions. Moreover, they do not undergo large volume
Soil permeability (m/s) changes due to variations in moisture content as in
expansive clays. Therefore, embankment fill is modelled
Figure 20. Maximum lateral deformation of the toe with soft as a frictional material without any cohesion in this study.
soil permeability, ks It is believed that the properties of the embankment fill
40 1.0 1.0
Efficiency coefficient of columns, f

Long term (f )
Long term
Short term (f )
Long term (ρ)
Maximum tension (kN/m)

Short term
35 0.8 Short term (ρ) 0.8
Arching ratio, ρ

30 0.6 0.6

25 0.4 0.4

20 0.2 0.2
1 ⫻ 10⫺12 1 ⫻ 10⫺10 1 ⫻ 10⫺8 1 ⫻ 10⫺6 1 ⫻ 10⫺12 1 ⫻ 10⫺10 1 ⫻ 10⫺8 1 ⫻ 10⫺6
Soil permeability (m/s) Soil permeability (m/s)

Figure 21. Maximum tension in the geosynthetic with soft Figure 22. Efficacy coefficient of columns and arching ratio
soil permeability, ks with soft soil permeability, ks
Geosynthetics International, 2014, 21, No. 3

Downloaded by [ UNIVERSIDADE CRUZEIRO DO SUL] on [12/09/16]. Copyright © ICE Publishing, all rights reserved.
A parametric study of geosynthetic-reinforced column-supported embankments 225
45
material highly influence the load transfer to the columns Long term
due to soil arching. In the parametric study the friction 40 Short term

Maximum tension (kN/m)


angle is varied from 25 to 408.
35
As shown in Figure 23, maximum settlement over the
clay surface and crest decreases as 9 increases. Maxi- 30
mum settlement over columns increase marginally with
9. However, 9 of the fill material does not have any 25

influence on the settlement difference between the short- 20


and long-term cases. It is clear from these results that a
higher value of 9 leads to less settlement at the base and 15
24 28 32 36 40
the crest of the embankment. Figure 24 shows the Friction angle of embankment fill material (°)
variation of differential settlement with 9. It clearly
shows that increasing 9 decreases the differential settle- Figure 25. Maximum tension in the geosynthetic with the
ment of the embankment because the frictional resistance friction angle of fill material, ö9
provided by the fill material for differential settlement
increases with increasing 9. These results are consistent 230

Maximum lateral deformation (mm)


Long term
with the influence of 9 on load transfer described at the Short term
200
end of this section.
According to the Figure 25, higher 9 results in lower 170
tension in the geosynthetic layer. The main reason for
140
decreasing tension with increasing 9 is the lower differ-
ential settlement and lateral deformation for higher 9 as 110
shown in Figures 24 and 26, respectively.
80
Lateral deformation basically depends on lateral spread-
ing force which is a function of vertical stress distribution 50
24 28 32 36 40
in the embankment and active earth pressure coefficient, Friction angle of embankment fill material (°)
tan2 (458  9=2): When 9 increases, the active earth
pressure coefficient decreases and hence the lateral Figure 26. Maximum lateral deformation of the toe with the
friction angle of fill material, ö9
Clay surface (long term) Clay surface (short term)
Column head (long term) Column head (short term)
120
Crest (long term) Crest (short term) spreading force decreases. Therefore, lateral deformation
at the toe decreases as shown in Figure 26 and the tensile
Maximum settlement (mm)

100
load required to resist the lateral spreading force decreases
80 as shown in Figure 25.
The efficiency coefficient of columns shown in Figure
60 27 slightly increases with 9. Conversely, the arching ratio
slightly decreases with increasing 9 as shown in Figure
40 27. According to these figures, when 9 increases, the
amount of embankment and traffic load transferred to
20 columns from soil arching increases whereas the load
24 28 32 36 40
Friction angle of embankment fill material (°) transferred to the geosynthetic decreases. The reduction in
vertical load transferred to the geosynthetic with increas-
Figure 23. Maximum settlement at the base and the crest ing 9 also contributes to decrease the tension developed
with the friction angle of fill material, ö9 in the geosynthetic as shown in Figure 25.

55 0.9 0.9
Efficiency coefficient of columns, f

Long term
Differential settlement (mm)

50 Short term
0.7 0.7
Arching ratio, ρ

45

40 0.5 0.5

35
Long term (f )
0.3 0.3
Short term (f )
30
Long term (ρ)
Short term (ρ)
25 0.1 0.1
24 28 32 36 40 24 28 32 36 40
Friction angle of embankment fill material (°) Friction angle of embankment fill material (°)

Figure 24. Maximum differential settlement with the friction Figure 27. Efficiency coefficient of columns and arching
angle of fill material, ö9 ratio with the friction angle of fill material, ö9
Geosynthetics International, 2014, 21, No. 3

Downloaded by [ UNIVERSIDADE CRUZEIRO DO SUL] on [12/09/16]. Copyright © ICE Publishing, all rights reserved.
226 Yapage and Liyanapathirana

The differential settlement decreases with increasing 9, marginally decreases at the clay layer and increases at the
indicating that the degree of arching decreases with column heads, with increasing J. Increasing J enhances
increasing 9, decreasing the loads transferred to the the loads transferred to columns while reducing the loads
columns. However, the arching ratio decreases whereas transferred to geosynthetic layer and foundation soil as
the efficiency coefficient increases with increasing 9, described in details later in this section. Consequently the
indicating an increase in loads transferred to columns with settlement of the soft foundation soil between columns
increasing 9. This shows that the influence of differential decreases while maximum settlement of columns slightly
settlement on the load transfer mechanism is not promi- increases. Figure 28 clearly shows that the geosynthetic
nent in this case. Using centrifuge test data, Kitazume layer is more effective at the crest than at the base in
(1996) and Broms (2003) showed that the increased lateral reducing the total settlement. However, there is no con-
deformation has the ability to reduce the load-carrying tribution from the stiffness on the long-term settlements at
capacity of the improved ground. In this case, increased the base of the embankment, because the post-construction
9 reduced the lateral deformation of columns and clay settlement (difference between short- and long-term cases)
increasing the load-carrying capacity of columns. Hence is nearly a constant.
more embankment load is transferred to DCM columns According to Figure 29, the maximum differential
with increasing 9. settlement decreases with increasing J. Based on these
These results clearly show that 9 has a significant results, it can be concluded that J is more important to
influence on the overall performance of the embankment. reduce the differential settlement than the total settlement
A fill material with a higher 9 is suitable for the at the embankment base. Furthermore, Figure 30 shows
embankment to increase the amount of load transfer to the the total settlement profile along the crest for different
columns, to reduce the lateral deformation at the toe and values of geosynthetic stiffness. The case without a
to increase the degree of smoothness at the crest of the geosynthetic layer is indicated by J ¼ 0 kN/m and no
embankment. Therefore, it is important to consider 9 of significant humps or depressions are visible in all cases,
the embankment fill material in the design of GRCS except for J ¼ 0 kN/m. When tensile stiffness increases,
embankments. the smoothness of the crest increases due to reduction in
differential settlement.
4.7. Influence of tensile stiffness The influence of J on the maximum tension developed
A geosynthetic reinforcement bridging layer is often within the geosynthetic reinforcement is shown in Figure
provided in GRCS embankments to transfer to DCM 31. With increasing J, the maximum tension developed in
columns a portion of the embankment load, which has not
been transferred due to soil arching. Furthermore, this 44
Long term
layer will assist in reducing the lateral spreading beneath
Differential settlement (mm)

40 Short term
the embankment. The stiffness of the geosynthetic layer, J,
is the parameter which governs its efficiency. Therefore, 36
the effect of the stiffness of the geosynthetic layer on the
32
maximum total and differential settlements of the embank-
ment, maximum tension developed in the geosynthetic, 28
lateral deformation of columns and load transfer to soil
24
and columns are investigated by varying J between 1000
and 10 000 kN/m. 20
0 2000 4000 6000 8000 10 000 12 000
As shown in Figure 28, the effect of J on the maximum Tensile stiffness (kN/m)
settlement at the base of the embankment (at the level of
column heads) is not significant. The maximum settlement Figure 29. Maximum differential settlement at the base with
tensile stiffness of geosynthetic, J
Clay surface (long term) Clay surface (short term)
Column head (long term) Column head (short term)
Distance from the edge of the crest to centre of embankment (m)
Crest (long term) Crest (short term)
120 0 2 4 6 8 10 12 14 16
0
Maximum settlement (mm)

100 100
0 kN/m
200 1000 kN/m
Settlement (mm)

80 1500 kN/m
300
2000 kN/m
400 3000 kN/m
60
5000 kN/m
500 10 000 kN/m
40 600
700
20
0 2000 4000 6000 8000 10 000 12 000 800
Tensile stiffness (kN/m)
Figure 30. Settlement profile along the crest of the
Figure 28. Maximum settlement at the base and crest with embankment with different stiffness of geosynthetic for long
tensile stiffness of geosynthetic, J term
Geosynthetics International, 2014, 21, No. 3

Downloaded by [ UNIVERSIDADE CRUZEIRO DO SUL] on [12/09/16]. Copyright © ICE Publishing, all rights reserved.
A parametric study of geosynthetic-reinforced column-supported embankments 227
70
membrane action of the geosynthetic layer and influence
60 of lateral deformation of columns on the overall embank-
Maximum tension (mm)

50 ment behaviour cannot realistically be simulated. There-


40
fore these results confirm that axisymmetric models do
not replicate the load transfer mechanism of GRCS
30 embankments as they do not take into account the behav-
20 iour of the group of columns, membrane action of the
Long term geosynthetic and the overall lateral behaviour of the
10
Short term embankment.
0
0 2000 4000 6000 8000 10 000 12 000
The lateral deformation of the embankment decreases
Tensile stiffness (kN/m) considerably as shown in Figure 33. The lateral deforma-
tion mainly occurs due to lateral spreading of the embank-
Figure 31. Maximum tension with tensile stiffness of ment fill and foundation soil, and lateral earth pressures
geosynthetic, J acting on columns. When J ¼ 10 000 kN/m, the maximum
lateral deformation for the short-term and long-term cases
the geosynthetic increases because developed tension is decrease by about 60% of the lateral deformation at
J, where  is the strain developed within the geosyn- J ¼ 1000 kN/m. Consequently, there is an increase in
thetic. At the same time the arching ratio decreases resistance to lateral spreading of the embankment with
marginally with increasing tensile stiffness of the geosyn- increasing J. As a result lateral deformations decrease with
thetic as shown in Figure 31, indicating that there is a increasing J. Therefore, J is an important factor to be
reduction in vertical loads transferred to the geosynthetic considered to avoid bending failure and subsequent propa-
with increasing J. If the differential settlements are con- gation of plastic hinges within the DCM columns in the
sidered, increasing J decreases the degree of arching and embankment design. Furthermore, in tall embankments, J
it should have increased the arching ratio. However, in this is an important parameter to avoid slip surface failure in
case, the reduction in arching ratio is due to the vertical sloping sides due to excessive lateral spreading.
component of the tension in the geosynthetic over clay.
With increasing J, tension in the geosynthetic increases
and it will reduce the vertical stress in soil above the 4.8. Influence of geosynthetic layer and thickness of
geosynthetic. This is called the membrane action of the platform layer
the geosynthetic. In addition, lateral deformation of the The thickness of the platform layer (the location of the
embankment significantly reduces with increasing J as geosynthetic reinforcement layer) is an important param-
shown in Figure 32. As a result embankment load eter affecting the transfer of the embankment load effi-
transferred to DCM columns increases, reducing the load ciently to the DCM columns minimising the risk of
transfer to geosynthetic. localised punching failure. To investigate the influence of
The overall column efficiency coefficient increases the thickness of the platform layer on the maximum
slightly for both long- and short-term cases as shown in tension developed in the geosynthetic layer, simulations
Figure 32, indicating that there is an increase in the loads were carried out by varying the distance to the geosyn-
transferred to columns with increasing J. This is also due thetic layer from the DCM column heads.
to the vertical component of the geosynthetic tension Figure 34 shows that when the thickness of the platform
acting over the column heads and increase in loads layer increases, the amount of load transferred to the
transfer to the columns due to the decrease in lateral geosynthetic reinforcement decreases. Furthermore, the
deformation of the embankment. However numerical amount of load transferred to the DCM columns de-
modelling based on axisymmetric models showed that the creases. Hence, in order to maximise the loads transferred
arching ratio increases with tensile stiffness (Han and to the columns and to reduce the loads transferred to the
Gabr 2002) because in an axisymmetric model, the soft foundation soil, it is advisable to use a thin platform

0.9 0.9 160


Maximum lateral deformation (mm)
Efficiency coefficient of columns, f

Long term
140 Short term
0.7 0.7 120
Arching ratio, ρ

100
0.5 0.5 80
60
0.3 Long term (f ) 0.3 40
Short term (f )
Long term (ρ) 20
Short term (ρ)
0.1 0.1 0
0 2000 4000 6000 8000 10 000 12 000 0 2000 4000 6000 8000 10 000 12 000
Tensile stiffness (kN/m) Tensile stiffness (kN/m)

Figure 32. Efficiency coefficient of columns and arching Figure 33. Maximum lateral deformation of the toe with
ratio with tensile stiffness of geosynthetic, J tensile stiffness of geosynthetic, J
Geosynthetics International, 2014, 21, No. 3

Downloaded by [ UNIVERSIDADE CRUZEIRO DO SUL] on [12/09/16]. Copyright © ICE Publishing, all rights reserved.
228 Yapage and Liyanapathirana
52
Huang et al. (2009) also investigated the embankment
behaviour without a geosynthetic reinforcement layer but
48
they did not observe similar behaviour due to the low
Tension (kN/m)

height (1.8 m) that was used for the embankment. The


44 results presented in this paper show clearly the importance
of placing a geosynthetic layer with sufficient stiffness for
40 embankments with relatively high fill thickness to resist
the lateral spreading forces. Furthermore, a significant
36
heave could be seen near the toe of the embankment when
0 0.1 0.2 0.3 0.4 0.5 there is no geosynthetic layer but this heave is not clearly
Thickness of the platform layer (m) visible when a geosynthetic layer is incorporated in the
embankment. In addition, the long-term column efficiency
Figure 34. Maximum tension with thickness of the platform
coefficient is lower than that of the short-term case
layer for short term
whereas the long-term arching ratio is higher than that of
the short-term case due to considerable column yielding
layer or keep the geosynthetic layer closer to the DCM and subsequent strain softening occurred as a result of
column heads at the base of the embankment. excessive lateral deformations which occurred towards the
To confirm the effect of geosynthetic layer on the embankment toe.
behaviour of GRCS embankments, another simulation is Considering these results, it can be concluded
conducted without a geosynthetic layer above column that controlling lateral deformation of columns by placing
heads. All results are summarised in Table 3 for the cases a single layer of geosynthetic or by constructing a
with and without the geosynthetic layer and they are geosynthetic-reinforced platform consisting of several
extracted at same locations as described in Section 4.1. layers of geosynthetic over column heads can increase the
When there is a geosynthetic layer over columns, lateral bearing capacity of columns by reducing lateral deforma-
deformation, maximum settlement over clay surface, dif- tions. In addition, when geosynthetic is not incorporated
ferential settlement at the base and arching ratio are in the design, it is important to consider inclined columns
reduced significantly. In contrast, settlement over the or secant wall panels in transverse direction under the side
columns and column efficiency coefficient are increased slope of the embankment to resist lateral spreading forces
due to an increase in load transfer to the columns. These and hence to avoid getting an uneven embankment surface
results clearly show that the presence of a geosynthetic and progressive failure of columns.
layer is important in reducing lateral deformation, and
total and differential settlements in column-supported 4.9. Evaluation of the significance of influencing
embankments. factors on embankment behaviour
Figure 30 also shows that the settlement towards the Huang and Han (2010) proposed a method to rank the
edge is higher than the settlement at the centre of the influencing factors based on their degree of influence on
embankment crest for J ¼ 0 kN/m resulting a significantly the behaviour of the embankment. A similar approach was
deformed embankment surface. When a geosynthetic layer adopted here to rank the influencing factors. The influence
is not included, resistance to lateral spreading forces are of various factors discussed in previous sections are
provided by the improved ground. In general, the lateral ranked based on their influence on maximum tension,
load-carrying capacity of columns and soft clay are not maximum settlement at the base and the crest, lateral
substantial. Hence, large lateral deformations are experi- deformation, arching ratio, column efficiency coefficient
enced by DCM columns towards the embankment toe, and differential settlement at the base. The degree of
resulting in column yielding and subsequent strain soft- influence (DOI) of any influencing parameter on any
ening of columns, increasing the settlements towards the assessing parameter is defined as the ratio of variation of
toe of the embankment. When a geosynthetic layer is assessing parameter to the mean of the same parameter
placed over columns, these lateral spreading forces are (Equation 6). The mean is calculated as described in
resisted by it. Hence the settlement difference between the Equation 6. Importantly, this ranking method is only
centre and edge of the crest are marginal. suitable for the studies in which influencing parameters

Table 3. Comparison of assessing factors when no geosynthetic involved

Parameter Without geosynthetic With geosynthetic

Short term Long term Short term Long term

Lateral deformation (mm) 1233 1232 123 108


Maximum settlement over clay surface (mm) 124.5 150.25 46.8 60.0
Maximum settlement over column heads (mm) 40.0 56.0 43.0 55.05
Differential settlement at base (mm) 96.0 116.25 36.5 37.8
Column efficiency coefficient 0.55 0.47 0.61 0.67
Arching ratio 0.70 0.77 0.43 0.36

Geosynthetics International, 2014, 21, No. 3

Downloaded by [ UNIVERSIDADE CRUZEIRO DO SUL] on [12/09/16]. Copyright © ICE Publishing, all rights reserved.
A parametric study of geosynthetic-reinforced column-supported embankments 229

are varied over a wide range, from its possible lower Maximum settlement at the base is governed by the
bounds to upper bounds. The calculated degree of influ- elastic modulus, spacing and diameter of columns and
ence based on the parametric study are summarised in elastic modulus of soft soil, whereas the maximum settle-
Table 4. ment at the crest is governed by the spacing and diameter
jAPmax  APmin j of columns, elastic modulus of soft soil and friction angle
DOI ¼ 3 100% (6) of the fill material. Only the elastic modulus of the
(APmax þ APmin )=2
columns has a low influence on the lateral deformation of
where APmax is the maximum value of the assessing the columns. Column efficiency coefficient is mainly
parameter and APmin is the minimum value of the governed by the column spacing followed by the column
assessing parameter. diameter and elastic modulus of soft soil. The arching
The degree of influence of any influencing factor on any ratio is mainly governed by column spacing and elastic
assessing parameter can be divided into three levels, i.e. modulus of soft soil followed by column diameter. There-
high, medium and low. However, there is no widely fore, column spacing and diameter and elastic modulus of
accepted standard to classify these levels. Therefore, the soft soil are the most important design parameters to be
significance of degree of influence is classified in Table 5 considered to improve the load transfer to columns. In
based on a commonly used method in geotechnical engi- addition, column spacing for an embankment should be
neering. When the degree of influence is higher than 60%, determined to improve the load transferred from the
in between 30% and 60% and lower than 30%, the influ- arching mechanism. Only permeability has a low influence
ence is considered as high, medium and low, respectively. on the maximum differential settlement and all other
For the tension in the geosynthetic, all influencing influencing parameters have a medium or high influence
factors have a high or medium degree of influence except on the differential settlement. However, maximum differ-
column modulus. Therefore, in the design process of the ential settlement is mainly governed by the column
geosynthetic reinforcement layer or geosynthetic rein- spacing and elastic modulus of the soft soil.
forced platform, it is advisable to consider the influence Overall, column spacing, diameter and elastic modulus
of these parameters. of soft soil have a medium to high influence on all

Table 4. The degree of influence as a percentage

Factors Tmax Smax,base Smax,crest Dmax Efficiency Arching Sdif


coefficient ratio
Column head Clay surface

DCM columns
Elastic modulus, Ec 24 119 68 51 28 29 27 37
Spacing, s 115 83 113 101 87 120 98 146
Diameter, d 95 106 84 92 73 59 45 52
Soft soil
Elastic modulus, Es 158 50 107 139 173 45 74 171
Permeability, ks 41 35 18 30 100 24 12 10
Geosynthetic
Tensile stiffness, J 106 3 16 58 69 10 22 37
Embankment
Friction angle of fill, 9 61 4 17 71 62 13 14 39

Tmax , maximum tension in geosynthetic; Smax,base , maximum settlement at base; Smax,crest , maximum settlement at crest; Dmax , maximum lateral
deformation; Sdif , differential settlement.

Table 5. The significance of degree of influence on the embankment performance

Factors Tmax Smax,base Smax,crest Dmax Efficiency Arching Sdif


coefficient ratio
Column head Clay surface

DCM columns
Elastic modulus, Ec Low High High Medium Low Low Low Medium
Spacing, s High High High High High High High High
Diameter, d High High High High High Medium Medium Medium
Soft soil
Elastic modulus, Es High Medium High High High Medium High High
Permeability, ks Medium Medium Low Low High Low Low Low
Geosynthetic
Tensile stiffness, J High Low Low Medium High Low Low Medium
Embankment
Friction angle of fills, 9 High Low Low High High Low Low Medium

Geosynthetics International, 2014, 21, No. 3

Downloaded by [ UNIVERSIDADE CRUZEIRO DO SUL] on [12/09/16]. Copyright © ICE Publishing, all rights reserved.
230 Yapage and Liyanapathirana

assessing parameters considered in this parametric study. excessive lateral deformations and hence to avoid getting
Accordingly, they can be considered as the most important an uneven embankment surface and progressive failure of
design parameters governing the embankment behaviour. columns due to strain-softening.
If the influence of columns on the embankment behaviour Low permeability of soft soil has unfavourable effects
is considered, column spacing and diameter have more on the overall performance of GRCS embankments. Low
influence on embankment behaviour than column modu- permeability significantly increases the tension in the
lus. Therefore, the use of a higher area replacement ratio geosynthetic, lateral deformation and the tendency for
would be more efficient than the use of stronger columns yielding of columns and embankment fill. Therefore,
to improve the performance of GRCS embankments. when soft soil has a low permeability, a geosynthetic layer
with sufficient tensile stiffness and higher area replace-
ment ratio by increasing column diameter and lowering
5. CONCLUSION column spacing would be advisable to avoid embankment
In this study, two-dimensional coupled mechanical and failure related to excessive lateral deformations.
hydraulic finite-element modelling was used to investigate Column spacing and diameter, which control the area
the behaviour of embankments constructed over deep ce- replacement ratio, and elastic modulus of soft soil are the
ment mixed (DCM) columns reinforced with geosynthetic. most important design parameters to be considered in
The analysis was carried out incorporating time-dependent GRCS design procedures. In addition, use of higher area
behaviour due to excess pore pressure generation and replacement ratio (column layout) is highly recommended
dissipation, strain-softening due to destructuring of the instead of using stronger columns to improve the perform-
soil–cement structure beyond yield and full geometry of ance of GRCS embankments.
the embankment.
According to the results, soil arching ratio decreases
ACKNOWLEDGEMENT
with an increase in column modulus, area replacement
ratio (by increasing column diameter or decreasing col- The authors would like to acknowledge the financial
umn spacing), friction angle of the fill material, and support for this research provided by the Australian
tensile stiffness of geosynthetic and decrease in soft soil Research Council and Coffey Geotechnics Pty Ltd under
modulus. Column efficiency can be increased by increas- the Linkage project LP0990581.
ing column modulus, area replacement ratio, tensile stiff-
ness of geosynthetic and friction angle of embankment fill
NOTATION
material, and by decreasing soft soil modulus. The maxi-
mum settlement at the clay surface and crest of the Basic SI units are given in parentheses.
embankment decrease with increasing column modulus,
area replacement ratio, soft soil modulus, tensile stiffness Acolumn area of a column (m2 )
of the geosynthetic and friction angle of embankment fill Aunit-cell area of a unit cell (m2 )
material. The differential settlement at the base decreases APmax maximum value of assessing parameter
with decreasing column modulus, and increasing area APmin minimum value of assessing parameter
replacement ratio, soft soil modulus, tensile stiffness of c9 effective cohesion of soil (Pa)
the geosynthetic and friction angle of embankment fill c9res residual cohesion (Pa)
material. Lateral deformation at the toe of the embank- c9p peak cohesion (Pa)
ment can be minimised by increasing column modulus, ci interaction coefficient between geosynthetic
soft soil modulus, area replacement ratio, tensile stiffness and platform fill (dimensionless)
of the geosynthetic and friction angle of embankment fill ct tensile strength of soil (Pa)
material. In addition, decreasing lateral deformation in- Dmax maximum lateral deformation (m)
creases the load transfer to columns. d column diameter (m)
In general all assessing parameters show a similar trend DOI degree of influence (dimensionless)
in variation with area replacement ratio calculated based E elastic modulus (Pa)
on column diameter and spacing. However, when the area Ec modulus of DCM columns (Pa)
replacement ratio is greater than 15% there is a significant Es soft soil modulus (Pa)
difference between results for the tension developed in the e void ratio (dimensionless)
geosynthetic. The area replacement ratio calculated based f column efficiency coefficient (dimensionless)
on column diameter decreases sharply beyond 15% due to H embankment height (m)
reduction in geosynthetic span with increasing column I area moment of inertia of columns (m4 )
diameter. Therefore, when the area replacement ratio is J stiffness of the geosynthetic layer (N/m)
greater than 15%, it is important to give consideration for k permeability (m/s)
the approach used to compute ARR. ks permeability of soft soil (m/s)
Based on the results, a thin platform layer (placing a kc permeability of cement mixed columns (m/s)
geosynthetic close to the column heads) is important to p average vertical stress applied on the
effectively transfer the un-arched load to columns. In tall geosynthetic over the soft soil within the unit
embankments, a geosynthetic layer is important to avoid cell (N/m2 )
the reduction of bearing capacity of columns due to Qc total load supported by a column (N)
Geosynthetics International, 2014, 21, No. 3

Downloaded by [ UNIVERSIDADE CRUZEIRO DO SUL] on [12/09/16]. Copyright © ICE Publishing, all rights reserved.
A parametric study of geosynthetic-reinforced column-supported embankments 231

Qt total load from the embankment fill and BSI (British Standard Institution) (2010). BS 8006: Code of Practice for
traffic load applied over a unit cell relevant Strengthened/Reinforced Soil and other Fills. BSI, London, UK.
Broms, B. B. (2003). Deep Soil Stabilization: Design and Construction of
for a column (N)
Lime and Lime/cement Columns. Royal Institute of Technology,
q traffic load (Pa) Stockholm, Sweden.
qu unconfined compressive strength (UCS) Bruce, D. A. (2001). An Introduction to the Deep Mixing Methods as
(N/m2 ) Used in Geotechnical Applications. Volume III: The Verification
Smax,base maximum settlement at base (m) and Properties of Treated Ground. US Department of Transporta-
Smax,crest maximum settlement at crest (m) tion, Washington, DC, USA, FHWA-RD-99–167, Final Report,
455.
Sdif differential settlement (m) Budhu, M. (2000). Soil Mechanics and Foundations. Wiley, New York,
s centre-to-centre spacing of column (m) NY, USA.
T yield strength of geosynthetic (N/m) Carlsson, B. (1987). Reinforced Soil, Principles for Calculation.
Tmax maximum tension in geosynthetic (N/m) Terratema AB, Linköping, Sweden (in Swedish).
t thickness of the geosynthetic (m) CDIT (Coastal Development Institute of Technology) (2002). The Deep
Mixing Method: Principle, Design and Construction. Balkema,
9 effective friction angle of soil (degrees) Rotterdam, the Netherlands.
9p peak friction angle (degrees) Chew, S. H., Kamruzzaman, A. H. M. & Lee, F. H. (2004).
9res residual friction angle (degrees) Physicochemical and engineering behaviour of cement treated clay.
 strain developed within the geosynthetic Journal of Geotechnical and Geoenvironmental Engineering, 130,
No. 7, 696–706.
(dimensionless)
Das, B. M. (2001). Principles of Geotechnical Engineering, 5th edition.
pd,peak deviatoric plastic shear strain at which the Brooks/Cole, PWS Engineering, Boston, MA.
strength reduction starts (dimensionless) EPRI (1990). Manual on Estimating Soil Properties for Foundation
pd,res deviatoric plastic shear strain at which the Design. Geotechnical Engineering Group, Cornell University,
softening is complete (dimensionless) Ithaca, NY, USA.
9 effective contact pressure (N/m2 ) EuroSoilStab (2002). Development of Design and Construction Methods to
Stabilize Soft Organic Soils: Design Guide Soft Soil Stabilization.
 column average vertical stress applied on a column IHS BRE Press, Watford, UK, CT97–0351, Project No: BE 96–3177.
(N/m2 ) Filz, G. M. (2007). Load Transfer, Settlement, and Stability of
ª unit weight of soil (N/m3 ) Embankments Founded on Columns Installed by Deep Mixing
ªfill unit weight of the fill material (N/m3 ) Methods, National Technical University of Athens School of Civil
r soil arching ratio (dimensionless) Engineering Geotechnical Department – Foundation Engineering
Laboratory, Athens, Greece.
 mobilised shearing resistance (N/m2 ) Forsman, J., Honkala, A. & Smura, M. (1999). Hertsby case: a column
 friction angle (degrees) stabilised and geotextile reinforced road embankment on soft
 Poisson’s ratio (dimensionless) subsoil. Dry Mix Method for Deep Soil Stabilization, Brandenburg,
H., Holm, G. and Broms, B. B., Editors, Balkema, Rotterdam, the
Netherlands, pp. 263–368.
ABBREVIATIONS GeoTesting Express (1996). Geotechnical Tests on Soil–Cement Mix,
GeoTesting Express, Atlanta, GA, USA, Central Artery/Tunnel
ARR area replacement ratio Project –C09A7 Final Report, a report prepared for Bechtel/Parsons
Brinkerhoff.
DCM deep cement mixed Han, J. & Gabr, M. A. (2002). A numerical study of load transfer
GRSC geosynthetic-reinforced column-supported mechanisms in geosynthetic reinforced and pile supported embank-
ments over soft soil. Journal of Geotechnical and Geoenvironmental
Engineering, 128, No. 1, 44–53.
Hewlett, W. J. & Randolph, M. F. (1988). Analysis of piled embankments.
REFERENCES Ground Engineering, 21, No. 3, 12–18.
ABAQUS (2010). ABAQUS Version 6.10 User’s Manual. ABAQUS, Inc, Huang, J. & Han, J. (2009). 3D coupled mechanical and hydraulic
Providence, RI, USA. modelling of a geosynthetic-reinforced deep mixed column-
Ariyarathne, P., Liyanapathirana, D. & Leo, C. (2013a). A comparison of supported embankment. Geotextiles and Geomembranes, 27, No. 4,
different two dimensional idealizations for a geosynthetic reinforced 272–280.
pile-supported embankment. International Journal of Geomechanics Huang, J. & Han, J. (2010). Two-dimensional parametric study of
13, No. 6, 754–768. geosynthetic-reinforced columns-supported embankments by
Ariyarathne, P., Liyanapathirana, D. S. & Leo, C. J. (2013b). Effect of coupled hydraulic and mechanical modelling. Computers and
geosynthetic creep on reinforced pile-supported embankment Geotechnics, 37, No. 5, 638–648.
systems. Geosynthetics International, 20, No. 6, 421–435. Huang, J., Han, J. & Oztoprak, S. (2009). Coupled mechanical and
Baker, S. (2000). Deformation Behaviour of Lime/Cement Column hydraulic modelling of geosynthetic-reinforced column-supported
Stabilized Clay. Swedish Deep Stabilization Research Centre, embankments. Journal of Geotechnical and Geoenvironmental
Rapport 7, Chalmers University of Technology, Gothenburg, Sweden. Engineering, 135, No. 8, 1011–1021.
Bergado, D. T. & Long, P. V. (1994). Numerical analysis of embankment Jacobson, J. R., Filz, G. M. & Mitchell, J. K. (2003). Factors Affecting
on subsiding ground improved by vertical drains and granular piles. Strength Gain in Lime-cement Columns and Development of a
In Proceedings of the International Conference on Soil Mechanics Laboratory Testing Procedure, Virginia Polytechnic Institute and
and Foundation Engineering –International Society for Soil State University, Blacksburg VA, Report prepared for the Virginia
Mechanics and Foundation Engineering, Morgantown, WV, USA, Transportation Research Council.
Balkema, Rotterdam, the Netherlands, vol. 4, pp. 1361–1361. Jacobson, J. R., Filz, G. M. & Mitchell, J. K. (2005). Factors affecting
Borges, J. L. & Marques, D. O. (2011). Geosynthetic-reinforced and jet strength of lime-cement columns based on a laboratory study of
grout column-supported embankments on soft soils: numerical three organic soils. In Deep Mixing ‘05: International Conference
analysis and parametric study. Computers and Geotechnics, 38, No. on Deep Mixing Best Practice and Recent Advances, Stockholm,
7, 883–896. Sweden.

Geosynthetics International, 2014, 21, No. 3

Downloaded by [ UNIVERSIDADE CRUZEIRO DO SUL] on [12/09/16]. Copyright © ICE Publishing, all rights reserved.
232 Yapage and Liyanapathirana

Japanese Geotechnical Society (2000). Standard Practice for Making and and Chai, J.-C., Editors, Geotechnical Special Publication no. 152,
Curing Stabilized Soil Specimens without Compaction (JGS 0821– ASCE, Reston, VA, USA.
2000). Geotechnical Test Procedure and Commentary. Japanese Stewart, M. E., Navin, M. P. & Filz, G. M. (2004). Analysis of a column-
Geotechnical Society, Tokyo, Japan. supported test embankment at the I-95/route 1 interchange.
Kempfert, H. G., Gobel, C., Alexiew, D. & Heitz, C. (2004). German Geotechnical Engineering for Transportation Projects, Yegian, M. K.
recommendations for reinforced embankments on pile-similar and Kavazanjian, E., Editors, Geotechnical Special Publication no.
elements. In EuroGeo3-Third European Geosynthetics Conference, 126, ASCE, Reston, VA, USA, pp. 1337–1346.
Geotechnical Engineering with Geosynthetics, Deutsche Gesell Yapage, N. N. S. (2013). Numerical Modelling of Geosynthetic
schaft fur Geotechnik, Munich, Germany, pp. 279–284. Reinforced Embankments over Soft Ground improved with Deep
Kitazume, M. (1996). Bearing capacity of improved ground with column Cement Mixed Columns. PhD thesis, University of Western Sydney,
type DMM. Grouting and Deep Mixing, Balkema, Rotterdam, the Penrith, Australia.
Netherlands, pp. 503–508. Yapage, N. N. S., Liyanapathirana, D. S., Poulos, H. G., Kelly, R. B. &
Kitazume, M. & Maruyama, K. (2006). External stability of group Leo, C. J. (2012). An investigation of progressive failure of
column type deep mixing improved ground under embankment. geosynthetic reinforced deep cement mixed column supported
Soils and Foundation, 46, No. 3, 323–40. embankments. International Conference on Ground Improvement
Lorenzo, G. A. & Bergado, D. T. (2004). Fundamental parameters of and Ground Control, Wollongong, Australia, vol. 2, pp. 1345–1351.
cement-admixed clay-new approach. Journal of Geotechnical and Yapage, N. N. S., Liyanapathirana, D. S., Poulos, H. G., Kelly, R. B. &
Geoenvironmental Engineering, 130, No. 10, 1042–1050. Leo, C. J. (2013). Towards the development of a new design
Madhyannapu, R. S., Puppala, A. J., Hossain, S., Han, J. & Porbaha, A. guideline for geosynthetic reinforced column supported embank-
(2006). Analysis of geotextile reinforced embankment over deep ments. Australian Geomechanics Journal, 48, No. 3, 35–50.
mixed soil columns: using numerical and analytical tools. In Yapage, N. N. S., Liyanapathirana, D. S., Poulos, H. G., Kelly, R. B. &
Proceedings of the ASCE GeoCongress 2006, Atlanta, GA, USA. Leo, C. J. (2014a). Numerical modelling of geotextile reinforced
Navin, M. P. (2005). Stability of Embankments Founded on Soft Soil embankment over deep cement mixed columns incorporating strain
improved with Deep Mixing Method Columns. PhD thesis, Virginia softening behaviour. International Journal of Geomechanics, in
Polytechnic Institute and State University, Blacksburg, VA, USA. press, http://dx.doi.org/10.1061/(ASCE)GM.1943–5622.0000341.
Navin, M. P. & Filz, G. M. (2005). Statistical analysis of strength data Yapage, N. N. S., Liyanapathirana, D. S., Poulos, H. G., Kelly, R. B. &
from ground improved with DMM columns. In Deep Mixing ‘05: Leo, C. J. (2014b). Modelling of an embankment over soft ground
International Conference on Deep Mixing Best Practice and Recent improved with deep cement mixed columns: a case history. Journal
Advances, Stockholm, Sweden, pp. 145–154. of Geotechnical and Geoenvironmental Engineering, in press.
Navin, M. P. & Filz, G. M. (2006). Numerical stability analysis of Yu, Y. & Gao, W. (2002). Field test study on DCM pile in soft foundation
embankments supported on deep mixed columns. Ground Modifica- improvement at bridge approaches. Chinese Civil Engineering
tion and Seismic Mitigation, Porbaha, A., Shen, S.-L., Wartman, J. Journal, 35, No. 4, 65–69.

The Editor welcomes discussion on all papers published in Geosynthetics International. Please email your contribution to
discussion@geosynthetics-international.com by 15 December 2014.

Geosynthetics International, 2014, 21, No. 3

Downloaded by [ UNIVERSIDADE CRUZEIRO DO SUL] on [12/09/16]. Copyright © ICE Publishing, all rights reserved.

You might also like