Computers and Geotechnics 72 (2016) 42–56

Contents lists available at ScienceDirect

Computers and Geotechnics

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

Research Paper

Three-dimensional numerical analysis of a DCM column-supported

highway embankment
Pitthaya Jamsawang a,⇑, Naphol Yoobanpot a, Nuttawut Thanasisathit a, Panich Voottipruex b,
Pornkasem Jongpradist c
a
Department of Civil Engineering, King Mongkut’s University of Technology North Bangkok, Thailand
b
Department of Teacher Training in Civil Engineering, King Mongkut’s University of Technology North Bangkok, Thailand
c
Department of Civil Engineering, Faculty of Engineering, King Mongkut’s University of Technology Thonburi, Bangkok, Thailand

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents a three-dimensional numerical analysis of a well-monitored DCM (deep cement
Received 11 August 2015 mixing) column-supported embankment for a highway in Thailand. A coupled three-dimensional
Received in revised form 9 November 2015 mechanical and hydraulic numerical model was used for this analysis, and the results are compared with
Accepted 11 November 2015
the field measurements, including settlement, load distribution between soil and DCM columns, excess
Available online 7 December 2015
pore water pressure and lateral movement. The numerical model reasonably simulated the performance
of the DCM column-supported embankment compared to the observed data. From the field measure-
Keywords:
ments, the surrounding soils carried approximately half an embankment load, while the rest was trans-
Deep mixing
Embankment
ferred to the stronger soil layers below. The computed maximum bending moment in the column under
Parametric study the toe of the embankment was located at the interlayer between the fill material and soft clay due to the
Simulation existence of thick fill material layer. A parametric study was conducted to investigate the influence of five
Soil stabilization key factors on the performance of the DCM column-supported embankment. The parametric study indi-
cated that the factor of safety against bending moment failure (FSbending) seems to be the most important
factor of safety to design in this study. The embedded column was more suitable than a fixed column to
support high embankment load due to the higher moment capacity of the column with the same
improvement area ratio ar, while the reduction ratios of settlement (Rs) and lateral movement (Rl) were
not significantly different. If the floating column is chosen, sufficient improvement depth to produce the
column tip laterally restrained is necessary to not only reduce the lateral movement but also increase the
FSbending. The dr of greater than 0.7 is suggested in this study.
Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction of China [49,59], the eastern coast of Singapore [3,6] and the Bang-
kok Plain [37,18,19]. The majority of these soft deposits were laid
Geotechnical engineering relates to civil engineering projects, down during the Quaternary period [22,48,23]. The Quaternary
such as the design of a bridge, construction of an embankment deposits generally have high water content with high compress-
dam, or prevention of slope instability in a geological structure ibility, high sensitivity, high viscosity, and low shear strength
[25]. Geotechnical engineers must investigate the geological condi- [61]. The construction of highway embankments overlying these
tions to identify the local soil conditions at a site and determine soils often leads to large lateral pressures and movement, excessive
whether they are problematic. Further, they must determine a settlements and slope and bearing failures, which usually result in
solution with which structures can be designed, constructed and long construction delays and costly remedial work [1,25].
operated safely and economically. Soft clays are geological materi- Civil and/or geotechnical engineers have faced major challenges
als that represent a well-known category of problematic soils and in designing highway embankments on this soft clay. To mitigate
are generally encountered in the form of deposited layers in coastal these potential problems, deep cement mixing (DCM) columns, a
areas [58]. Soft clay deposits are universally distributed along ground improvement technique, have been introduced [9,14,15].
coastal regions such as Ariake Bay, Japan [39], the eastern coast DCM column-supported embankments, as a rapid construction
technique, have been increasingly adopted to raise the bearing
⇑ Corresponding author. capacity and reduce the total and differential settlements [29].
E-mail address: [email protected] (P. Jamsawang). The case histories reported that DCM columns have been widely

http://dx.doi.org/10.1016/j.compgeo.2015.11.006
0266-352X/Ó 2015 Elsevier Ltd. All rights reserved.
 P. Jamsawang et al. / Computers and Geotechnics 72 (2016) 42–56 43

used in many application across the world, such as railway further investigating the behaviors in more detail have also been
embankments in Malaysia [4]; road and railway embankments conducted by two-dimensional (2D) finite element analysis (FEA)
[33,45] and metropolitan highways in Japan [41]; road embank- – e.g., Huang and Han [21], Borges and Marques [7], Oliveira
ments in Sweden [53]; urban carriage ways in Poland [53]; and et al. [43], Horpibulsuk et al. [19], Zhang et al. [62] and Jiang
flood levees and floodwalls in the USA [13]. et al. [29]. The most studied parameters include the spacing and
The DCM columns under the embankment are usually installed stiffness of the columns, soil stiffness and permeability because
through soft soil layers to transfer the embankment and traffic they significantly affect the amount of settlement and the consol-
loads to deep and firm strata [16,42,35,19]. Square or triangular idation rate. However, a comparison between 2D-FEA [2] and 3D-
grid patterns of a single column are usually applied for settlement FEA [64] on the same problem of slope stability of embankments
reduction purposes with improvement area ratios (ar) of 0.1–0.5 over column-improved soft clay reveals that the 2D numerical
[52,54]. The improvement depth is generally classified into two analysis slightly overestimated the factor of safety compared to
improvement types, floating and fixed columns (as schematically its 3D counterpart. Chai et al. [12] concluded that 2D-FEA predicts
shown in Fig. 1a and b, respectively), which depends on whether incorrect results in terms of the lateral movement and bending
the tips of DCM columns are placed on the stronger clay layer. moment in the columns under the toe of the embankment com-
Fixed columns are normally used to increase stability and reduce pared to 3D-FEA. The most recent FEA analyses of DCM-
settlement and lateral movement. However, floating columns are supported embankments were under the 3D condition – e.g.,
selected when the thickness of the soft clay is quite high [30]. Huang and Han [20], Voottipruex et al. [56] and Yapage et al. [60].
A number of field studies have been reported on the use of However, limited studies have been conducted on the compar-
column-supported embankments, such as Lai et al. [32], Arulrajah ison results of improved and unimproved ground (no DCM column
et al. [4], Hello and Villard [17], Liu et al. [35], Nunez et al. [40], support) even though such studies are needed to determine the
Yapage et al. [60], Zhang et al. [63], Liu et al. [36], Lu and Miao degree of improvement or degree of settlement reduction. More-
[38] and Rowe and Liu [46]. These field studies verified the load over, the types of improvement (Fig. 1), which are important
transfer mechanisms, the development of differential settlements parameters to achieve the required degree of settlement reduction,
between columns, and the consolidation behavior of column- have not been considered in previous studies. In addition, the ‘‘per-
supported embankments. Additional numerical studies aiming at formances” in previous research contain deformation, pore water
pressure and load transfer, whereas the columns located under
the toe of an embankment can fail when the maximum bending
moment in the column reaches the moment capacity of the column
[12]. Thus, the maximum bending moment in the column should
be a parameter considered in the design.
This paper focuses on the numerical analysis of a field case
study on an instrumented DCM column-supported embankment
along the Bangbo-Klongdan highways. In the field study, axial
compression and lateral load tests in individual column tests were
performed, and settlements, excess pore water pressure, and earth
pressure during the embankment construction were monitored
over time for approximately 235 days. 3D-FEA incorporated in
the commercial software PLAXIS 3D Version 2013 was adopted
for the numerical analysis. The numerical analysis simulated the
consolidation behavior by a coupled mechanical and hydraulic
model. In addition, a parametric study was conducted to investi-
gate five key influence factors on the performance of the DCM
column-supported embankment including maximum bending in
the column. The influence factors included no DCM column sup-
port; floating, fixed and embedded columns (Fig. 1c); and embank-
ment height.

2. Site conditions and soil properties

The highway embankment in this study was located at Station

KM 6+055 of highway No. 3117 (Bangbo-Klongdan) within the
Bangbo district of Samutprakan province, Thailand. The project site
mainly consisted of marine deposits. The soil profile at this station
was a 3-m-thick fill material underlain by an 11-m-thick soft clay,
a 9-m-thick medium stiff clay, and a 5-m-thick stiff clay soft clay.
Fig. 2a–h present the soil properties of the studied area. Water con-
tents, liquid limits and plastic limits are the average profiles gener-
ated based on the laboratory data of soil samples from the field.
The coefficients of permeability (k) of the soil were calculated
based on the consolidation test data. The compression indices,
recompression indices and over-consolidation ratio (OCR) values
were determined by odometer tests. The OCR profile adopted for
the numerical study was estimated by the average of the OCR data
available for the same layer of soil. The undrained shear strength of
Fig. 1. Types of improvement depths: (a) floating (b) fixed and (c) embedded. soils (su) was obtained from unconfined compression tests on the
44 P. Jamsawang et al. / Computers and Geotechnics 72 (2016) 42–56

Fig. 2. Soil profiles and properties.

collected specimens. The groundwater table was near the ground as shown in Fig. 4 corresponding to the test results by Jongpradist
surface. Fig. 2h shows that the undrained shear strengths of the et al. [28] and Jamsawang et al. [24–27].
soils increased almost linearly with depth. According to the Unified
Soil Classification System (USCS), the soft clay in this project site
4. Instrumentations
was classified as high plasticity clay (CH).

To observe the performance of a DCM column-supported high-

3. A DCM column-supported embankment way embankment, various types of geotechnical instruments were
installed at the project site. The geotechnical instruments used in
Fig. 3 shows the cross-sectional and plan views of the DCM this project were similar to those used in other projects related
column-supported highway embankment, which is 1.5-m high to embankment loading on soft ground as reported by Arulrajah
with a 12-m wide crest and 1(V): 2(H) side slopes. The elastic et al. [5], Liu et al. [35] and Yapage et al. [60]. These include two
modulus and shear strength of the embankment fill were deter- vibrating-wire piezometers to measure the piezometric head
mined from laboratory tests. The DCM columns under the highway placed at depths of
5 and
15 m along the centerline of the
embankment were installed in a square pattern at a spacing of embankment; two total earth pressure cells to measure the total
1.5 m using a mechanical mixing method. The water–cement ratio stress transferred to the DCM column (TP1) and the surrounding
of the slurry and the cement content were 1.5 and 150 kg/m3 of soil between columns (TP2); four rod extensometers to measure
soil, respectively. An additional tangential DCM column was con- the settlements below the embankment at different depths of
structed at the toe of the embankment to increase the rigidity,
5,
10,
15 and 20 m; and a vibrating wire settlement cell to
resisting lateral movement of the subsoil foundation. All DCM col- measure settlements on the surrounding soil installed at a depth
umns had a diameter of 0.6 m and a depth of
14.00 m, where the of 0.4 m from the original ground surface at the center of the
column tip was placed on the medium stiff clay layer, which is embankment. Fig. 3a and b shows the instrumentation in cross-
stronger than the upper soft clay layer. This improvement depth section and plan views, respectively. In addition to the abovemen-
is the so-called ‘‘fixed type” improvement. Core samples were ran- tioned instruments installed under the embankment, an incli-
domly extracted from the DCM columns at various depths over the nometer casing was installed close to the toe of the embankment
construction area for use in unconfined compression tests in the on the right to a depth of
26.5 m to measure the lateral move-
laboratory to determine the unconfined compressive strength, qu, ment induced by the subsoil foundation and to compute the lateral
and the secant modulus of elasticity corresponding to 50% of the movement–settlement ratio of the embankment, as shown in
unconfined compressive strength, EDCM. The values of the uncon- Fig. 3a. Moreover, a deep reference point is installed in a stiff clay
fined compressive strength ranged from 0.2 to 1.6 MPa, with an layer at locations on the site far from other permanent survey
average value of 0.94 MPa, whereas the modulus of elasticity benchmarks as the survey datum reference point. All instruments
ranged from 10 to 170 MPa, with an average value of 92 MPa, indi- were monitored daily during the construction of the embankment
cating an empirical relationship of EDCM = 101qu, with R2 = 0.8698 and weekly after the filling of the last layer was completed by
 P. Jamsawang et al. / Computers and Geotechnics 72 (2016) 42–56 45

Fig. 3. Cross-sectional and plan views of the DCM column-supported embankment and locations of instrumentations: (a) cross-sectional view and (b) plan view of the
instrumented embankment.

can be computed from the piezometric head. All settlement values

were calibrated with the surveyed data of deep reference points.

5. Full-scale field loading tests

The embankment weight and traffic load are transferred to the

DCM columns and surrounding soils, and they tend to concentrate
on the DCM columns rather than the surrounding soils. The ulti-
mate bearing capacity of a single DCM column (Qult) should be
examined. An axial load test on a DCM column was conducted by
the application of a load until continuous settlement occurred at
a slight or no increase in axial load. The improvement depth and
diameter of the test DCM column were the same as those of the
DCM columns used for embankment support. Fig. 5 shows the axial
compression load versus settlement relationships for the test col-
umn. Qult was determined by the slope tangent method at the point
of intersection of the initial and final tangents to the load settle-
ment curve following Butler and Hoy [8]. Qult was found to be
300 kN. The result of a previous test on a 0.6-m-diameter and
7 m long fixed end type DCM column in soft Bangkok clay [24] is
Fig. 4. Secant modulus versus unconfined compressive strength from cored DCM also illustrated in the figure. The surface layer of soil in the study
samples.
was a 1-m-thick weathered crust followed by a 6-m-thick soft
and medium clay layer. By the same criteria, Qult from the previous
reading with a data logger. The recorded values were adjusted by work by Jamsawang et al. [24] was 200 kN, which is lower than
the calibration factors of each instrument. Together with the static that of DCM for this study by a factor of approximately 1.5. More-
water level, the excess pore pressure due to embankment loading over, the settlements were larger than those at the same load. This
46 P. Jamsawang et al. / Computers and Geotechnics 72 (2016) 42–56

Fig. 5. Axial compression load–settlement curves of DCM columns from field tests.

implies that the improvement depth affects both Qult and the
settlement of a fixed column.
The maximum bending moment (Mmax) induced in a single
DCM column must be known to prevent local failure due to
exceeding the moment capacity of the column section (Myield)
[12]. Generally, DCM columns have higher compressive strength
than tensile strength, and the tensile strength is typically approx-
imately 0.10–0.16 times qu [24]. Therefore, the columns located
under the toe of an embankment are displaced laterally by the lat- Fig. 6. (a) Lateral load–lateral displacement curves of DCM columns from field
eral earth pressure in the embankment. Those columns can fail tests. (b) Moment capacity versus DCM column diameter.
when Mmax reaches Myield. Thus, the lateral load tests on two
DCM columns were also performed to measure ultimate lateral
load (Pult) and Myield. The loads were applied in increments of
0.4 kN until continuous lateral displacements occurred at a slight
or no increase in load. Pult of the DCM columns were 16 kN and
12 kN, with an average Pult of 14 kN. Excavation after the test
revealed cracks approximately 0.10 m below the base of the exca-
vated pit, which is the location of the plastic hinge referring to
Mmax in the DCM columns. This means that the failure mode of
two DCM columns was column material failure due to Mmax
induced by the lateral load reaching Myield, as calculated from
Fig. 6a. The lateral load–lateral displacements of a similar study
by Petchgate et al. [44] are plotted together in Fig. 6a for compar-
ison. The column diameter (DDCM) and improvement depth were
0.5 m and 15 m, respectively. Pult of the test DCM column was
8 kN, and the crack location found by Petchgate et al. [44] was
the same as that in this study. The column size affects the flexural
stiffness and ultimate load as well as Myield. The flexural strength of
the DCM column (rf) can be calculated from the following relation-
ship [44]:

M yield ¼ Pult ycr ð1Þ

M yield ðDDCM =2Þ

rf ¼ ð2Þ
IDCM
Fig. 7. (a) Plan view of the analysis zone. (b) 3D finite element mesh.
where IDCM = the moment of inertia of the column section and
ycr = the distance of the moment arm from the load application to
the crack location. The average Myield calculated from Eq. (1) was et al. [24]. Based on Eq. (2) and the field test results, Myield depends
2.8 kN-m, which corresponds to a rf of 130 kPa (Eq. (2)) or is equiv- on the column diameter, whereas rf is a property of the soil–cement
alent to 0.15 times the average qu (900 kPa), which is close to the material. Assuming that all DCM columns in this study comprise a
test results from the beam test in the laboratory of Jamsawang homogeneous and isotropic material, the Myield for various column
 P. Jamsawang et al. / Computers and Geotechnics 72 (2016) 42–56 47

sizes can be determined by Eq. (2) and are shown in Fig. 6b using boundary, the x = 60 plane. In addition, symmetrical conditions
the same rf value of 130 kPa, and they are used for comparison with imply zero displacement in the y direction for nodes on the two
the computed Mmax in the DCM column as obtained from 3D-FEM. vertical planes, the y = 0 and y = 1.5 planes. With regard to the drai-
nage boundary conditions, the water table is assumed to be at a
6. Numerical modeling depth of 1.5 m below ground level, and the initial pore pressures
prior to the embankment construction are taken to be hydrostatic.
6.1. Model mesh and boundary condition Water can freely drain out at the ground surface. The bottom of the
finite-element mesh is defined as impermeable owing to the exis-
The DCM column-supported embankment shown in Fig. 3a is tence of low-permeability soil, and the lateral flow is not permitted
essentially three-dimensional because each column is not continu- across the x = 0, x = 60, y = 0, and y = 1.5 planes because of the sym-
ous in the out-of-plane direction. However, by taking the planes of metry. The construction stages are followed in the finite-element
symmetry, it is possible to analyze a three-dimensional slice in iso- calculation as tabulated in Table 1.
lation. In this study, three-dimensional (3D) FEA was conducted to
simulate the behavior of the DCM column-supported embankment 6.2. Constitutive model and model parameters
using the commercial FE software PLAXIS 3D version 2013. The 3D
FE model consisted of the DCM columns, highway embankment The DCM columns, fill material and embankment were modeled
and foundation soils. There were no interface elements among as linearly elastic to perfectly plastic materials with the Mohr–
the periphery, the DCM columns and the surrounding foundation Coulomb model (MCM) [20,55–57,25–27]. The tensile strength of
soils because the shear strength at the interface between the the DCM columns, considering the use of the tension cutoff in
DCM columns and the surrounding clay was higher than that of the model, was 0.15qu. The soft clay layer and medium stiff clay
the surrounding soft clay foundation [54,55]. Owing to the symme- layer were modeled by the soft soil model (SSM). The SSM is a
try of this embankment, half of the embankment was modeled as Cam-clay-type model. In addition to the strength parameters, the
shown in Fig. 8 for the finite element mesh, which corresponds modified compression index (k⁄) and modified swelling index
to the DCM column-supported embankment configuration in (j⁄) were determined based on the one-dimensional consolidation
Fig. 3. To ensure that the boundary effect will be minimized, the test using well-recognized equations k⁄ = Cc/2.3(1 + e0) and
finite-element mesh is extended to a depth of 28 m (down to a j⁄ = 2Cr/2.3(1 + e0). The coefficients of permeability of the DCM
rigid and impermeable layer underneath) and a horizontal length columns were assumed to be the same as those of the surrounding
of 60 m. soils according to the experimental data by Horpibulsuk et al. [18].
At the bottom of the finite-element mesh z = 0 plane, the dis- The hardening soil model (HSM) is an advanced model for sim-
placements are set to zero in the three directions x, y, and z. The ulating the behavior of different types of stiff soils [50,47,25–27,57].
displacements in the x direction are set to zero on the center line In addition to the plasticity concept, the model is equipped with
of the embankment, the x = 0 plane, and the far-field lateral many key features regarding the soil responses, including double
yield surfaces with hardening functions, stress-dependent stiffness,
and loading/unloading. The HSM utilizes four basic deformation
parameters: the secant stiffness in standard drained triaxial tests
(Eref ref
50 ), the tangential stiffness for primary odometer loading (Eoed ),

the unloading and reloading stiffness (Eref ur ; v ur ) and the power of

the stress-level dependency of the stiffness (m) and shear strength
parameters according to the MCM (c0 ; /0 and w). The HSM was used
to model the behavior of the stiff clay in this study. The permeabil-
ity of all soils is assumed to be equal all directions. The FEM input
parameters of all materials for MCM, SSM and HSM are shown in
Tables 2–4, respectively.

Table 1
Construction stages in the finite-element calculation.

Stage Description
0 Generation of the initial stresses (Ko – condition)
1 Installation of the DCM columns
2 Construction of a 0.5-m-high embankment
3 Construction of a 0.8-m-high embankment
4 Construction of a 1.2-m-high embankment
5 Construction of a 1.5-m-high embankment

Table 2
Parameters used in MCM.

Embankment Fill material DCM column

Unit weight, c (kN/m3) 20 20 15
Elastic modulus, E0 (MPa) 20 30 80
Poisson’s ratio, m0 0.33 0.33 0.33
Cohesion, c (kPa) 1 1 cu ¼ 450
Friction angle, / (degree) 30 32 0
Permeability coefficient, – – 5  10
4
k (m/d)
Fig. 8. (a) Embankment load history. (b) Observed and computed settlements Material behavior Drained Drained Undrained type B
versus time.
48 P. Jamsawang et al. / Computers and Geotechnics 72 (2016) 42–56

Table 3
Parameters used in SSM.

Soft clay Medium stiff clay

Unit weight, c (kN/m )3
14 16
Modified compression index, k 0.18 0.12
Modified swelling index, j 0.04 0.06
Poisson’s ratio, m 0.35 0.15
Cohesion, c (kPa) 1 10
Friction angle, / (degree) 23 25
Over consolidation ratio (OCR) 1.5 2.0
Permeability coefficient, k (m/d) 5  10
4 2.5  10
4
Material behavior Undrained Undrained

Table 4
Parameters used in HSM.

Stiff clay
Unit weight, c (kN/m3) 20
Secant stiffness, Eref 50
50 (MPa)
Tangential stiffness, Eref (MPa) 50
oed
Unloading and reloading stiffness, Eref 150
ur (MPa)
Power of the stress-level dependency of the stiffness, m 1
Poisson’s ratio, m 0.2
Cohesion, c (kPa) 18
Friction angle, / (degree) 25
Over consolidation ratio (OCR) 2.5
Permeability coefficient, k (m/d) 2.5  10
4
Material behavior Undrained

Fig. 9. Observed and computed settlements with depth.

7. Numerical results and comparisons

7.1. Settlement of the surface settlement, which implies that approximately one-
third of the total settlement was induced in soft clay and the rest
Fig. 8a shows the history of the embankment construction for in stronger soil below. The computed results show that the settle-
up to 235 days, whereas Fig. 8b presents the corresponding ment percentages induced in soft clay, medium stiff clay and stiff
measured and computed settlements over the elapsed time at clay were 49%, 48% and 3% of the computed total settlement,
the base of the embankment near the centerline. The observed sur- respectively.
face settlements at the end of the fill placement and at 235 days
after the commencement of construction reached 66 and 78 mm,
7.2. Load transfer from the soil to the DCM columns
respectively. The corresponding computed values were 72 and
82 mm, respectively. Therefore, the computed surface settlements
Fig. 10 shows the pressures acting on the columns and the soil
slightly overestimated the observed values by 8% and 5%, respec-
surface between the columns as measured by the earth pressure
tively. It is shown that the computed results agree reasonably well
with the observed data considering both settlement magnitudes
and settlement rates, which implies that the stiffness of materials
and permeability values of soft clay and DCM columns modeled in
this study are reasonable. However, the overestimation is likely
due to the assumption of a constant permeability coefficient during
the 3D-FEA, whereas in reality, the permeability coefficient may be
decreased with the reduction of the void ratio under a constant
embankment load in the field. This phenomenon was not consid-
ered because of the insufficient information to establish the per-
meability coefficient and void ratio relationship for these soils.
The observed maximum settlement rate is 2.1 mm/day from the
beginning to 10 days during the filling of the first layer correspond-
ing to the high embankment filling rate. The design code for the
minimum requirement of the maximum settlement rate is unavail-
able in Thailand. Thus, the Chinese Technical Code for Ground
Treatment for Buildings reported by Jiang et al. [29] is adapted
for this project. This settlement rate (2.1 mm/day) is much less
than the requirement of 10 mm/day based on this code.
Fig. 9 shows the settlement versus depth profile 235 days after
the beginning of construction. The computed settlement captured
the overall trend but slightly underestimated the magnitude. The
observed differential settlement between the surface and the Fig. 10. Observed and computed vertical stresses carried by the column and soil
depth of 14 m was approximately 28 mm, corresponding to 36% between columns.
 P. Jamsawang et al. / Computers and Geotechnics 72 (2016) 42–56 49

cells. The embankment load is also included in Fig. 8a for reference. columns, and more loads were transferred to the column tips, so
When the embankment height increased to 1.5 m, the embank- the soft clay beneath the embankment was subjected to a much
ment load was approximately 30 kPa, and the measured pressures smaller compressive load. The remaining excess pore water pres-
acting on the soil surface increased by approximately 17 kPa, sure at the end of the construction, as shown in Fig. 11, can be used
which is approximately 50% of the embankment load. However, to explain why the settlements still increased over time in Fig. 7b.
the pressure acting on the column head increased to 34 kPa, which The continuous increase in the settlements might also result from
is approximately equal to the embankment load. It is clear that the penetration of the columns and the compression of the med-
there was load transfer from the soil to the columns as a result ium stiff clay below the column tips, as shown in Fig. 3a.
of soil arching in the embankment fill. The load transfer mecha- Typically, it is impractical to construct an embankment at a
nism may be quantified using a stress concentration ratio, which slow rate such that no excess pore pressures can develop in the
is defined as the ratio of an average vertical stress on top of a col- underlying soft clay. Lui et al. [34] warned that if the embankment
umn to an average vertical stress applied on the foundation soil. is constructed at a rate such that Bmax, which is the ratio of the
Typical reported values of the stress concentration ratio for maximum excess pore pressure to the change in total vertical
column-supported embankments range from 1 to 8 [34]. However, stress, is substantially greater than 0.34, bearing capacity failure
the measured stress concentration ratio in this study is approxi- of an embankment may occur during construction. By comparing
mately 2, which is within the range of column-supported embank- the measured excess pore pressure of 2 kPa at a depth of 5 m with
ments. Based on the measured pressures, the load carried by a the estimated applied soil pressure of 18 kPa (Fig. 10), the esti-
single column was estimated to be 9.6 kN. Qult from the field load mated Bmax is approximately 0.11, which is much lower than the
test was 300 kN, as estimated from Fig. 5; therefore, a very high recommended value. No failure was observed during the construc-
factor of safety against bearing capacity failure of a single column tion of the embankment. Moreover, the factor of safety analysis
(FSbearing) of 31 was obtained, which confirms that Qult is an was conducted for confirmation by the use of the phi-c reduction
unimportant parameter in the design of DCM column-supported approach; the strength parameters tanuand c of the soil are
highway embankments. The trends of pressures acting on the col- reduced until failure of the structure occurs. A factor of safety
umn and the surrounding soils versus time were reasonably well against global failure (FSglobal) of 4.21 was obtained.
captured, and the computed magnitudes were generally in good
agreement with observed data. The maximum difference between 7.4. Lateral movement
the observed and computed pressures on the column and the soil
surface were only approximately 9% and 13%, respectively. The improved ground consisting of DCM columns and sur-
rounding soils often shows lateral movement owing to the
7.3. Excess pore water pressure embankment weight and the earth pressures acting on the
improved ground. When the purpose of improvement includes
Fig. 11 presents the observed and computed excess pore water the reduction of lateral movement that may adversely affect
pressures over time at depths of 5 and 15 m. The computed results nearby existing structures, the examination of lateral movement
are consistent with the general trends of the observed data. As
expected, both the observed and computed results illustrate an
increase in excess pore pressures during the period of construction
owing to an increase in the surcharge load of 30 kPa, followed by a
dissipation of pore water pressure during the constant load period.
The computed results slightly underestimated the excess pore
water pressures. Fig. 11 shows that the measured and computed
excess pore water pressures at a depth of 5 m were lower than
those at a depth of 15 m. The smaller increase in pore water pres-
sures at a depth of 5 m was the result of load transfer from the
foundation soft clay to adjacent columns owing to soil arching
and some dissipation of pore water pressures during construction.
The piezometer at a depth of 15 m was close to the tips of the

Fig. 11. Observed and computed excess pore water pressures with time. Fig. 12. Observed and computed lateral movements with depth at 235 days.
50 P. Jamsawang et al. / Computers and Geotechnics 72 (2016) 42–56

is necessary. Fig. 12 shows the observed and computed lateral

movement profiles at the toe of the embankment at 235 days.
The trends of lateral movement profiles were reasonably well
captured, and the computed magnitudes were generally in fair
agreement with observed data, although under-prediction and
over-prediction were found in the soft clay and medium stiff clay
layers, respectively. The lateral movement became small and could
be neglected at a depth of 18 m, which is approximately twice the
width of the embankment base. A measured maximum lateral dis-
placement of 36 mm is located at a depth of 8 m in the soft clay,
which is approximately half the soft clay thickness. The measured
maximum lateral displacement was 18 mm in the medium stiff
clay layer, which is approximately 0.5 times the maximum lateral
displacement in the soft clay thickness. The computed maximum
lateral movement at a depth of 8 m (36 mm) in this study was
4 mm less than the observed value (an 11% underestimate),
whereas the calculated value of 18 mm in medium stiff clay was
4 mm greater than the field value (an 11% overestimate). This is
consistent with the findings of previous researchers, who noted
difficulty in obtaining consistent predictions in soil settlements
and that lateral movements are usually overestimated [51]. This
may be attributed to the combined effects of significant rotation
of principal stress under the embankment slope and the modeling
of what is likely anisotropic foundation soil with an isotropic soil
model owing to the lack of data regarding the nature of the soil
anisotropy.
According to Liu et al. [34], the ratio of lateral movement–
settlement is a good indicator of embankment stability. The ratio
of lateral movement–settlement is defined as the ratio of the max-
imum lateral movement at the toe to the maximum settlement at Fig. 13. Computed bending moments of DCM columns at the embankment toe with
the center line of an embankment. Chai et al. [10] showed that depth.

when an embankment approaches failure, the ratio increases

rapidly. At this embankment failure, the value was larger than
DCM column diameter. The maximum moment carried by a single
0.5. The observed and computed lateral movement–settlement
column was computed to be 1.54 kN-m, whereas the observed
ratios in this study (Figs. 10 and 12) were 0.42 and 0.46 at
bending moment capacity of the column was 2.8 kN-m (Fig. 6b).
235 days, respectively. The increase in the lateral movement–set-
Thus, a factor of safety against bending failure (FSbending) of 1.82
tlement ratio during the embankment construction indicates that
was obtained. It should be noticed that FSbearing, FSglobal and FSbending
the use of the DCM column-supported system enhanced the stabil-
were summarized as 31, 4.12, and 1.82, respectively. Thus, the last
ity of the highway embankment significantly. However, Zhang
parameter should be the main parameter to design. If the embank-
et al. [63] presented that the column-supported embankment with
ment height is greater than 1.5 m and/or traffic load is greater than
a maximum ratio of 0.65 can be constructed safely.
the design value, risk of local bending failure may occur.

7.5. Bending moment

8. Parametric study
The bending moments in the column under the toe of the
embankment were calculated from the following relationship: To further investigate the performance of the DCM column-
supported highway embankment, five influence factors were
MðDDCM =2Þ
rmax ¼ ra þ ð3Þ selected for the parametric study: no DCM column support, float-
IDCM ing column design, fixed-column design, embedded-column design
and embankment height. In the following discussion, the field
MðDDCM =2Þ
rmin ¼ ra
ð4Þ study is considered as the baseline case, and each influence factor
IDCM was varied one at a time to investigate its effects. The computed
where rmax and rmax are the maximum and minimum normal ver- results below are analyzed at the 235th day.
tical effective stress in the column obtained from FEM results,
which occur at points on the cross-sectional area farthest from 8.1. Effect of no DCM column support
the neutral axis, assuming that the normal vertical effective varia-
tion between integration points is linear and using the extrapola- The effect of no DCM columns to support the 1.5-m-high
tion method [11,12]. ra is the normal vertical effective stress embankment on the settlement was investigated by the FEM pre-
assuming that it is distributed uniformly on the cross-sectional dictions. Generally, prior to design of the ground improvement
area. M is the bending moment in the DCM column calculated by method, the construction of the embankment on unimproved
solving Eqs. (3) and (4). The computed bending moment profile of ground (e.g., no DCM column supports) was first analyzed to esti-
the DCM column located at the embankment toe corresponding to mate the magnitude of the maximum settlement and lateral move-
235 days is presented in Fig. 13. The shape is similar to that of a sin- ment as well as the settlement rate. The performance of the DCM
gle pile under lateral load. The maximum moment is located at a column-supported highway embankment computed by 3D-FEM
depth 3 m below the ground surface at the interlayer between the (presented in Section 6) is used as a ‘‘baseline” for the analysis of
fill material and soft clay, which is approximately 5 times the the results of the field study and the parametric study in this section.
 P. Jamsawang et al. / Computers and Geotechnics 72 (2016) 42–56 51

Fig. 14a shows the effect of no DCM column support on the evo- water pressures at a depth of 5 m were dissipated faster than the
lution over time of the settlements. The maximum settlement at corresponding value at a depth of 15 m owing to the significant
235 days and the maximum settlement rate of the foundation effect of the drainage path, causing the rate of consolidation of
without ground improvement would be approximately 404 mm the unimproved clay to be less than that of the improved clay with
and 9.1 mm/day, respectively. Obviously, these values are higher DCM columns, as shown in Fig. 14a. The computed maximum
than those with the ground improvement. This result demon- excess pore water pressures at a depth of 5 m for the unimproved
strated the effectiveness of the rigid DCM columns to increase and improved soft clay foundation were 11.5 and 1.5 kPa, respec-
the global stiffness of the foundation with the improvement factor tively. The DCM columns can effectively reduce the excess pore
of 4.9 or reduction ratio of settlement (Rs) of 0.78 and to reduce the water pressure by as much as 87% in soft clay and transfer the load
consolidation time. The fixed columns not only effectively reduce to a stronger and lower compressible bearing stratum, which is the
the total settlement of the soft soil deposit under the embankment main objective of the fixed column design. There were no signifi-
loading but also accelerate the settlement due to the much higher cant differences in the computed excess pore water pressures at
stiffness of the columns. a depth of 15 m for the unimproved and improved soft clay foun-
Fig. 14b shows the computed settlement versus depth profiles dation, which implies that the compression in the medium stiff
at 235 days for the improved and unimproved foundation. The clay layer was similar for both unimproved and improved soft clay.
large differential settlement between the surface and the depth Therefore, lower settlements were obtained because most com-
of 14 m was 349 mm, which corresponds to 86% of the computed pressions in the soft clay layer were improved. However, the rate
total settlement for the unimproved foundation. The rest of the set- of dissipation of the improved ground was higher than that of
tlement (14%) was induced in the stronger soil below. This means the unimproved ground, implying that the consolidation rate of
that the inclusion of DCM columns reduced the compression in the the improved ground is higher than that of the improved ground,
soft clay by a factor of 1.76 and increased the compression in the as shown in Fig. 7b. The computed Bmax is approximately 0.40,
stronger soil by a factor of 3.64. which corresponds to an FSglobal of 2.71. This confirms that the
For the soft clay foundation without DCM column support, the main function of DCM columns in this project is only to reduce
excess pore water pressures show similar magnitudes and trends the expected settlements because FSglobal (2.71) is sufficient to pre-
from the beginning of filling to the maximum height of filling vent slope and/or bearing capacity failure in the soft clay
(Fig. 14c), which indicates that the embankment loads were trans- foundation.
ferred from the ground surface to the medium stiff clay below. The inclusion of DCM columns reduced the computed lateral
After completion of the 1.5-m-high embankment, the excess pore movement of unimproved ground below the toe of embankment

Fig. 14. Effects of no DCM column support on computed: (a) settlement versus elapsed time (b) excess pore water pressure versus elapsed time (c) settlement movement
versus depth and (d) lateral movement versus depth.
52 P. Jamsawang et al. / Computers and Geotechnics 72 (2016) 42–56

The inclusion of the DCM columns reduces not only the total lateral
movement but also the different lateral movement in soft clay and
medium stiff clay.

8.2. Floating column design

The height of the highway embankment presented in this study

is relatively smaller than that of others reported in the literature,
such that the design by way of fixed columns may be uneconomi-
cal for some projects for which the settlement issue is not serious.
The redesign by use of the floating column concept seems reason-
able. The improvement depth is an important parameter in design-
ing the floating columns. The improvement depth ratio (dr) in this
study is defined as the depth of DCM columns under an embank-
ment divided by the depth of soft soil from the embankment base
to the top of a stronger soil layer. The improvement depths were
varied from 7 to 13 m, measured from the ground surface, corre-
sponding to dr = 0.50 to 0.93, respectively. The un-improvement
(dr = 0) and full improvement depths (dr = 1) were also plotted
for comparisons of lower and upper limits in designing the floating
columns. The reduction ratios of settlement (Rs) and lateral move-
ment (Rl) for improved ground compared to unimproved ground
are used as parameters to indicate the efficiency of the floating col-
umn design. The degree of the reduction ratio was divided into four
levels: extreme, high, medium, and low. Because there was no
standard to define these levels, the common concept of signifi-
cance in geotechnical engineering was adopted. The degrees of set-
tlement and lateral movement reductions of less than 0.4, between
0.4 and 0.6, between 0.6 and 0.8, and greater than 0.8 were consid-
ered to be low, medium, high, and extreme, respectively.
Fig. 15a shows that Rs increased with increasing dr, as expected
because deeper columns produce thinner unimproved soft clay
below. The reduction rate is significant only when dr varies from
0 to 0.7, but it becomes lower for higher values of dr. Thus,
dr = 0.7 may be considered an optimal value for an economical
Fig. 15. Effects of dr on: (a) reduction ratios and (b) factor of safety against bending
failure for floating columns. design offering Rs and Rl of 0.65 and 0.55, respectively. Moreover,
Rl increased with increasing dr because deeper columns provide
more fixity of the improved ground, leading them to behave like
(Fig. 14d). The location of the maximum lateral movement was the long columns; therefore, the lateral movement of the improved
same as for the improved ground with the magnitude of 142 mm, ground was smaller [31,25]. Rs is close to Rl when dr is greater than
corresponding to a reduction ratio of as much as 0.73. The large lat- 0.85, and it becomes different for lower values of dr, which implies
eral movement developed in both soft clay and medium stiff clay. that the floating columns have the efficiency to improve settlement
The computed maximum lateral displacement was 35 mm in the more than lateral movement. Therefore, when the purpose of
medium stiff clay layer, which is approximately 0.25 times the improvement includes the reduction of lateral movement
computed maximum lateral displacement induced in soft clay. that may yield an adverse influence on nearby existing structures,

Fig. 16. Effect of dr on computed: (a) bending moment and (b) lateral movement profiles of floating column at the embankment toe.
 P. Jamsawang et al. / Computers and Geotechnics 72 (2016) 42–56 53

dr should be high to minimize it. dr less than 0.4 can be considered

low improvement because Rs is smaller than 0.4.
Fig. 15b shows that FSbending gradually decreased with decreas-
ing dr from 1 to 0.7. In contrast, when dr is less than 0.7, FSbending
increased. This phenomenon can be explained by plotting the
bending moments and lateral movements versus the column
depths, as shown in Fig. 16a and b, respectively. Fig. 16a shows
effect of dr on the computed bending moment profiles of the
DCM column located at the embankment toe corresponding to
235 days. It can be noted that both positive (Mpos) and negative
(Mneg) bending moments developed in the columns. The magni-
tudes of maximum Mpos and Mneg increase and decrease with
increasing dr from 0.5 to 0.71 and 0.71 to 1.0, respectively. Mmax
is always positive, and its locations are the same for all values of
dr, which was located at a depth of 3 m below the ground surface
at the interlayer between the fill material and soft clay. On the
other hand, the locations of the maximum Mneg changed with
increasing dr. The existence of 3-m-thick fill material layer in the
field, of which the stiffness is much higher than that of the soft
clay, affected the magnitude and the location of Mmax. It acts as a
thick platform to restrain the column head, causing the locations
of the maximum bending moment to always be located at an inter-
layer of fill material and soft clay.
The lateral movement was significantly large for dr = 0.50, espe-
cially the movement at the column tip (Fig. 16b), which indicates
that the columns rotate similarly to rigid bodies without substan-
tial flexural distortion. The columns do not develop a substantial
bending moment [31]. Therefore, the lateral movement of the col-
umn was large, but the bending moment was small in the columns,
resulting in high FSbending. The lateral movement can be drastically
reduced whereas the induced bending moment significantly
increased (lower FSbending) with increasing dr from 0.5 to 0.71
because the column tip started to develop the fixity condition.
However, when dr was greater than 0.71, the induced bending
moments gradually decreased. This is because the column tips
are more restrained at greater depths, causing the lateral move-
ment profile trend to move backward the direction of the embank-
ment load, especially for dr = 1.0. Owing to the restrained tip
condition, the lateral soil pressure near the tip serve as the stabi-
lized force, resulting in smaller induced bending moment. Thus,
higher FSbending is obtained.

8.3. Fixed-column design

To limit the deformation of the embankment, an increase in the

size of the DCM columns by increasing the diameter of the columns
is introduced in this paper. The increase in the column size results
in a larger soil improvement area ratio [25]. Thus, the global stiff-
ness of the soil under the embankment increases, resulting in
reduced deformation. To evaluate the influence of this parameter,
four values of DDCM were considered, 0.3, 0.5, 0.6, and 1 m, corre-
sponding to the improvement area ratios of 0.03, 0.09, 0.13 and
0.35, respectively, which are based on the minimum to maximum
Fig. 17. Effects of ar on: (a) reduction ratios (b) stress concentration ratio and
sizes that can be constructed in geotechnical engineering practices
(c) factor of safety against bending failure for fixed and embedded columns.
in Thailand. The values of ar were calculated by varying the column
diameters and keeping the column spacing constant. The result
shows that the increase in ar induces smaller settlements and lat- with a diameter of 0.3 m (ar = 0.03) seem unsuitable to resist bend-
eral movement, as shown in Fig. 17a. The smaller amount in the ing failure (FSbending = 0.85) even though the degree of Rs was high.
settlements, as a result of the increase in load transferred to the This is also the reason why DCM columns with a diameter of 0.3 m
columns by the arching effect with higher n values, leads to an are unusually employed to stabilize embankment slopes. The opti-
increase of Rs (Fig. 17b). The values of n varied from the minimum mal ar for the fixed column design was 0.09 with FSbending = 1.47.
value of 1 to 5.5 for ar of 0 to 0.35, respectively. In the case of n = 1,
the embankment load was transferred directly to the unimproved 8.4. Embedded-column design
ground because of the lack of DCM column support. Extreme
reduction ratios were obtained only for ar = 0.35. Fig. 17c shows The best performance of the highway may be obtained by
that FSbending increased linearly with increasing ar. DCM columns the minimization of settlement and lateral movement. The
54 P. Jamsawang et al. / Computers and Geotechnics 72 (2016) 42–56

improvement depth is a very important parameter, as described in type, and the consequent results show that Rs, Rl and FSbending were
Section 8.2, because long columns transfer the loads exerted by the 0.90, 0.90 and 1.65, respectively. Thus, the settlement and lateral
embankment load to the bearing stratum. The column tip in the movement decreased, and FSbending increased to prevent local
field study was placed on top of the medium stiff clay layer at bending failure.
the location of the bearing stratum because it has a higher
undrained strength than that of the soft clay layer by a factor of 9. Conclusion
approximately 1.5 (e.g., su = 25 kPa). Based on the soil profiles
shown in Fig. 2, the improvement depth can be extended to the In this study, axial compression and lateral load tests in individ-
stiff clay layer by mixing medium stiff clay with cement to create ual column tests were conducted, and a field DCM column-
the column tip on stiff clay, which is a higher bearing stratum supported highway embankment was selected and analyzed using
(with su = 80 kPa) than the medium stiff clay layer. The wet process a coupled mechanical and hydraulic numerical model. The three-
with blade mixing is able to properly mix medium stiff clay with dimensional numerical results were compared to the field data. A
cement. Therefore, the improvement depth was increased from parametric study was conducted to examine five key influence fac-
14 m to 23 m to evaluate the effect of the bearing stratum. Fig. 17a tors. From the comparisons of the numerical results with the field
shows that the embedded columns have more efficiency to reduce data and from the parametric study, the following conclusions can
settlements and lateral movements than fixed columns because be drawn:
the n values are higher and there is more fixity, respectively, at
the same ar (Fig. 17b). Moreover, the embedded column with 1. The three-dimensional coupled mechanical and hydraulic
ar = 0.09 has the potential to produce the only provided ratio for model reasonably simulated the performance of the DCM
an extreme degree of reduction. column-supported highway embankment, and the numerical
The embedded column with ar = 0.09 (volume/ results compared reasonably well with the field data.
column = 4.52 m3) provides the same Rs = 0.85 as the fixed column 2. Fixed DCM columns carry most of the embankment load,
with ar = 0.35 (volume/column = 11 m3), which implies that the whereas the surrounding soils carried approximately half with
embedded column is much more economical than the fixed col- the measured stress concentration ratio of 2 by earth pressure
umn to obtain the same reduction ratios, even though the fixed cells. This shows that the fixed columns can effectively generate
column provides higher FSbending (Fig. 17c). Fig. 18a and b present a soil arching effect and transfer more loads to the column tips,
effect of dr on the computed bending moment and lateral move- causing the observed total settlements to be induced in the soft
ment profiles, respectively, for fixed and embedded columns with clay, and the stronger soil layers below were 42% and 58%,
the same ar = 0.13. The embedded column provided smaller respectively.
induced bending moment than the fixed column because of the 3. The DCM columns for this case study were originally designed
effect of the restrained column tip embedded in the medium stiff as a foundation to carry the vertical loads from the embank-
clay layer, which is similar to floating columns as discussed in Sec- ment and traffic above and to prevent global failure (slope fail-
tion 8.2. This confirms the significance of the improvement depth ure); however, the field test and analysis results reveal that
of the DCM columns in this respect. FSbending (1.82) is much lower than FSglobal (4.71) and FSbearing
(31). This emphasizes the necessity of considering the bending
8.5. Influence of embankment height failure evaluation in the design of DCM-supported embank-
ment, particularly the high embankment. Thus, a field lateral
The embankment height was increased from 1.5 m (baseline) to load test may be required to determine the moment capacity
3 m to investigate the performance of the DCM column-supported of the DCM columns, whereas the ultimate bearing capacity of
highway embankment. The results show that Rs, Rl and FSbending a single column obtained from the field axial load test may be
were 0.84, 0.86 and 1.05, respectively. Therefore, local failure of unnecessary.
the DCM column at the embankment toe may occur. The fixed type 4. The improvement depths of the DCM columns have a drastic
is improper in this case. The design is changed to the embedded impact on the lateral movements and the associate FSbending.

Fig. 18. Effect of dr on computed: (a) bending moment and (b) lateral movement profiles of fixed and embedded columns at the embankment toe for ar = 0.13.
 P. Jamsawang et al. / Computers and Geotechnics 72 (2016) 42–56 55

Low reduction in lateral movement is observed for the floating [17] Hello BL, Villard P. Embankments reinforced by piles and geosynthetics –
numerical and experimental studies dealing with the transfer of load on the
column with small improvement depths, even though FSbending
soil embankment. Eng Geol 2009;106:78–91.
is high. The increase in the improvement depths until the col- [18] Horpibulsuk S, Shibuya S, Fuenkajorn K, Katkan W. Assessment of engineering
umn tips approach or are embedded into the strong strata sig- properties of Bangkok clay. Can Geotech J 2007;44(2):173–87.
nificantly reduces both the settlement and lateral movement [19] Horpibulsuk S, Chinkulkijniwat A, Cholphatsorn A, Suebsuk J, Liu MD.
Consolidation behavior of soil cement column improved ground. Comput
and increases FSbending. The embedded column introduced in Geotech 2012;43:37–50.
this study is much more economical than the fixed column to [20] Huang J, Han J. 3D coupled mechanical and hydraulic modeling of a
obtain the same reduction ratios, owing to the higher FSbending geosynthetic reinforced deep mixed column-supported embankment.
Geotextiles Geomembrances 2009;27:272–80.
provided by the embedded columns. [21] Huang J, Han J. Two-dimensional parametric study of geosynthetics-reinforced
5. With the cases considered in this study, dr = 0.7 is considered as column-supported embankments by coupled hydraulic and mechanical
the critical value for an economical design with floating col- modeling. Comput Geotech 2010;37(5):638–48.
[22] Hong ZS, Shen SL, Deng Y, Negami T. Loss of soil structure for natural
umns if the performances regarding lateral movement and pre- sedimentary clays. Geotech Eng, Proc ICE 2007;160(3):153–9.
vention of bending failure are expected. The fixed-DCM column [23] Hong ZS, Zeng LL, Cui YJ, Cai YQ, Lin C. Compression behaviour of natural and
design with ar = 0.03 is unsuitable to stabilize embankment reconstituted clays. Géotechnique 2012;62(4):291–301.
[24] Jamsawang P, Bergado DT, Voottipruex P. Field behavior of stiffened
slopes owing to low bending moment capacity. However, this deep cement mixing piles. Proc Inst Civil Eng Ground Improve 2011;164
conclusion is based on limited data under specific conditions. (1):33–49.
A broader set of studies is needed to enhance the findings from [25] Jamsawang P, Voottipruex P, Boathong P, Mairaing W, Horpibulsuk S. Three-
dimensional numerical investigation on lateral movement and factor of safety
this study.
of slopes stabilized with deep cement mixing column rows. Eng Geol
2015;188:159–67.
[26] Jamsawang P, Voottipruex P, Jongpradist P, Bergado DT. Parameters affecting
the lateral movements of compound deep cement mixing walls by numerical
Acknowledgements
simulations and parametric analyses. Acta Geotech 2015. http://dx.doi.org/
10.1007/s11440-015-0417-5.
This research was funded by King Mongkut’s University of [27] Jamsawang P, Boathong P, Mairaing W, Jongpradist P. Undrained creep failure
Technology North Bangkok under Contract no. KMUTNB-GEN-59- of a drainage canal slope stabilized with deep cement mixing columns.
Landslides 2015, http://dx.doi.org/10.1007/s10346-015-0651-9.
10. The authors also extend their appreciation to the Office of the [28] Jongpradist P, Jumlongrach N, Youwai S, Chucheepsakul S. Strength
Higher Education Commission for granting financial support characteristics of fly ash-cement admixed clay at high water content. J
through the Higher Education Research Promotion and National Mater Civil Eng (ASCE) 2010;22(1):49–58.
[29] Jiang Y, Han J, Zheng G. Numerical analysis of a pile–slab-supported railway
Research University (NRU) Project of Thailand. The authors are embankment. Acta Geotech 2014;9:499–511.
grateful to the Department of Highway (DOH) of Thailand for per- [30] Kitazume M, Terashi M. The deep mixing method. London, UK: Taylor &
mitting the publication of the field measured data of the test Francis Group; 2013. 297–298.
[31] Kourkoulis R, Gelagoti F, Anastasopoulos I, Gazetas G. Slope stabilizing piles
embankment reported in this study. and pile-groups: parametric study and design insights. J Geotech Geoenviron
Eng, ASCE 2011;137(7):663–7.
References [32] Lai YP, Bergado DT, Lorenzo GA, Duangchan T. Full-scale reinforced
embankment on deep jet mixing improved ground. Ground Improve J
2006;10(4):153–64.
[1] Abusharar SW, Zheng JJ, Chen BG. Finite element modeling of the consolidation
[33] Lawson CH, Spink TW, Crawshaw JS, Essler RD. Verification of dry soil mixing
behavior of multi-column supported road embankment. Comput Geotech
at Port of Tilbury, UK. In: Proceedings of the international conference on deep
2009;36(4):676–85.
mixing best practice and recent advances, May 23–25, 2005. Stockholm,
[2] Abusharar SW, Han J. Two-dimensional deep-seated slope stability analysis of
Sweden; 2005. pp. 453–462.
embankments over stone column-improved soft clay. Eng Geol
[34] Lui HL, Charles W, Ng W, Fei K. Performance of a geogrid-reinforced and pile-
2011;120:103–10.
supported highway embankment over soft clay: case study. J Geotech
[3] Arulrajah A, Bo MW. Characteristics of Singapore marine clay at Changi.
Geoenviron Eng, ASCE 2007;113(12):1483–93.
Geotech Geol Eng 2008;26(4):431–41.
[35] Liu SY, Du YJ, Yi YL, Puppala AJ. Field investigations on performance of T-
[4] Arulrajah A, Abdullah A, Bo MW, Bouazza A. Ground improvement techniques
shaped deep mixed soil cement column-supported embankments over soft
for railway embankments. Proc Inst Civil Eng, Ground Improve 2009;162
ground. J Geotech Geoenviron Eng ASCE 2012;138(6):718–27.
(1):3–14.
[36] Liu H, Kong G, Chu J, Ding X. Grouted gravel column-supported highway
[5] Arulrajah A, Bo MW, Chu J, Nikraz H. Instrumentation at Changi land
embankment over soft clay: case study. Can Geotech J 2015. http://dx.doi.org/
reclamation project, Singapore. Proc Inst Civil Eng, Geotech Eng 2009;162
10.1139/cgj-2014-0284.
(1):33–40.
[37] Lorenzo GA, Bergado DT. Fundamental parameters of cement-admixed clay-
[6] Bo MW, Arulrajah A, Sukmak P, Horpibulsuk S. Mineralogy and geotechnical
new approach. J Geotech Geoenviron Eng ASCE 2004;130(10):1042–50.
properties of Singapore marine clay at Changi. Soils Found 2015;55(3):600–13.
[38] Lu W, Miao L. A simplified 2-D evaluation method of the arching effect for
[7] Borge JL, Marques DO. Geosynthetic-reinforced and jet grout column-
geosynthetic-reinforced and pile-supported embankments. Comput Geotech
supported embankments on soft soil: numerical analysis and parametric
2015;65:97–103.
study. Comput Geotech 2011;38:883–96.
[39] Ma L, Shen SL, Luo CY, Xu YS. Field evaluation on the strength increase of
[8] Butler HD, Hoy HE. The Texas quick load method for foundation load testing-
marine clay under staged construction of embankment. Mar Georesour
User’s Manual. Federal Highway Administration Report No. FHWA-IP-77-8; 1977.
Geotechnol 2011;29(4):317–32.
[9] Chai JC, Liu SY, Du YJ. Field properties and settlement calculation of soil cement
[40] Nunez MA, Briançon L, Dias D. Analyses of a pile-supported embankment over
improved soft ground – a case study. Lowland Technol Int 2002;4(2):51–8.
soft clay: full-scale experiment, analytical and numerical approaches. Eng Geol
[10] Chai JC, Miura N, Shen SL. Performance of embankments with and without
2013;153:53–67.
reinforcement on soft subsoil. Can Geotech J 2002;39(4):838–48.
[41] Ohdaira H, Hashimoto H, Gotoh K, Nozu M. Observation results for the
[11] Chai JC, Shen SL, Ding WQ, Zhu HH, Carter JP. Numerical investigation of the
embankment on soft ground improved by DJM, Tsuchi-to-kiso; 2002. p. 31–3
failure of a building in Shanghai, China. Comput Geotech 2014;55:482–93.
[in Japanese].
[12] Chai JC, Shrestha S, Hino T, Wo Ding, Kamo Y, Cater J. 2D and 3D analyses of an
[42] Okyay US, Dias D. Use of lime and cement treated soils as pile supported load
embankment on clay improved by soil–cement columns. Comput Geotech
transfer platform. Eng Geol 2010;114:34–44.
2015;68:28–37.
[43] Oliveira PJV, Pinheiro JLP, Correia AAS. Numerical analysis of an embankment
[13] Dahlström M. Dry soil mixing. In: Kirsch K, Bell A, editors. Ground
built on soft soil reinforced with deep mixing columns: parametric study.
improvement. London, UK: CRC Press Taylor & Francis Group; 2013. p. 482–5.
Comput Geotech 2011;38:566–76.
[14] Han J, Gabr MA. Numerical analysis of geosynthetic reinforced and pile
[44] Petchgate K, Jongpradist P, Jamsawang P. Field flexural behavior of soil–
supported earth platforms over soft soil. J Geotech Geoenviron Eng 2002;128
cement column. In: Proceedings of the fifth symposium 2004 on soil/ground
(1):44–53.
improvement and geosynthetics. Thailand: King Mongkut’s University of
[15] Han J, Zhou HT, Ye F. State of practice review of deep soil mixing techniques in
Technology Thonburi; 2004. p. 85–90.
China. Transportation Research Record 1808. Transportation Research Board,
[45] Pye N, ÓBrien A, Essler RD, Adams D. Deep dry soil mixing to stabilize a live
Washington, DC; 2002. p. 49–57.
railway embankment across Thrandestone Bog. In: Proceedings from the
[16] Han J, Oztoprak S, Parsons RL, Huang J. Numerical analysis of foundation
international conference on grouting and deep mixing, February 16–18, 2012.
columns to support widening of embankments. Comput Geotech 2007;34
New Orleans, Louisiana, United States; 2012.
(6):435–48.
56 P. Jamsawang et al. / Computers and Geotechnics 72 (2016) 42–56

[46] Rowe RK, Liu KW. Three-dimensional finite element modelling of a full-scale [56] Voottipruex P, Bergado DT, Suksawat T, Jamsawang P, Cheang W. Behavior and
geosynthetic-reinforced, pile-supported embankment. Can Geotech J 2015. simulation of deep cement mixing (DCM) and stiffened deep cement mixing
http://dx.doi.org/10.1139/cgj-2014-0506. (SDCM) piles under full scale loading. Soils Found 2011;51(2):307–20.
[47] Sexton BG, McCabe BA. Numerical modeling of the improvements to primary [57] Wonglert A, Jongpradist P. Impact of reinforced core on performance and failure
and creep settlements offered by granular columns. Acta Geotech behavior of stiffened deep cement mixing piles. Comput Geotech 2015;69:93–104.
2013;8:447–64. [58] Wu HN, Shen SL, Ma L, Yin ZY, Horpibulsuk S. Evaluation of the strength
[48] Shen SL, Xu YS. Numerical evaluation of land subsidence induced by increase of marine clay under staged embankment loading: a case study. Mar
groundwater pumping in Shanghai. Can Geotech J 2011;48(9):1378–92. Georesour Geotechnol 2015;33(6):532–41.
[49] Shen SL, Ma L, Xu YS, Yin ZY. Interpretation of increased deformation rate in [59] Xu YS, Shen SL, Du YJ, Chai JC, Horpibulsuk S. Modelling the cutoff behavior of
aquifer IV due to groundwater pumping in Shanghai. Can Geotech J 2013;50 underground structure in multi-aquifer-aquitard groundwater system. Nature
(11):1129–42. Hazards 2013;66(2):731–48.
[50] Surarak C, Likitlersuang S, Wanatowski D, Balasubramaniam A, Oh E, Guan H. [60] Yapage NNS, Liyanapathirana DS, Kelly RB, Poulos HG, Leo CJ. Numerical
Stiffness and strength parameters for hardening soil model of soft and stiff modeling of an embankment over soft ground improved with deep cement
Bangkok clays. Soils Found 2012;52(4):682–97. mixed columns: case history. J Geotech Geoenviron Eng 2014;140(11):10.
[51] Taechakumthorn C, Rowe RK. Performance of a reinforced embankment on a [61] Yin ZY, Yin JH, Huang HW. Rate-dependent and long-term yield stress and
sensitive Champlain clay deposit. Can Geotech J 2012;49:917–27. strength of soft Wenzhou marine clay: experiments and modeling. Mar
[52] Terashi M, Ooya T, Fujita T, Okami T, Yokoi K, Shinkawa N. Specifications of Georesour Geotechnol 2015;33(1):79–91.
Japanese dry methods of deep mixing deduced from 4300 projects on land. In: [62] Zhang J, Zheng JJ, Chen BG, Yin JH. Coupled mechanical and hydraulic modeling
Proceedings of international symposium on deep mixing & admixture of a geosynthetic reinforced and pile-supported embankment. Comput
stabilization, May 19–21, 2009. Okinawa, Japan; 2009. p. 647–52. Geotech 2013;52:28–37.
[53] Topolnicki M. In situ soil mixing. In: Moseley MP, Kirsch K, editors. Ground [63] Zhang C, Jiang G, Xianfeng Liu X, Wang Z. Deformation performance of cement-
improvement. New York, USA: Spon Press; 2004. p. 331–423. fly ash-gravel pile-supported embankments over silty clay of medium
[54] Topolnicki M. In situ soil mixing. In: Kirsch K, Bell A, editors. Ground compressibility: a case study. Arab J Geosci 2014. http://dx.doi.org/10.1007/
improvement. London, UK: CRC Press Taylor & Francis Group; 2013. p. 377–8. s12517-014-1559-8.
[55] Voottipruex P, Suksawat T, Bergado DT, Jamsawang P. Numerical simulations [64] Zhang Z, Han J, Ye G. Numerical investigation on factors for deep-seated slope
and parametric study of SDCM and DCM piles under full scale axial and lateral stability of stone column-supported embankments over soft clay. Eng Geol
loads. Comput Geotech 2011;38(3):318–29. 2014;168:104–13.