Professional Documents
Culture Documents
Centrifugal and Numerical Modeling of Stiffened Deep Mixed Column-Supported Embankment With A Slab Over Soft Clay
Centrifugal and Numerical Modeling of Stiffened Deep Mixed Column-Supported Embankment With A Slab Over Soft Clay
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
1Assistant Professor. Department of Geotechnical Engineering, Tongji University, Shanghai 200092, China.
dyzhangzhen@gmail.com
2Professor. Department of Geotechnical Engineering, Tongji University, 1239 Siping Road, Shanghai 200092, China.
3Engineer. SGIDI Engineering Consulting (Group) Co., Ltd. Shanghai 200093, China. Email: cys036@163.com
4Master student. Department of Geotechnical Engineering, Tongji University, Shanghai 200092, China. Email:
1395483534@qq.com
Abstract: A stiffened deep mixed (SDM) column can significantly increase the bearing capacity,
reduce settlement, and enhance the slope stability of soft clays as compared with a conventional deep
mixed (DM) column. This technique involves inserting plain concrete, reinforced concrete, or a steel
pile into the center of the DM column after the DM column is constructed. In this paper, a series of
column-supported embankment over soft clay. A model embankment supported only by DM columns
was constructed for comparison. Two ideal numerical models of column-reinforced soil under equal
stress and equal strain conditions were established, respectively, to explore the role the column played
in accelerating soil consolidation. A parametric study was conducted to investigate the influence
factors of the length of the core pile, column spacing, thickness of the underlying soil, modulus and
thickness of the cushion, and modulus of the slab on the load transfer of the system, and some
recommendations were proposed for its application. The load transfer mechanism of an SDM
column-supported embankment system with a slab was established based on the development of the
Keywords: Centrifugal modeling, stiffened deep mixed column, embankment, soft clay,
INTRODUCTION
With the rapid growth of highways and high-speed railways in China in recent decades, some
transportation lines have to be constructed across regions consisting of thick saturated soft soil. Such
unfavorable geotechnical conditions lead to geotechnical problems such as low bearing capacity,
excessive settlement, and slope instability. Among the techniques used to stabilize the soft soil (e.g.,
jet grouting, stone columns, deep mixed (DM) columns and vertical drains), the DM column has been
a commonly used alternative around the world (Lorenzo and Bergado 2006; Ye et al. 2013). However,
since the increase in soil strength is limited through mixing soil with cement, DM column-reinforced
soft soil cannot sufficiently meet the desired bearing capacity and/or allowable deformation, especially
stiffened deep mixed (SDM) column was proposed (Dong et al. 2002). An SDM column is formed by
installing plain concrete, reinforced concrete, or a steel pile into the center of the DM column
immediately after the DM column is constructed (Zheng et al. 2009). Installing a core pile into a DM
column significantly increases the bearing capacity and reduces the deformation of soft clay owing to
the high strength and stiffness of the core pile (Tanchaisawat et al. 2008; Jamsawang et al. 2011;
Voottipruex et al. 2011a; Raongjant and Jing 2013; Wonglert and Jongpradist 2015). Voottipruex et
al. (2011a) and Wang et al. (2014) stated that the bearing capacity of a single SDM column was 3 to
3.6 times that of a conventional DM column with the same column geometry. Jamsawang et al. (2011)
further indicated that the length of a concrete core pile significantly affected the axial ultimate bearing
capacity of an SDM column, while the cross-sectional area of a concrete core column significantly
affected its lateral ultimate bearing capacity. As compared to the conventional DM column, an SDM
column has high bending stiffness, and it can reduce the lateral deformation of soft soil (Voottipruex
Page 3 of 54
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
et al. 2011b; Jamsawang et al. 2011). Voottipruex et al. (2011b) and Raongjant and Jing (2013)
indicated that the lateral bearing capacity of an SDM column was 11 to 15 times and 3 to 4 times
greater than that of a DM column under a static load and cyclic load, respectively.
In a system of SDM column-reinforced soft soil, the external load is mostly concentrated onto the
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
core pile head, and then the axial load carried by the core pile is transferred to the soil-cement column
socket through the skin friction along the pile shaft. The additional stress is spread to the surrounding
soil. The gradual reduction in stiffness from the core pile and DM column to the soft soil becomes a
reasonable load transfer system to effectively play the bearing performances of the individual
components. Tanchaisawat et al. (2008) investigated the interface behavior between a DM column and
a concrete core pile, and indicated that the adhesion intercept was sensitive to the cement content
compared to the interfacial friction angle. Voottipruex et al. (2011a) and Wonglert and Jongpradist
(2015) observed three possible failure modes for an SDM column when it was subjected to static axial
load: soil failure, pile failure at the core tip, and pile failure in the DM socket at the top part of the
pile. Ye et al. (2017) investigated the load transfer effect of SDM column-reinforced ground under an
Owing to the large strength and stiffness of an SDM column, the spacing of SDM columns can be
enlarged to make a single column carry a higher embankment load. The commonly used area
replacement ratio of an SDM column is in the range of 5 to 10% in practice, which is significantly
smaller than those used for conventional DM columns (typically in a range of 15 to 30%). The large
spacing of SDM columns results in a large differential settlement between the column and soil. Such a
large differential settlement is reflected on the embankment surface when the embankment height is
not high enough. Therefore, an adequate embankment height is required in design. (For example,
BS8006 (2010) requires that the embankment height is greater than 1.4 times the clearing spacing of
promote the load transfer onto the pile and reduce the differential settlement on the embankment base.
However, the required embankment height is still needed to meet the critical height for a complete soil
arch. In some regions such as the east costal area in China, the resources of high-quality backfilling
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 4 of 54
material (i.e., crushed stone or gravel) are scarce and must be transported from other places, resulting
in an increase in the construction cost. To effectively increase the load carried by the SDM column
and reduce the embankment height, a combination of SDM columns with concrete slabs is proposed to
support embankments over soft clay (see Fig. 1). A gravel cushion (typically 20-50 cm in thickness) is
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
placed above the SDM column head, followed by placing a slab on top of the gravel cushion. Only a
relatively thin embankment is necessary for embankment construction. The concrete slab is used to
transfer more embankment load onto the SDM column head, and the gravel cushion acts to adjust the
stress distribution between the SDM column and the surrounding soil. This embankment system has
been used to support the embankments of high-speed railways in China (for example, Qu-Ning
high-speed railway in Fujian province). However, the behavior of this new system of SDM
column-supported embankments with a slab over soft clay has not been well investigated so far.
This paper presents a series of centrifugal modeling tests of SDM column-supported embankments
over soft clay. A model embankment was constructed in a strongbox, and mini-earth pressure cells,
displacement transducers, strain gauges, and mini-piezometers were installed to monitor the
performance of the SDM column-supported embankment. For comparison, a model test of the DM
column-supported embankment was also performed. A parametric study based on a unit cell concept
model was conducted to investigate the influence factors of the lengths of the DM columns and
concrete core piles, slabs, and bearing stratum conditions (i.e., end bearing column and floating
column) on the load transfer behavior. The load transfer mechanism of the embankment system with a
slab was discussed and established based on the development of the volumetric strains and the
CENTRIFUGE APPARATUS
The centrifugal model tests were conducted in the Geotechnical Centrifuge TLJ-150 at Tongji
University (see Fig. 2). The centrifuge had a radius of 3 m, maximum acceleration of 200 g, and
maximum capacity of 150 g-tons. The details of the centrifuge were introduced by Ye et al. (2015).
The model tests were conducted in a strongbox with inside dimensions of 800 mm (length) × 500 mm
Page 5 of 54
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
(width) × 500 mm (height). The strongbox was made of stainless steel plates.
MODEL MANUFACTURING
To minimize the side effects owing to the friction of the steel side walls of the box, two layers of
polytetrafluoroethene (PTFE) membrane were placed on the inside of the steel sidewalls. One layer of
PTFE membrane was pasted on the inside of the steel sidewall, and the other layer was directly placed
on the first layer so that it was allowed to move freely along with the model foundation. Four drainage
pipes were installed at the four corners of the strongbox to collect the water drained from the soil
The third stratum in Shanghai, called a soft silty clay layer, was sampled from the field to
manufacture the model ground. Two subsoil conditions were manufactured for the model ground: (a)
subsoil consisting of a 300-mm-thick soft soil layer and (b) subsoil consisting of a 200-mm-thick soft
soil layer, underlain by a 100-mm sand layer. The model ground was prepared using the following
procedures. For the model ground with a 300-mm-thick soft soil layer, the soft silty clay was air-dried,
coarsely ground, and sieved to pass a 2-mm screen. The soil powder was mixed with water to
manufacture uniform slurry with a water content of 200 % of its liquid limit, and then the clay slurry
was deaired for 48 h in the strongbox. The model ground was preconsolidated under the self-weight at
a centrifuge acceleration of 50 g for about 18,000 s. For the model ground with two soil layers, a fine
sand layer with a thickness of 100 mm was placed at the bottom of the strongbox. The preparation of
the upper soft soil layer was the same as that introduced previously.
To assess the consolidation effect under the self-weight, mini piezometers were installed at three
depths (150, 250, and 350 mm from the ground surface) to monitor the dissipation of excess pore
water pressure during preconsolidation. Fig. 3 shows the variation of the excess pore water pressures
with time during the flight. It can be seen that the excess pore water pressures at the three depths were
almost dissipated after 18,000 s since the ratios of the excess pore water pressure at the end of
consolidation to the initial excess pore water pressure ( ut u0 ) were small (4.3%, 7.1%, and 18.8%
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 6 of 54
for the three depths, respectively). After completion of the model ground, laboratory tests were
conducted to test the soil properties. Table 1 tabulates the main properties of the materials used in the
centrifugal tests. It can be seen that the model ground was almost identical in each test.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
The DM column was manufactured using a mixture of soft silty clay and Portland cement. The
core pile was modeled by an aluminum tube with an outer diameter of 8 mm and an inner diameter of
6 mm. The soil-cement slurry was filled into a hollow polyvinyl chloride (PVC) tube with an internal
diameter of 16 mm. Before the soil-cement was hardened, the aluminum tube was inserted into the
center of the soil-cement column. The ratio of the cross-sectional area of the core pile to that of the
SDM column was 0.25, which approximately equals the average value of the typical ranges of 0.04 to
0.4 used in the field (Zhao et al. 2010). After curing for 28 days, the model SDM column was
extracted from the PVC tube. The unconfined compression strengths of the soil-cement samples at 28
days varied from 0.97 to 1.15 MPa with an average of 1.10 MPa. The unconfined compressive
strength of the DM column obtained from the field typically ranged from 0.5 to 1.5 MPa (Voottipruex
et al. 2011b; Zhang et al. 2014). The aluminum tube had a compressive rigidity of 1.52 × 106 N, which
followed the compressive rigidity in a manner similar to the prototype concrete core pile.
After the self-weight consolidation of the model foundation soil, the centrifuge was stopped to
produce the SDM column-reinforced ground. The model ground was trimmed to 300 mm in thickness.
A plastic plate with prepunched apertures was placed on top of the model ground to precisely locate
the column positions. Apertures with a diameter of 18 mm were arranged in a square grid pattern at a
spacing of 58 mm. A thin cylindrical wall tube with an outer diameter of 18 mm was vertically pushed
into the soil through the apertures. The soil inside the tube was carefully removed, and a model SDM
column was inserted after removing the tube. This procedure was repeated until all of the model
columns were installed in the soil. After that, the improved model ground was brought up again to the
centrifuge at an acceleration of 50 g to minimize the soil disturbance effect and mitigate the gaps
After completion of the ground improvement work, the model embankment was placed on the
model foundation. The model embankment had a base width of 340 mm, a slope of 1.5H:1V, and a
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
height of 80 mm. The embankment consisted of a gravel cushion, and a slab and embankment fill
from bottom to top. The model gravel cushion was made of medium sand. The model slab was made
of a high-density polyethylene (HDPE) plate with a thickness of 8 mm. The use of the HDPE plate
followed the flexural rigidity in a manner similar to the prototype concrete slab. The model
embankment fill was made of fine sand with a side slope of 1.5H:1V. Fig. 4 shows the grain size
distribution curves of the adopted sand. The uniformity coefficients (Cu) and the coefficients of
curvature (Cc) of the sand used to model the gravel cushion and the embankment fill were 3.80 and
Table 2 shows the model test program. Six model tests were conducted to investigate the influence
factors of column length (i.e., DM column length and core pile length), slab in the embankment, and
bearing stratum conditions (i.e., end bearing or floating column) on the performance of the SDM
column-supported embankment over soft clay. For comparison purposes, the ground improved only
reinforcement is commonly included in the gravel cushion. However, to make a direct comparison of
the performances between the DM column with a slab and the SDM with a slab, the inclusion of
geosynthetic reinforcement in the gravel cushion is not considered in the centrifugal test.
Both the SDM columns and the DM columns had a diameter of 16 mm and were installed in a
square pattern at a spacing of 58 mm. The area replacement ratio of the column (i.e., SDM column or
DM column), which is defined as a ratio of the cross-sectional area of a single column to its
corresponding influence zone, was 6% for all of the centrifugal model tests. In the following
discussion, the model tests are denoted as F/E-SDM/DM-S/NS(LX/X), in which F/E represents the
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 8 of 54
end-bearing column or the floating column, SDM/DM represents the SDM column or the DM column,
S/NS represents the case with a slab or without a slab, and LX/X represents the length of the core pile
(former X) and DM column (latter X). For instance, F-SDM-S(L160/200) denotes a model with a
floating SDM column with a 160-mm-long core pile, 200-mm-long DM column, and a slab on the
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
embankment base.
Fig. 5 shows a schematic of the model tests. The monitoring instruments were all installed near the
central area of the model embankment. Mini-earth pressure cells were placed at midspan in the
surrounding soil. A displacement transducer was installed on the embankment base to monitor the
ground settlement. Mini piezometers were installed at depths of 50, 150, and 250 mm from the ground
surface.
After completion of the model SDM column-supported embankment, the strongbox was installed
in the centrifuge. The embankment construction was simulated by increasing the centrifugal
(equivalent to 23 days in the prototype). Then, a model test was run in a 50-g centrifugal acceleration
field for 50,000 s to simulate a four-year soil consolidation under the embankment load in the
prototype. Since the vertical and lateral bearing capacities of the SDM column and ground
al. 2011; Jamsawang et al. 2011), the possibility of instability in SDM column-supported
embankments was significantly mitigated. Therefore, the construction time can be reduced by a fast
rate of embankment construction. This is another advantage of the application of SDM columns.
Fig. 6 presents the ground settlements on the centerline of the embankment in the process of
embankment construction and consolidation. Fig. 6(a) shows that the SDM column significantly
reduced the ground settlement as compared to the DM column (i.e., at the end of the test, the ground
settlements were 14.2 mm and 21.6 mm for the model tests denoted as F-SDM-S(L160/200) and
Page 9 of 54
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
F-DM-S(L200), respectively). This can be explained by the fact that the SDM column had a modulus
that was much greater than that of the DM column (the concrete core pile commonly had a modulus
greater than 10 GPa, and the DM column had a modulus typically in a range of 30 to 150 MPa). Since
the area replacement ratio was only 6%, the composite modulus of the SDM column-reinforced zone
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
was greater than that of DM column-reinforced zone. Thus, the model test with a DM column had a
larger ground settlement than the model test with an SDM column. Fig. 6(b) shows that increasing the
length of the DM column can more effectively reduce the ground settlement as compared with
increasing the length of the core pile. At the end of testing, the model tests denoted as F-DM-S(L200),
F-SDM-S(L160/200), and F-SDM-S(L120/200) had ground settlements of 21.6, 14.2, and 15.0 mm,
F-SDM-S(L160/200) and F-SDM-S(L120/200) had ground settlements that were 34.3% and 30.6%
less than that of the DM model test (F-DM-S(L200)). Voottipruex et al. (2011a) stated that
end-bearing SDM columns reduced the settlement by 40% as compared to the DM columns based on
Fig. 6(c) shows that the floating-type SDM column-reinforced soil with and without a slab had
settlements of 14.2 mm and 20.1 mm, respectively. The inclusion of a slab significantly reduced the
soil compression within the column length, as more load was concentrated on the stiffer inclusion. On
the other hand, the inclusion of a slab in the embankment fill significantly reduced the differential
settlement between the column and the surrounding soil (see Fig. 7). Fig. 6(d) shows that the
end-bearing-type SDM column significantly mitigated the ground settlement as compared to the
floating-type SDM column. In the end-bearing case, most loads were concentrated onto the SDM
column head through the slab and were transmitted to the underlying firm soil through the pile shaft.
Owing to the large modulus of the SDM column, the compression of soil within the column length
was low. As a result, only 2.3 mm of total settlement was observed in the centrifugal test of the
end-bearing case.
It is interesting to note that in Fig. 6(a) and 6(c), the settlements in the model test of
initial stage. In the process of column installation, the model column was inserted into the hole after
the soil inside the hole was removed. The large initial settlement could result from the fact that the
hole cleaning was not conducted very well and some disturbed soil remained at the bottom of the hole.
With the increase of the centrifugal acceleration up to 50 g within 800 s, the column with a certain
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
thickness of disturbed soil under the column tip experiences a quick settlement, resulting in a large
Fig. 8 presents the variation of the earth pressures at the midspan surrounding soil between four
columns in the process of consolidation. In all of the model tests, the earth pressures at the midspan
surrounding soil decreased with time, which indicates that with a process of soil consolidation, more
embankment load was transferred onto the columns. This stress transfer process from soil to columns
is called the “stress concentration.” Han and Ye (2001) theoretically demonstrated this load transfer
mechanism. Ye et al. (2012) and Huang and Han (2009) also found the additional total stress in the
soil decreased and the additional total stress in the column increased during the consolidation process.
The earth pressure at the midspan surrounding soil in the model test E-SDM-S(L160/200) was
smallest among all of the model tests and decreased rapidly to a constant value. In the model test with
a floating-type column, the earth pressures gradually reduced to a constant value. As compared with
the DM column model, the earth pressure at the midspan surrounding soil in the SDM column model
(i.e., F-SDM-S(L160/200)) significantly decreased since the SDM column had higher stiffness.
However, the SDM column model without a slab (i.e., F-SDM-NS(L160/200)) had higher earth
pressure than the SDM column model with a slab (i.e., F-SDM-S(L160/200)). Therefore, the
installation of a rigid slab on the top of the embankment base helped transfer more embankment load
To further investigate the load transfer effect, the load transfer between the columns and the
surrounding soil was evaluated by the efficacy of the load transfer, which is defined as a proportion of
the total vertical load carried by the columns. Low et al. (1994) determined that the efficacy can be
Page 11 of 54
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
expressed as
Efficacy 1 s 1 m 100% (1)
hm
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
where s is the earth pressure on the surrounding soil, is the unit weight of the embankment fill,
hm is the height of the embankment, and m is the area replacement ratio of the column. Fig. 9
presents the variation of the load transfer efficacy with time. With the elapse of time, the efficacy of
the load transfer gradually increased to a constant value. The model of the end-bearing SDM column
(i.e., E-SDM-S(L160/200)) rapidly increased to a constant value and achieved the maximum final
efficacy of the load transfer (i.e, 89.3%). A comparison between model tests F-SDM-S(L160/200) and
F-SDM-NS(L160/200) shows that the inclusion of a slab significantly increased the efficacy of the
load transfer from 33.2% to 63.9%. Chevalier et al. (2010) investigated the load transfer in a
pile-supported thin-fill platform and indicated that when a rigid slab was applied, the load transfer was
much greater than that without a slab. A comparison between F-DM-S(L160/200) and
F-SDM-S(L160/200) shows that a greater difference in stiffness between the column and the
The load transfer mechanism of the SDM column-supported embankment system without a slab is
a result of the soil arching effect, which was termed by Terzaghi (1943). The differential settlement
between the soil and the column is restrained by the shear stress developing in the embankment fill,
which reduces the load acting on the surrounding soil and increases the load acting on the column. Ye
et al. (2017) proposed a modified method for the use of EBGEO (2011) to calculate the load transfer
efficacy of SDM column-supported embankments over soft soil. It was expressed as follows:
where Efficacy0 is the original efficacy of the load transfer calculated by EBGEO (2011), f ( c ) is
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
the modified factor, αc is the area ratio of the concrete core pile, s is the spacing of the SDM column,
dc is the diameter of the core pile, and hm is the embankment height. In the centrifugal modeling test of
F-SDM-NS(L160/200), the ratio of the embankment height to the clearing spacing between the core
piles was 1.6, which was greater than 1.5. Thus, the modified factor was 0.828 based on Eq. (2b). The
original efficacy of the load transfer by EBGEO (2011) was 44%. Thus, the calculated load transfer
efficacy was 36% after multiplying by the modified factor of 0.828, which agreed with the modeling
test result of 33%. The above analysis demonstrates that the modified method proposed by Ye et al.
(2017) is feasible for the calculation of the load transfer efficacy of SDM column-supported
Fig. 10 presents the variations of the measured excess pore water pressures with time at depths of
50, 150, and 250 mm along the centerline of the embankment. In all of the tests, the excess pore water
pressures increased in the process of increasing the centrifugal acceleration until the centrifugal
acceleration was 50 g, and then they dissipated gradually with time. Fig. 10(a) shows that the
maximum excess pore water pressure in the model test with DM columns (i.e., F-DM-S(L200)) was
higher than that in the model test with SDM columns (i.e., F-SDM-S(L160/200)) at depths of 50 mm
and 150 mm. This can be explained by the fact that as the SDM column had a larger modulus than the
DM column, a greater embankment load was carried by the SDM columns as compared with the
model test with a DM column. The reduction of the embankment load carried by the surrounding soil
resulted in a reduction of the excess pore water pressure in the soil. Han and Ye (2001) and Ye et al.
(2012) found this phenomenon when investigating the consolidation characteristics of soft soil with
Page 13 of 54
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
stone columns or DM columns. However, below the column base, the increment of excess pore water
pressure in the model test of SDM columns was slightly larger than that in the model test of a DM
column. Since the overall stiffness of the SDM column-reinforced area was higher than that of the DM
column-reinforced area, additional stress was concentrated in the SDM column-reinforced area and
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
Fig. 10(b) shows the dissipation of the excess pore water pressure at a depth of 50 mm. In the
model tests denoted as F-SDM-NS(L160/200) and F-SDM-NS(L120/200), increasing the core pile
from 120 to 160 mm had a minor effect on changing the rate of pore water dissipation. A possible
explanation is that since the length of the DM column and the thickness of the unreinforced soft soil
below the column base were the same in the two tests, the stiffness difference between the SDM
column and the surrounding soil did not experience a significant change after increasing the core pile
from 120 to 160 mm under the floating column condition. As a result, the load carried by the
surrounding soil was similar. However, the maximum excess pore water pressure decreased by 5.8
kPa after the length of the DM column was increased from 200 to 240 mm by comparing the model
tests of F-SDM-NS(L160/200) and F-SDM-NS(L160/240). When the SDM column was seated on
firm soil (i.e., the model denoted as E-SDM-S(L160/200)), the increment of excess pore water
pressure was small. This confirms that increasing the improvement depth in a thick soft soil stratum
has a stronger effect on reducing the maximum excess pore water pressure in the soil between the
piles than increasing the length of the core pile but maintaining the total length of the SDM column.
Fig. 10(c) shows the influence of the slab on the excess pore water pressure in the soil when it is
subjected to the embankment load. As compared with the model without a slab, the maximum
increments of excess pore water pressure in the soil decreased by 7.5 kPa and 5.9 kPa at depths of 50
mm and 250 mm, respectively, when the slab was installed. It can be speculated that the inclusion of a
slab enhanced the load transfer effect onto the SDM column, resulting in a low stress carried by the
soil between the columns. The load transfer mechanism will be discussed in a later section.
It is assumed that the soil consolidation at different depths in the model tests followed the
one-dimensional consolidation theory under one-way drainage conditions by Terzaghi (1943). The
2u0 Mz M 2Tv
u z , t sin e , M 2m 1 2, m 0, 1, 2... (3)
m 0 Mπ Hd
where Tv Cv t H d2 is the time factor for vertical drainage, Cv is the vertical coefficient of
consolidation, u0 is the initial excess pore water pressure, and H d is the drainage path. Bromwell
and Lambe (1968) proposed an incremental time method to back calculate the coefficient of
consolidation of soil based on the measured pore water pressure in the field. Based on Eq. (3), the
ratio of the excess pore pressure at two consecutive elapsed times can be expressed as
π 2 Cv
ui t t
2 i i 1
Fv e 4 H d (4a)
ui 1
π2
2u0 Mz M 4 Tvi
2
m0 M
sin e
Fv Hd (4b)
2 π2
2u0 Mz M 4 Tvi1
m0 M
sin e
Hd
where ui and ui -1 are the excess pore water pressures, and Tvi and Tvi-1 are the time factors at
time ti and ti 1 , respectively. Cao et al. (2001) demonstrated that the value of Fv converges to unity
Cv π 2 ti ti 1
4 H d2
ui e ui 1 (5)
Page 15 of 54
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Thus, the measured excess pore water pressures at different depths during a later preloading period
(e.g., a degree of consolidation greater than 60%) were divided into a number of data items at
successive equal time intervals ( t =800 c. in this study), and then the excess pore water pressure for
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
the current time step ui was plotted against the excess pore water pressure for the previous time
step ui -1 . We draw a best-fit line from the origin through these data points, and the vertical
coefficient of consolidation is determined based on the slope of the best-fit line denoted as β. Cv can
4 H d2
Cv ln (6)
t 2
As the soil consolidation at different depths was assumed to be under one-dimensional consolidation
with one-way drainage conditions, the drainage path ( H d ) is the thickness of the model foundation.
Fig. 11 shows the measured excess pore water pressure and the best-fit equations for ui vs. ui -1 .
Fig. 12 presents the back-calculated vertical coefficients of consolidation based on the measured
pore water pressures. It is worth noting that the back-calculated Cv represents the value at a certain
location, which varied in a narrow range from 0.015 to 0.023 cm2/s with an average of 0.0188 cm2/s.
The soft soil in the centrifugal test was relatively homogeneous. Since the back-calculated values of
Cv were almost independent for depth, the back-analysis results validate the feasibility of the
Similarly, based on the measured excess pore water pressures in the model ground during
preconsolidation under the self-weight, the back-calculated values of Cv were from 0.020 to 0.027
cm2/s. The coefficients of consolidation of the model foundation experienced a small reduction owing
To further investigate the effect of columns on soil consolidation, three ideal numerical models
based on the axisymmetric unit cell concept were established using the software FLAC2D 5.0 (see
Fig. 13): (a) unreinforced soil deforms under an equal stress-flexible loading, (b) column-reinforced
soil deforms under an equal stress-flexible loading, and (c) column-reinforced soil deforms under an
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
equal strain-rigid loading. The DM column was selected for the analysis and was modeled using the
Mohr-Coulomb model. The soft soil was modeled using a modified Cam-Clay model. Most of the
properties of the soft soil and column were the same as those used in the centrifugal test, and the
parameters of the modified Cam-Clay model referred to a study on the remolded third stratum in
Shanghai by Sheng (2012). The elastic modulus of the DM column was estimated based on the typical
relationship of E = 100qu, where qu = unconfined compressive strength (1.1 MPa in this study).
Poisson’s ratios of the soils and DM columns were assumed based on typical values. The permeability
of the DM column was assumed as 10 times less than the soil. Table 3 presents the parameter
Under flexible loading, a uniform pressure was directly applied on the ground surface. Under the
rigid loading condition, a uniform pressure was exerted on the soil through a rigid plate without mass
to maintain the equal strain condition. A frictionless contact was created in the interface between the
rigid plate and the ground. In all of the models, a uniform pressure of 30 kPa was applied
instantaneously, followed by a long consolidation for 360 days. The boundary conditions of the model
were set as follows: the bottom boundary was fixed in both the horizontal and vertical directions, and
the two side boundaries were fixed in the horizontal direction but free in the vertical direction. The
bottom and side boundaries were defined as impermeable, and a free drainage boundary was defined
Fig. 14 shows the variations of the excess pore water pressure at the monitored locations (see Fig.
13) with the elapse of time in the three models. It can be seen that the excess pore water pressures in
the soil dissipated faster when a column was installed, especially in the case under the equal strain
condition. The soil properties related to soil consolidation (i.e., modulus and permeability) were kept
constant in the numerical models. The drainage paths for the monitored soil elements were the vertical
Page 17 of 54
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
distance from the monitored locations to the ground surface, as the three cases were almost under a
one-dimensional consolidation condition. At the end of preloading for 360 days, the vertical distances
had small reductions (i.e, 7.8 mm, 4.8 mm, and 3.4 mm) in the unreinforced soil model, the
column-reinforced soil model under flexible loading, and the column-reinforced soil model under
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
rigid loading, respectively. Therefore, the faster dissipation of the pore water pressure in the soil is not
attributed to the changes in the soil consolidation properties and the drainage path.
Fig. 15 shows the variations of the additional total vertical stress at the monitored locations in the
soil and in the column with time. It can be seen that the total vertical stress applied to the soil element
was equal to the applied external pressure in the unreinforced soil model. However, when a column
was included, the total vertical stress applied to the soil element decreased with time, while the total
vertical stress applied to the column element increased with time. This load transfer phenomenon is
more significant under the equal strain condition. In the column-reinforced soil, the shear stress was
generated in the soil owing to the uneven deformation between the soil and the column under the
equal stress condition. The shear stress transferred the additional stress in the soil onto the stiffer
column. Under the equal strain condition, owing to the great difference in stiffness between the
column and the soil, the applied load was concentrated onto the column through the rigid plate during
the soil consolidation. The above two load transfer mechanisms result in a reduction in the applied
additional total stress in the soil, which helps accelerate the dissipation of the excess pore water
pressure in the soil when a column is included. Based on the numerical analyses, the equal strain
condition had more efficiency in load transfer than the equal stress condition. Therefore, the inclusion
of a slab over the column-reinforced soil has a function to accelerate the soil consolidation.
As presented in the centrifugal test, the inclusion of a slab significantly increased the load transfer
efficacy of the SDM column as compared with that without a slab. Owing to the complexity of the
SDM column-supported embankment system, many factors such as the gravel cushion, core pile, and
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 18 of 54
bearing stratum conditions (i.e., end bearing column and floating column) also affect the load transfer
behavior of the system. A parametric study based on the unit cell concept was performed in this
section to investigate the load transfer behavior of the SDM column-supported embankment system.
The load transfer efficacy of the column was used to evaluate the degree of load transfer.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
To validate the feasibility of the unit cell model, the results from the unit cell of the prototype
model were compared with the test results of the corresponding centrifugal model. Considering the
acceleration of 50 g used in the centrifuge test, a scale factor of 50 was applied to establish the unit
cell model. The geometries of the prototype model were equal to 50 times those of the centrifuge
model. The diameter of the unit cell model was equal to the diameter of the influence zone of a single
SDM column, which was 1.13 times the column spacing for a square column pattern. The baseline
model was established according to the centrifugal test denoted as F-SDM-S(L160/200), as shown in
Fig. 16. Other models were modified based on the baseline model (see Table 4).
The core pile and the slab were simplified as elastic materials. The DM column, gravel cushion,
and embankment fill were modeled as linearly elastic-perfectly plastic materials with the
Mohr-Coulomb failure criteria. The soft soil was modeled using a modified Cam-Clay model. For
granular materials (i.e., the gravel cushion and the embankment fill), the dilation behavior was
considered based on the empirical correlation (i.e., ψ = φ-30º, in which ψ is the dilation angle and φ is
the friction angle of the soil). Most of the properties used in the numerical models are the same as
those used in the centrifugal test. The compressive rigidity of the core pile and the flexural rigidity of
the slab of the unit cell model followed the similarity theory to that in the corresponding centrifugal
test. Table 5 lists the properties used in the baseline numerical model.
An interfacial model with the Mohr-Coulomb failure criterion was created on the interface
between the core pile shaft and the DM column. The frictional angle was 25º, and the adhesion was 50
kPa, which approximates the average values reported by Tanchaisawat et al. (2008). The values of the
interfacial normal and shear stiffness were estimated by the following formula (Itasca Consulting
10 K 1.33G
k n ks (6)
z
where kn and ks = interface normal and shear stiffness, respectively; K and G = bulk and shear moduli,
respectively; and Δz = smallest width of an adjoining zone near the interface. The bottom boundary
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
was fixed in both the radial and vertical directions. The side boundaries were fixed in the radial
Table 6 lists the load transfer efficacies obtained from the unit cells of the prototype models and
their corresponding results by the centrifugal models at the end of testing. It can be seen that the
numerical results of the prototype models and the centrifugal modeling agreed well. For the end
bearing condition (i.e., E-SDM-S(L160/200)), the numerical result underestimated the load transfer
efficacy. The gravel layer might have a large deformation in the end bearing condition as the gravel
material is likely to be squeezed to the area above the surrounding soil owing to the upward
penetration of the column. This was unable to be simulated well using the Mohr-Coulomb failure
behavior. In general, however, the comparisons demonstrate the reliability of the unit cell of the
prototype model.
Fig. 17(a) shows the influence of the core pile length on the load transfer efficacy, in which the
length of the core pile changes to 0, 1, 5, 8, and 10 m. A length of the core pile equal to 0 m represents
the soil improved by the conventional DM column, and a length of the core pile equal to 10 m means
the core pile and the DM column have equal lengths. The SDM column had an obviously higher load
transfer efficacy than the DM column. The load transfer efficacy increased with an increase in the
length of the core pile, and gradually achieved a stable value after the length ratio (i.e., the ratio of the
core pile length to the DM column length) was greater than 0.5. From a practical point of view,
blindly increasing the length of the core pile cannot effectively exploit the load transfer performance
of an SDM column. Based on the numerical results, a length ratio from 0.5 to 0.8 is recommended in
Column spacing
Fig. 17(b) shows the influence of column spacing on the load transfer efficacy. The column
spacing in the baseline model was changed to 2.6, 3.3, and 4 m, which corresponds to area
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
replacement ratios of 0.04, 0.06, and 0.09, respectively. The load transfer efficacy increased with an
increase in the area replacement ratio, as expected. When the area replacement ratio increased from
0.06 to 0.09, the improvement in load transfer efficacy was not that significant as compared with the
Fig. 17(c) shows the influence of the thickness of the underlying soft soil on the load transfer
efficacy. A thickness of the underlying soft soil equal to 0 m means the SDM column was seated on a
firm soil layer (for example, bedrock). The load transfer efficacy had a significant decrease with an
increase in the thickness of the underlying soft soil. Caution should be taken when using a
floating-type SDM column in a thick soft soil layer owing to the relatively low load transfer efficacy.
Modulus of slab
Fig. 17(d) shows the influence of the slab modulus on the load transfer efficacy. A model without
a slab is also included for comparison. It can be seen that when a slab was involved in the SDM
column-supported embankment system, the load transfer efficacy had a significant enhancement as
compared to the model without a slab. When the modulus of the slab increased from 1 GPa and 10
GPa to 30 GPa, the load transfer efficacy had a slight increase owing to the already significantly high
stiffness of the slab as compared with the soft clay in the system. From a practical point of view, the
commonly used concrete grade C30 for the slabs (i.e, elastic modulus is 30 GPa) has sufficient
The gravel cushion acts to adjust the stress distribution between the SDM column and the
surrounding soil. Fig. 17(e)(f) show that the load transfer efficacy decreased with an increase in the
cushion thickness and a decrease in the cushion modulus. Therefore, it is important to design a
cushion with appropriate thickness and modulus to give full play to the bearing performances of the
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
It is interesting to note from the centrifugal tests that when a rigid slab was involved, the load
transfer efficacy of the SDM column was significantly almost twice that of the SDM column without a
higher load transfer efficacy might be owing to the fact that involving a slab in the SDM
column-supported embankment system had a different load transfer mechanism as compared to that
without a slab. To investigate the slab’s role in the load transfer of the SDM column-supported
embankment, the above unit cells of the prototype models corresponding to the centrifugal tests
investigated.
Fig. 18 shows the volumetric strains of the models with and without a slab. For better presentation,
the results in the upper portion of the model are examined herein. Owing to the inclusion of a slab, the
relatively large volumetric strains of the gravel cushion were concentrated mostly above the SDM
column head, especially above the core pile head. The strain of the embankment fill above the slab
was equal at elevation as compared to the case without a slab. The relatively stiff slab cuts off the
reflection of the differential settlement between the column and the surrounding soil to the
embankment fill. As a result, the differential settlement in the embankment fill is effectively mitigated
Fig. 19 illustrates the principal stress vectors of the model with a slab. Following the deflections of
the principal stress vectors starting from the edges of the DM column and core pile, more stress was
concentrated onto the SDM column, especially onto the core pile through the slab. The upper surface
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 22 of 54
of the slab above the column and its lower surface above the surrounding soil generated tensile
stresses. Therefore, the slab behaves like a rigid beam that carries the embankment load and transfers
the embankment fill load onto the column head through the gravel cushion. The concept of the stress
distribution angle was used to evaluate the load transfer behavior of the system with a slab. As shown
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
in Fig. 19, α and α’ represent the distribution angles along the edges of the DM column and core pile
in the gravel cushion, and β and β’ represent the distribution angles in the slab, respectively.
The stress distribution angles (i.e., α and α’ in Fig. 19) were determined by integrating the vertical
stresses distributed on a series of interface lengths (see bb’ in Fig. 19) until the integral force on the
interface of bb’ was equal to the total vertical forces on the SDM column and the core pile. Then, a
similar procedure was followed to determine the stress distribution angles β and β’. Involving a slab
above the gravel cushion significantly increases the distribution angle (i.e., β > α and β’ > α’),
resulting in more of the embankment load being transferred on the column through the slab. As
compared with the floating-type column, the end bearing column had a high distribution angle in the
gravel cushion. This explains why the end bearing column had a maximum load transfer efficacy in
Based on the preceding analysis, the load transfer mechanism of the column-supported
embankment with a slab can be established as shown in Fig. 20: when a slab is involved, the slab
behaves like a rigid beam that carries the embankment load. Owing to the great difference in the
stiffness between the column and the surrounding soil, most of the embankment load is concentrated
onto the localized area of the slab above the columns and is transferred to the column, especially the
core pile, through the soil wedge between the column and the slab.
Conclusions
This paper presented a series of centrifugal modeling tests and numerical modeling to investigate
the performance of an SDM column-supported embankment over soft clay. The influence factors of
the length of the DM column and concrete core pile, the slab in the embankment, and the bearing
stratum conditions (i.e., end bearing column and floating column) were considered in this study.
Page 23 of 54
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
1. For the floating-type column, installation of a slab significantly reduced the total and differential
settlement and increased the load carried by the SDM columns as compared to that without a slab.
2. The SDM column-supported embankment with a slab minimized the maximum increment of the
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
excess pore water pressure in the surrounding soil under the embankment load. The numerical
analyses demonstrated that the installation of a slab accelerates the soil consolidation as compared
with the unreinforced case and the reinforced case without a slab.
3. Based on numerical analyses, a length ratio in the range of 0.5 to 0.8 is recommended in terms of
the load transfer effect of an SDM column. The commonly used concrete slab with a modulus of
30 GPa has sufficient stiffness to transfer the embankment load to the SDM column. Appropriate
selections of the thickness and modulus of the gravel cushion are necessary to take full advantage
4. The slab behaves as a rigid beam to carry the embankment load. Most of the embankment load is
concentrated onto the localized area of the slab above the SDM columns and is transferred to the
column, especially the core pile, through the soil wedge between the column and the slab.
Acknowledgments
The authors appreciate the financial support provided by the Natural Science Foundation of China
(NSFC) (grant no. 51508408 and no. 41772281) and by the Fundamental Research Funds for the
References
Asaoka, A. 1978. Observational procedure for settlement prediction. Soils and Foundations., 18(4):
87-101.
Bromwell, L.G., and Lambe, T.W. 1968. Comparison of laboratory and field values of Cv for Boston
Carlsson, B. 1987. Reinforced soil, principles for calculation. Terratema AB, Linköping. (In Swedish).
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 24 of 54
Cao, L.F., Chang, M.F., Teh, C.I., and Na, Y.M. 2001. Back-calculation of consolidation parameters
from field measurements at a reclamation site. Canadian geotechnical journal, 38(4): 755-769.
Chevalier, B., Villard, P., and Combe, G. 2010. Investigation of load-transfer mechanisms in
geotechnical earth structures with thin fill platforms reinforced by rigid inclusions. International
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
Dong, P., Chen, Z. Z., and Qin, R. 2002. Use of concrete-cored DCM pile in soft ground. Chinese
EBGEO. 2011. Recommendations for Design and Analysis of Earth Structures using Geosynthetic
Han, J., and Ye, S.L. 2001. Simplified method of consolidation rate of stone column reinforced
Huang, J., and Han, J. 2009. 3D coupled mechanical and hydraulic modeling of a
Itasca Consulting Group Inc. 2006. FLAC2D User's Guide, Version 5.0.
Jamsawang, P., Bergado, D. T., and Voottipruex, P. 2011. Field behaviour of stiffened deep cement
33-49.
Lorenzo, G.A., and Bergado, D.T. 2006. Fundamental characteristics of cement-admixed clay in deep
Low, B. K., Tang, S. K., and Choa, V. 1994. Arching in piled embankments. Journal of Geotechnical
Øiseth, E., Svanø, G., Watn, A., and Emdal, A. 2002. A computer program for designing reinforced
embankments. Proceedings of the 7th international conference on geosynthetics, France, vol. 1, pp.
201-204.
Raongjant, W., and Jing, M. 2013. Field testing of stiffened deep cement mixing piles under lateral
Tanchaisawat, T., Suriyavanagul, P., and Jamsawang, P. 2008. Stiffened Deep Cement Mixing
Construction, Kingston University, United Kingdom, 9-10 September 2008. CRC Press,
Sheng, J.-R. 2012. Laboratory tests and constitutive modeling on the mechanical behavior of shanghai
clays. Master of Engineering thesis, School of Naval Architecture, Ocean and Civil Engineering
Voottipruex, P., Bergado, D.T., Suksawat, T., Jamsawang, P., and Cheang, W. 2011a. Behavior and
simulation of deep cement mixing (DCM) and stiffened deep cement mixing (SDCM) piles under
Voottipruex, P., Suksawat, T., Bergado, D. T., and Jamsawang, P. 2011b. Numerical simulations and
parametric study of SDCM and DCM piles under full scale axial and lateral loads. Computers and
Wang, C., Xu, Y. F., and Dong, P. 2014. Working characteristics of concrete-cored deep cement
mixing piles under embankments. Journal of Zhejiang University (Science A), 15(6): 419-431
Wonglert, A., and Jongpradist, P. 2015. Impact of reinforced core on performance and failure
behavior of stiffened deep cement mixing piles. Computers and Geotechnics, 69: 93-104.
Ye, G., Zhang, Q., Zhang, Z., and Chang, H. 2015. Centrifugal modeling of a composite foundation
combined with soil–cement columns and prefabricated vertical drains. Soils and Foundations,
55(5): 1259-1269.
Ye, G.B., Cai, Y.S., and Zhang, Z. 2017. Numerical Study on Load Transfer Effect of Stiffened Deep
Mixed Column-supported Embankment over soft soil. KSCE Journal of Civil Engineering, 21(3):
703-714.
Ye, G.B., Zhang, Z., Han, J., Xing, H.F., Huang, M.S., and Xiang, P.L. 2013. Performance Evaluation
of an Embankment on Soft Soil Improved by Deep Mixed Columns and Prefabricated Vertical
Zhang, Z., Han, J., and Ye, G.B. 2014. Numerical analysis of failure modes of deep mixed
Proceedings of the 2014 GeoShanghai International Conference, Shanghai, 24-26 May 2014,
Zhao, X., Wu, M., Chen, S., and Kong, D. 2010. Study on bearing behaviors of single axially loaded
SDCM pile. Deep Foundations and Geotechnical In Situ Testing - Proceedings of the 2010
GeoShanghai International Conference, 3-5 June 2010, Geotechnical Special Publication, n 205
Zheng, G., Liu, S.Y., and Chen, R.P. 2009. State of advancement of column-type reinforcement
element and its application in China. Advances in Ground Improvement: Research to Practice in the
United States and China-Proceedings of the 2009 US-China Workshop on Ground Improvement
Technologies, March 14, 2009, Orlando, Florida, Geotechnical Special Publication, n 188, p 12-25.
Page 27 of 54
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Figure Captions
Fig. 8 Variation of earth pressure at the mid-span surrounding soil between four columns with time
Fig. 10 Variation of excess pore water pressure with time: (a) at depth of 5 cm; (b) at depth of 25 cm
Fig. 13 Ideal models for comparison of consolidation process (not to scale): (a) unreinforced soil under a
flexible loading; (b) column-reinforced soil under a flexible loading; (c) column-reinforced soil under a
rigid loading.
Fig. 15 Variation of additional total vertical stress with time in soil and in column
Fig. 18 Volumetric strain of the unit cell models corresponding to (unit: m): (a) F-SDM-NS(L160/200); (b)
F-SDM-S(L160/200)
Fig. 19 Principal stress vectors of the numerical model: (a) F-SDM-S(L160/200); (b) E-
SDM-S(L160/200)
Slab
Soft soil
Firm soil
Gravel cushion
Embankment fill
SDM pile
Core pile
DM column
0
40
80
120
160
200
0.0
0.5
350
250
150
Depth(mm)
1.0
7.1
4.3
18.8
Time, 104s
ᇞut/ᇞu0(%)
1.5
350 mm
250 mm
150 mm
2.0
Percent passing, %
0
20
40
60
80
100
110
0.1
Particle size, mm
0.01
Sand cushion
Embankment fill
1E-3
Soft soil
SDM pile
SDM pile
Embankment
Embankment
transducer
Displacement
100
340
800
(a)
(b)
1.5
1
Piezometer
Settlement plate
500 300 80
Earth pressure cell
0 0
8
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
8
Settlement, mm
Settlement, mm
16 16
24 24
32 32 F-SDM-S (L160/200)
F-DM-S (L200) F-SDM-S (L120/200)
F-SDM-S (L160/200) F-SDM-S (L160/240)
40 40
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Time, 104s Time, 104s
(a) (b)
0 0
8 8
Settlement, mm
Settlement, mm
16 16
24 24
32 32
F-SDM-S (L160/200) F-SDM-S (L160/200)
F-SDM-NS (L160/200) E-SDM-S (L160/200)
40 40
0 1 2 3 4 5 6 0 1 2 3 4 5 6
4
Time, 10 s Time, 104s
(c) (d)
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
(a)
(b)
0
20
40
60
80
100
0
1
2
3
Time, 104s
4
F-DM-S (L200)
5
F-SDM-S (L160/200)
E-SDM-S (L160/200)
F-SDM-NS (L160/200)
Fig. 8 Variation of earth pressure at the mid-span surrounding soil between four columns with time
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
Efficacy, %
0
20
40
60
80
100
1
2
3
Time, 104s
4
5
F-DM-S (L200)
89%
75%
47%
50%
F-SDM-S (L160/200)
E-SDM-S (L160/200)
F-SDM-NS (L160/200)
50 50
F-SDM-S (L160/200)
50 mm 50 mm 50 mm
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
10 10
0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
Time, 10 s 4
Time, 104s
(a) (b)
60
F-SDM-S (L160/200)
Excess pore water pressure, kPa
50 50 mm
250 mm
F-SDM-NS (L160/200)
40 50 mm
250 mm
30
20
10
0
0 1 2 3 4 5 6
Time, 104s
(c)
Fig. 10 Variation of excess pore water pressure with time: (a) at depth of 5 cm; (b) at depth of 25 cm
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
∆ui, kPa
0
5
10
15
20
25
0
5
10
∆ui-1, kPa
15
150 mm: ∆ui=0.956∆ui-1, R2=0.9989
50 mm: ∆ui=0.9599∆ui-1, R2=0.9991
20
250 mm: ∆ui=0.9625∆ui-1, R2=0.9997
50 mm
250 mm
150 mm
25
Page 38 of 54
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 39 of 54
300
250
200
150
100
50
0
0.00
0.02
2
Ave. Cv
0.04
Cv, cm /s
F-DM-S(L200)
0.06
F-SDM-S(L160/240)
F-SDM-S(L120/200)
F-SDM-S(L160/200)
E-SDM-S(L160/200)
F-SDM-NS(L160/200)
z z z P
P P
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
Rigid plate
2.5 m
2.5 m
2.5 m
0.2 m 0.2 m
a B b C c
1.02 m 1.02 m 1.02 m
8m
O O O
r r r
1.64 m 0.4 m
(a) (b) (c)
Fig. 13 Ideal models for comparison of consolidation process (not to scale): (a) unreinforced soil
under a flexible loading; (b) column-reinforced soil under a flexible loading; (c) column-reinforced
0
10
20
30
40
0
100
200
Time, Day
300
In soil
c
b
a
400
0
30
60
90
0
100
200
Time, Day
In soil
c
b
a
300
B
C
In column
400
0
30
60
90
-30
120
150
15 4
10
z
O
0.8
3.27
Slab
Soft soil
8
Gravel cushion
r
Embankment fill
SDM
column
100 100
80 80
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
Efficacy, %
Efficacy, %
60 60
40 40
(DM column)
20 20
-2 0 2 4 6 8 10 12 0.02 0.04 0.06 0.08 0.10
Length of core pile, m Area replacement raito
(a) (b)
100 100
90 80
(End bearing column)
Efficacy, %
Efficacy, %
80 60
70 40
(No slab)
60 20
50 0
-1 0 1 2 3 4 5 6 0 10 20 30 40
Thickness of underlying soft soil, m Modulus of slab, GPa
(c) (d)
100 100
80
80
Efficacy, %
Efficacy, %
60
60
40
20 40
100 200 300 400 500 20 40 60 80 100 120
Cushion thickness, mm Modulus of gravel cushion, MPa
(e) (f)
(b) F-SDM-S(L160/200
Fig. 18 Volumetric strain of the unit cell models corresponding to (unit: m): (a) F-SDM-NS(L160/200);
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record.
(a)
(b)
E-SDM-S(L160/200)
Fig. 19 Principal stress vectors of the numerical model: (a) F-SDM-S(L160/200); (b)
Page 46 of 54
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
For personal use only. This Just-IN manuscript is the accepted manuscript prior to copy editing and page composition. It may differ from the final official version of record. Page 47 of 54
Soil
Slab
Granular
Embankment
β' β
α' α
fill
SDM Column
Table Captions
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
Material Properties
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
Test No. 1 2 3 4 5 6
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by GLASGOW UNIVERSITY LIBRARY on 10/28/18
Test case
F-DM-S(L160/200)
F-SDM-S(L160/200)
E-SDM-S(L160/200)
F-SDM-N(L160/200)
33.2
89.3
63.9
37.2
Centrifuge,%
30.3
78.3
64.8
38.6
Numerical,%
Table 6 Efficacies by numerical analysis and centrifugal test
Page 54 of 54