Computers and Geotechnics 36 (2009) 676–685

Contents lists available at ScienceDirect

Computers and Geotechnics

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

Technical Communication

Finite element modeling of the consolidation behavior of multi-column

supported road embankment
Sari W. Abusharar, Jun-Jie Zheng *, Bao-Guo Chen
Institute of Geotechnical and Underground Engineering, Huazhong University of Science and Technology, 1037 Luoyu Road, Hongshan District, Wuhan, Hubei Province 430074, China

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

Article history: A multi-column composite foundation is a new concept utilizing different column types with varying
Received 29 February 2008 lengths and diameters to support the embankment fill and to mobilize the strength and stiffness of
Received in revised form 22 September the soil at shallow depths. This study presents the results of finite element analyses using the finite ele-
2008
ment software PLAXIS to investigate the consolidation behavior of a road embankment constructed on a
Accepted 23 September 2008
multi-column composite foundation. The finite element results are calibrated for a period of 200 days.
Available online 1 November 2008
The settlement, horizontal displacement, differential settlement, column axial force, and the develop-
ment and dissipation of excess pore pressure are presented and discussed in detail. It is concluded that
Keywords:
Multi-column
a multi-column composite foundation allows a fast rate of consolidation and significantly increases the
FEM embankment stability. A multi-column composite foundation formed by CFG–lime columns is more
Settlement effective than one formed by SC–lime columns. The CFG–lime columns improve the long-term stability
Consolidation analysis of the embankment because the compression modulus of CFG columns is significantly greater than that
Elasto-plasticity of SC columns.
Embankment Ó 2008 Elsevier Ltd. All rights reserved.

1. Introduction the latter to be utilized as a load-supporting system. In many areas

in China, this ground improvement method is being increasingly
The construction of embankments overlying soft soil with a adopted.
high groundwater level often leads to large lateral pressures and In recent years, a substantial amount of research has been con-
movement, excessive settlements as well as slope and bearing fail- ducted on such multi-column composite foundations. For instance,
ures, which usually result in long construction delays and costly Liu et al. [4] and Chen et al. [5] performed field tests on composite
remedial works. In such instances, ground improvement measures foundation comprising two types of columns. Zhou et al. [6] carried
are often used to enhance stability and minimize ground move- out field tests on a composite foundation with three to four differ-
ment. Examples of such soil improvement methods are preloading, ent column types. Chen [7] reported finite element analyses on
vertical drains or grouting injection [1–3]. multi-column composite foundation. Yan et al. [8] developed a sys-
In China, combinations of cement–fly ash–gravel (CFG), soil–ce- tematic design method for multi-column composite foundation
ment (SC), and lime columns are often used as ground improve- while Zheng et al. [9] proposed a design methodology for CFG–lime
ment measures to support highway embankments overlying soft column composite foundation, which is now being used exten-
soils. This system consists of installing different column types with sively in China and has proven to be an economical way of control-
different lengths and diameters to support the embankment fill ling settlements within the allowable levels and increasing the
and to mobilize the strength and stiffness of the soil at shallow bearing capacity. Liang et al. [10] performed three-dimensional fi-
depths. Fig. 1 shows an example of such a composite foundation, nite element method analyses of a composite piled raft foundation
where the embankment is supported by long CFG columns or SC with different pile types. Zheng et al. [11] reported three-dimen-
columns and shorter lime columns (also called chemico columns). sional finite element method analyses of a multi-element compos-
The long columns are generally much stiffer than the short col- ite foundation with SC–lime columns. Zheng et al. [12] conducted a
umns and they are used much like piles to mobilize the bearing series of three-dimensional finite element analyses on CFG–lime
capacity of the deeper soil strata. The more flexible lime columns columns composite foundation with varying lengths and diame-
strengthen and stiffen the shallower soil strata, thereby allowing ters. However, there are very few studies on multi-column com-
posite foundations formed by CFG–lime columns or SC–lime
* Corresponding author. Tel.: +86 27 87557024; fax: +86 27 87542231.
columns at present.
E-mail addresses: [email protected] (S.W. Abusharar), zhengjj@hust. To the best of the authors’ knowledge, this is the first attempt to
edu.cn (J.-J. Zheng), [email protected] (B.-G. Chen). address the behavior of multi-column composite foundation for

0266-352X/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compgeo.2008.09.006
 S.W. Abusharar et al. / Computers and Geotechnics 36 (2009) 676–685 677

Reinforcement Embankment
(Optional)
Short columns

Long columns
Soft soil

Firm soil or bedrock

Fig. 1. A multi-column supported embankment.

supporting embankment fill. Numerical analyses considering real- both sides. The clear spacing between adjacent columns was set
istic parameters are particularly needed for better utilization of the at 1.3 m. The ground water table was taken at the ground surface.
technique. This paper presents the results of a series of two-dimen-
sional (2D) finite element analyses of the consolidation behavior of 2.2. Modeling range
a multi-column supported embankment using the finite element
software PLAXIS. Parametric studies are presented that shed light The size of the modeled domain was determined on the basis of
on effective combinations of columns. trial calculations, during which the mesh was progressively refined
and its boundaries extended until stresses and deformations at the
2. Plane strain finite element analysis highly-stressed zones have sufficiently stabilized. Results from the
trial calculation showed that the lateral boundaries should be ex-
2.1. Problem dimensions tended to 15 m on each side of the embankment.

As Fig. 2 shows, the model considered here consists of an 2.3. Boundary conditions
embankment fill supported by a multi-column composite founda-
tion. The typical subsoil profile, together with the general soil The soil beneath 16 m is a very stiff soil and its contribution to
properties of Nong Ngu Hao test embankment was chosen for this ground movement was considered to be negligible. Thus, the bot-
study [13]. The soil profile consists of three layers as follows: 8.5 m tom boundary was set at this depth and was assumed to be com-
of very soft clay; 3 m of soft clay; and 4.5 m of medium clay. The pletely rigid and impermeable. The nodes on the two vertical
problem was solved as a two-dimensional (2D) plane strain consol- boundaries were fixed against horizontal movement but allowed
idation problem with a full cross-section using the software PLAXIS to move freely in the vertical direction. The vertical boundaries
by restricting the planes perpendicular to the side of the embank- and the ground surface were set as drainage or recharge boundaries.
ment (i.e., displacements fixed only in the longitudinal directions).
The column dimensions and spacing and the overall foundation 2.4. Material model and parameters
dimensions were typical values used in engineering practice. They
are as follows: long columns – diameter = 0.4 m, length = 16 m; The embankment fill and in situ soils were modeled as linearly
short columns – diameter = 0.4 m, length = 8.5 m; embankment – elastic-perfectly plastic materials with the Mohr–Coulomb failure
crest width = 11 m, height = 2 m, gradient of side slope = 1.0 on criterion, while the columns were assumed to be linearly elastic

15m 2m 11m 2m 15m

B
+2.0
A
Embankment Water table 0.0
E D 2m
C F

Short columns G
Very soft clay 8.5m

-8.5
H
Soft clay -11.5 3m

Long columns I
Medium clay 4.5m
-16.0

Firm soil
Fig. 2. Cross-section of multi-column supported embankment model for FEM analysis.
678 S.W. Abusharar et al. / Computers and Geotechnics 36 (2009) 676–685

Table 1
Material properties of the embankment and subsoil [13]

Parameter Unit Embankment fill Very soft clay Soft clay Medium clay
Material model – M–C M–C M–C M–C
Type of material behavior – Drained Undrained Undrained Undrained
Dry unit weight kN/m3 20 16 17 18
Sat. unit weight kN/m3 20 16 17 18
Horizontal permeability m/d 0.009 0.005 0.0009 0.0007
Vertical permeability m/d 0.009 0.005 0.0009 0.0006
Effective Young’s modulus kPa 8000 2100 2300 2900
Effective shear modulus kPa 3076.92 807.69 884.62 1115.39
Effective Poisson’s ratio – 0.3 0.3 0.3 0.3
Effective cohesion kPa 1 4 2.5 5
Effective friction angle deg 30 23 23 23
Effective dilatancy angle deg 0 0 0 0
Interface reduction factor – – 0.8 0.8 0.8

Table 2 soft clayey soils has been taken as 0.30, and the dilatancy angle
Material properties of the columns [11–13] has been assumed to be zero for all soils. The Young’s modulus
was set as a multiple of its undrained shear strength, that is
Parameter Unit Lime CFG column SC column
column E ¼ a  Su ð1Þ
Material model – Linear Linear Linear
elastic elastic elastic
where Su is the undrained shear strength and a = 75–100 for Nong
Type of material – Non-porous Non-porous Non-porous Ngu Hao clay [13]. For numerical modeling, the coefficient a has
behavior been taken here to be 100.
Young’s modulus kPa 20,000 250,00,000 100,000
Axial stiffness kN/m 2520.5 31,50,625 12602.5
Flexural rigidity kNm2/ 26.47 33088.13 132.35
m
Poisson’s ratio – 0.25 0.2 0.3
Equivalent diameter m 0.355 0.355 0.355
Reference – [13] [12] [11]

material. Previous studies have considered the Mohr–Coulomb (a) Triangular element (b) Beam element
model for the soft soil layers [13–17]. In addition, the Mohr–Cou-
lomb model is considered as a first order approximation for real
soil behavior and highly recommended where soil parameters
are not known with great certainty [18]. The soil input parameters
were determined from triaxial and in situ tests and prior publica-
tions [13]. The material properties of the various components are
shown in Tables 1 and 2.
The elastic-perfectly plastic Mohr–Coulomb model requires five
parameters: friction angle, u0 , cohesion, c0 , dilatancy angle, w,
Young’s modulus, E, and Poisson’s ratio, m0 . The strength parame- (c) Interface element
ters, u0 and c0 are obtained from large-scale laboratory testing of
the soil compacted to field conditions. The Poisson’s ratio for the Fig. 4. Basic elements considered for mesh generation.

Fig. 3. Finite element mesh of a multi-column supported embankment.

 S.W. Abusharar et al. / Computers and Geotechnics 36 (2009) 676–685 679

The permeability of the clay is one of the most difficult param- tions. For most natural deposits, the hydraulic conductivity in the
eters to determine. There are many uncertainties regarding the ra- horizontal direction is higher than that in the vertical direction
tios of horizontal permeability to vertical permeability due to the [19,20]. In this paper, the horizontal permeability is assumed to
presence of silt seams, sand lenses and fissures in the in situ condi- be approximately the same as the vertical [13].

Fig. 5. Settlement at (a) point A; (b) point B; (c) point C; (d) point D; (e) point E; and (f) point F.
680 S.W. Abusharar et al. / Computers and Geotechnics 36 (2009) 676–685

2.5. Mesh generation interface, and plastic interface behavior when permanent slip oc-
curs. For the interface to remain elastic, the shear stress s is given
The FEM analysis was carried out using PLAXIS version 8.2 to by
discretize the embankment, columns and ground soil geometry
jsj < rn tan /i þ ci ð2Þ
model into finite elements as shown in Fig. 3. Fifteen-noded trian-
gular elements (Fig. 4a) were used to discretize the embankment and for plastic behavior, s meets the criterion
and ground soils. The 15-noded triangular elements are accurate jsj ¼ rn tan /i þ ci ð3Þ
elements that provide high quality stress results for complex prob-
lems. Five-noded beam elements (Fig. 4b) with two translational where /i and ci are the friction angle and cohesion of the soil–struc-
degrees of freedom and one rotational degree of freedom were ture interface. The strength properties of interfaces are linked to the
used to model the long and short columns. Five pairs of interface strength properties of the corresponding soil layer in which it is
elements (Fig. 4c) were used to model the slippage between soil embedded. Each data set has an associated strength reduction fac-
and columns. Interface elements have been extended to 0.5 m be- tor for the interface (Rinter) [18].
neath the short columns in order to allow for sufficient flexibility
around the column tip. In Fig. 4c, the interface elements are shown 2.6. Loading procedure and calculation type
to have a finite thickness, but in the finite element formulation the
coordinates of each node pair of interface elements are identical, The construction process of the embankment consisted of two
which means that the elements have zero thickness. The stiffness phases, each phase adding 1 m of height over 2 days (the average
matrix for interface elements is obtained by means of Newton– rate of construction was 0.5 m/d). A post-embankment construc-
Cotes integration. The position of the Newton–Cotes integration tion period of 196 days is allowed for consolidation in the numer-
points coincides with the node pairs. An elastic–plastic model is ical analysis. Consolidation analyses have been performed to
used to describe the behavior of the interface of the soil and simulate the development or dissipation of excess pore water pres-
structure. The Coulomb criterion is used to distinguish between sure in the soft soil as a function of time, therefore, the model was
elastic behavior, where small displacements can occur within the analyzed under an undrained condition.

Fig. 6. Variation of settlement with depth beneath (a) point F; (b) point C; (c) point D; and (d) point E.
 S.W. Abusharar et al. / Computers and Geotechnics 36 (2009) 676–685 681

3. Results and discussion

Three different cases have been considered to clarify the behav-

ior of the new composite foundation and for comparison purposes.
These cases are as follows:

(1) M1C1: without reinforcement,

(2) M2C1: the embankment fill supported by SC–lime columns,
and
(3) M3C1: the embankment fill supported by CFG–lime
columns.

Results are presented for points A–I located in Fig. 2.

3.1. Settlement analysis

Fig. 5 shows the variation of settlement with time at points A–F

for the three cases. As can be seen, settlement has more or less sta-
bilized about 200 days after construction. As can be seen, settle-
ment is largest for M1C1, followed by M2C1 and then M3C1. Fig. 8. Differential settlement at the top surface of embankment.
Furthermore, the settlement of M1C1 also takes much longer to
stabilize than the corresponding settlement of M2C1 and M3C1.

Fig. 7. Variation of settlement with horizontal distance at (a) ground surface; (b) 4.25 m beneath ground surface; (c) 10 m beneath ground surface; and (d) 13.75 m beneath
ground surface.
682 S.W. Abusharar et al. / Computers and Geotechnics 36 (2009) 676–685

Thus, the settlement is strongly influenced by the column type, umns, compared to that of the SC columns. The settlement curves
with the CFG–lime columns being the most effective in curtailing for points D and E show initial heave, followed by subsequent set-
both the magnitude and time of consolidation. This is readily tlement. The heave is maximum just after completion of the
attributable to the higher compression modulus of the CFG col-

Fig. 10. Variation of lateral displacement with depth beneath (a) point C; (b) point
Fig. 9. Lateral displacement at (a) point C; (b) point D; and (c) point E. D; and (c) point E.
 S.W. Abusharar et al. / Computers and Geotechnics 36 (2009) 676–685 683

embankment. Once again, case M1C1 shows the largest ground 3.2. Differential settlement
movement, followed by M2C1 and finally M3C1.
Fig. 6 shows the variation of settlement with depth beneath Fig. 8 shows the build-up of differential settlement between
points C–F. As shown in Fig. 6, the settlement decreases mono- points A and B with time at the crest surface of the embankment.
tonically with depth beneath points F and C, while the settlement It can be seen that the differential settlement increases first and
beneath point D increases first then decreases with depth. It can reaches a maximum at the end of embankment construction then
be seen in Fig. 6d that the upper part of the soil beneath point E gradually decreases with time as consolidation occurs. As can be
is heaving whereas the lower part is settling. It can also be seen seen, the differential settlement is strongly influenced by the col-
that the maximum settlement occurs at the ground surface at umn type; once again, the CFG–lime combination is the most effec-
point F. tive, followed by SC–lime and then the untreated case.
Fig. 7 shows the variation of settlement along the base of the
embankment at the ground surface, 4.25 m beneath ground surface 3.3. Lateral deformation
(at the middle of very soft clay layer), 10 m beneath ground surface
(at the middle of soft clay layer), and 13.75 m beneath ground sur- Fig. 9 shows the variation of lateral displacement with time at
face (at the middle of medium clay layer). The settlement profile points C–E. It can be seen that the lateral displacement starts
shows that the maximum settlement occurs at the middle and at immediately during the construction and reaches a maximum
the ground surface and keeps decreasing towards the toe of the upon completion of the embankment construction. Thereafter, it
embankment and with depth. It can also be seen that surface heave decreases with time as consolidation occurs.
starts at a distance 10 m outside the embankment and continues Fig. 10 shows the variation of lateral displacement along depth
increasing. Moreover, the maximum heave occurs at the ground beneath points C–E. It is clear that the lateral displacement slightly
surface and decreases with depth. The differential settlement be- increases first then decreases significantly along the depth. The re-
tween the columns and surrounding subsoil arising from the dif- sults show that the lateral displacement of soft ground for the third
ferences in stiffness is also evident in Fig. 8. This differential case is larger than that for the second case at shallow depths, be-
settlement also decreases with depth as settlement decreases. cause the stiffness of CFG columns is greater than that of SC columns,

Fig. 11. Excess pore pressure at (a) point F; (b) point G; (c) point H; and (d) point I.
684 S.W. Abusharar et al. / Computers and Geotechnics 36 (2009) 676–685

while at larger depths, the lateral displacement of the soft ground for It is also seen that the maximum axial force of columns is lower
the third case is smaller than that for the second case. It is also clear from the head of the columns to a certain depth. This is because
that multi-column ground treatments are effective in reducing the that the displacements of subsoil are larger than those of the col-
lateral displacement of the composite foundation along depth. umns in a range of depth along the shaft. Additionally, the negative
friction is generated by the relatively larger settlement of shallow
3.4. Excess pore pressure analysis subsoil. Below the neutral plane, the displacements of the columns
are larger than that of the subsoil with the further increase in
In this study, it is assumed that the excess pore water pressure depth, and positive skin friction is generated. Thus, the column ax-
induced by the installation of columns has dissipated before the ial force decreases with the depth again.
construction of the embankment over the soft ground. An embank-
ment construction period of 200 days is allowed for consolidation 4. Summary and conclusions
in the numerical analysis. Fig. 11 shows the variation of excess
pore pressure with time at points F–I. It is clearly shown that dis- A multi-column composite foundation utilizes different column
sipation starts immediately during the construction and continues types with different lengths and diameters to support the embank-
afterwards. The rate of dissipation reaches a maximum once the ment fill and mobilize the strength and stiffness of the soil at shal-
embankment construction is completed and then decreases with low depths. The long columns constructed from cement–fly ash–
time. It is also clear that the excess pore pressure increases with gravel (CFG) columns and/or soil–cement (SC) columns are used
the depth from the ground surface. to reduce the settlement of the underlying soft ground. The short
columns constructed from lime columns (also called chemico col-
3.5. Axial forces of columns umns) are used to improve and strengthen the shallow soft soil
strata. In this study, two-dimensional finite element analyses were
Fig. 12 shows the variation with depth of the maximum axial used to investigate the consolidation behavior of a road embank-
force in the columns. It can be seen that the greater the elastic ment constructed on multi-column composite foundation.
modulus of the columns, the greater the axial force of the columns. Based on the finite element results, the following conclusions
can be drawn:

1. Multi-column ground treatment can significantly reduce total

and differential settlements and restrict the lateral movement
of the embankment; as a result, the stability of the embank-
ment can be improved.
2. Maximum settlement occurs at the center and at the ground sur-
face. Moreover, the settlement is more at the center than that at
the toe of the embankment, which agrees with field observations.
3. Multi-column support allows for a faster rate of consolidation
and significantly increases embankment stability.
4. The multi-column composite foundation formed by CFG–lime
columns is more effective than that formed by SC–lime col-
umns. The CFG–lime columns combination improves the long-
term stability of the embankment because the fact that the
compression modulus of CFG columns is significantly greater
than that of SC columns.

References

[1] Ariema F, Butler BE. Embankment foundations-guide to earthwork

construction. Washington, DC: Transportation Research Board, National
Research Council; 1990. p. 59–73.
[2] Magnan JP. Methods to reduce the settlement of embankments on soft clay: a
review. Vertical and horizontal deformations of foundations and
embankments. ASCE Geotechnical Special Publication 40; 1994. p. 77–91.
[3] Shen SL, Chai JC, Hong ZS, Cai FX. Analysis of field performance of
embankments on soft clay deposit with and without PVD-improvement.
Geosynth Geomembranes 2005;23(6):463–85.
[4] Liu FY, Yang XB, Liu X. Field test of a composite foundation including mixed
pile. Chin J Geotech Eng 2003;25(1):71–5 [in Chinese].
[5] Chen LZ, Liang FY, Huang DZ, Wang GC. Field study on behavior of composite
piled raft foundation for high-rise buildings. Chin J Geotech Eng
2004;26(2):167–71 [in Chinese].
[6] Zhou DQ, Liu HL, Zhang KN. Experimental comparison study on behavior of
three and four-element composite foundation. J Build Struct 2004;25(5):124–9
[in Chinese].
[7] Chen Q. Behavior of combined composite ground and simulation by study FEM.
Chin Civil Eng J 2001;34(1):50–5 [in Chinese].
[8] Yan ML, Wang MS, Yan XF, Zhang DG. Study on the calculation method of
multi-pile composite foundation. Chin J Geotech Eng 2003;25(3):352–5 [in
Chinese].
[9] Zheng JJ, Abusharar SW, He C. Design theory and application of CFG–lime piles
composite ground. In: 6th International conference on ground improvement
techniques, Coimbra, Portugal; 2005. p. 651–6.
[10] Liang FY, Chen LZ, Shi XG. Numerical analysis of composite piled raft with
Fig. 12. Maximum axial force of columns (a) long column and (b) short column. cushion subjected to vertical load. Comput Geotech 2003;30(6):443–53.
 S.W. Abusharar et al. / Computers and Geotechnics 36 (2009) 676–685 685

[11] Zheng JJ, Ou JH, Wang XZ. Three-dimensional finite element analysis of multi- [16] Han J, Huang J, Porbaha A. 2D numerical modeling of a constructed
element composite foundation. In: The 9th international conference on geosynthetic-reinforced embankment over deep mixed columns.
enhancement and promotion of computational methods in engineering and Contemporary issues in foundation engineering (GSP 131). ASCE; 2005.
science, Macao; 2003. [17] Chen RP, Chen YM, Xu ZZ. Interaction of rigid pile-supported embankment on
[12] Zheng JJ, Abusharar SW, Wang XZ. Three-dimensional nonlinear finite element soft soil. Advances in earth structures: research to practice (GSP 131). ASCE;
modeling of composite foundation formed by CFG–lime piles. Comput Geotech 2006. p. 231–8.
2008;35(4):637–43. [18] PLAXIS 2D V8. Reference manual; 2002.
[13] Hossain MS, Haque MA, Rao KN. Embankment over soft soil improved with [19] Bergado DT, Enriquez AS, Sampaco CL, Alfaro MC. Inverse analysis of
chemico pile – a numerical study. Advances in earth structures: research to geotechnical parameters on improved soft Bangkok clay. ASCE J Geotech Eng
practice (GSP 151). ASCE; 2006. 1990;18(7):1012–30.
[14] Huang J, Han J, Porbaha A. Two and three-dimensional modeling of DM [20] Hansbo S. Design aspects of vertical drains and lime column installations. In:
columns under embankments. GeoCongress. ASCE; 2006. Proceedings of 9th southeast Asian geotechnical conference, vol. 2. Bangkok;
[15] Madhyannapu RS, Puppala AJ, Hossain S, Han J, Porbaha A. Analysis of 1987. p. 8-1–12.
geotextile reinforced embankment over deep mixed columns: using numerical
and analytical tools. GeoCongress. ASCE; 2006.