You are on page 1of 5

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/270587625

Finite element analysis of embankments supported on geocell layer

Conference Paper · January 2001

CITATIONS READS

7 248

3 authors, including:

Madhavi Latha Gali Rajagopal Karpurapu


Indian Institute of Science Indian Institute of Technology Madras
120 PUBLICATIONS   1,591 CITATIONS    111 PUBLICATIONS   2,401 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Application of Image Processing in Geotechnical Engineering View project

Advanced Technologies for Post-Disasters Reconnaissance, Forensic and Environmental Impact Studies-Geotechnical View project

All content following this page was uploaded by Madhavi Latha Gali on 09 January 2015.

The user has requested enhancement of the downloaded file.


Finite element analysis of embankments supported on geocell layer

K. Rajagopal G. Madhavi Latha N.R. Krishnaswamy


Associate Professor Research Scholar Professor
Department of Civil Engineering
Indian Institute of Technology Madras, Chennai India 600 036

ABSTRACT: The geocell reinforcement provides 3-dimensional confinement to the soil thus improving both
the strength and stiffness of the soil. The improvement in these two parameters is quantified in this paper
through standard triaxial compression tests on geocell reinforced sand samples. The results from these exper-
iments have been analyzed to develop simple relations to estimate the shear strength and stiffness of the geo-
cell reinforced soil in terms of the geocell properties. The frame work of hyperbolic model has been employed
to develop simple equation to estimate the stiffness of the geocell reinforced soil. This model has been im-
plemented in a finite element program to analyze the performance of geocell reinforced soils. The perfor-
mance of laboratory-scale embankments has been back-predicted using this model to verify its accuracy. This
model was found to be able to simulate the effect of the various parameters, viz. dimensions of the geocell
pockets, modulus of geocell material etc.

1 INTRODUCTION these tests are reported in Rajagopal et al. (1998).


Typical test results from the triaxial compression
The geocell is a 3-dimensional cellular mattress tests on geocell confined soil are shown in Figure 1.
made of geogrids or geotextiles, which provides all The results have shown that the geocell confinement
round confinement to the soil. Because of this con- does not change the friction angle of the soil where-
finement, the strength and stiffness of the soil in- as it induces some apparent cohesion even to granu-
creases. This type of soil reinforcement is ideally lar soils.
suited for the construction over extremely soft soils
1200
where the use of planar geosynthetics may not be
adequate.
The 3-dimensional finite element simulation of
problems with geocell reinforcement requires the
generation of membrane type elements in a criss-
cross manner to model geocell walls. This is rather 800
(s1-s3) kPa

cumbersome and hence it is preferable to work with


a 2-dimensional equivalent model that can very well
replicate the behavior of a 3-dimensional system.
This paper deals with development of such a model four geocells
and its verification from back-analysis of laboratory 400 three geocells
tests on large-scale model embankments.
two geocells
single geocell
2 TRIAXIAL COMPRESSION TESTS
0
The strength and stiffness behavior of geocell en-
0 5 10 15 20
cased granular sands was investigated through a se-
ries of triaxial compression tests. The tests have used axial strain %
configurations of one geocell, two, three and four in-
terconnected cells. The geocells were made of four
different types of geotextiles whose 5% secant stiff- Fig 1. Typical stress-strain results from triaxial
ness ranged from 0.5 to 70 kN/m. The details of compression tests
It could be observed that as the number of geocells cell-soil composite was expressed in the proposed
is increased, the strength and stiffness of the soil in- model in terms of the Ku of the unreinforced soil and
creased. However, when the number of cells was in- the secant modulus of geocell walls (M in kN/m) as
creased beyond three, there was not much further in- follows,
crease in the strength and stiffness. This result shows
that the strength and stiffness behavior observed in Kr - Ku =  50 M0.16 (4)
laboratory tests with three inter-connected cells is
representative of geocells in real cases with many in-
in which  is an interaction parameter between the
ter-connected cells.
geocell pockets. Its value for the case with three and
four cells was found to be equal to 4. The above
3 EQUIVALENT COHESIVE STRENGTH OF constants in the equation were determined by regres-
GEOCELL CONFINED SOIL sion analysis of the test data. The validity of the
above equation was verified by successfully back-
Based on the results from the above triaxial com- predicting the experimental data by implementing
pression tests, an equation has been derived to esti- the above constitutive model in a finite element pro-
mate the apparent cohesion induced to the soil in gram. In these analyses, the geocell confined soil
terms of geometric and material properties of the was represented using the equivalent shear strength
geocell based on the membrane correction theory. properties derived from Equations 1 and 2 and the
The additional cohesive strength (cr) due to single stiffness derived from equations 3 and 4.
geocell encasement was obtained as,
s 3
cr  kp (1) 5 LARGE-SCALE LABORATORY
2
in which kp is the Rankine’s passive earth pressure EMBANKMENT MODEL TESTS
coefficient and s3 is the additional confining pres-
sure due to the membrane stresses given as, These tests were performed within a steel test tank
1800 mm long, 800 mm wide and 1200 mm height.
The tank was fitted with 20 mm thick perspex sheet
2M c 1 2M 1 - 1 -  a  on one longitudinal side and thick steel plates on all
s 3     (2) the other sides. All the sides of the tank were lined
d (1 -  a ) d o  1 -  a 
 with two layers of plastic sheets whose inner layers
were coated with lubricating oil to create perfect
The modifications to above equation to consider plane strain conditions in the tank.
multiple cells are reported in Rajagopal et al. (1998). The clay soil was initially mixed with excessive
amounts of water and poked with steel rods to obtain
4 STIFFNESS OF GEOCELL CONFINED SOIL uniform mixing and then consolidated under a sur-
charge pressure of 10 kPa for one week. After this
Due to the increase in the confining pressures exert- consolidation, the layer was trimmed to a 600 mm
ed on the soil by geocell walls, its stiffness will also depth for all tests. This procedure was found to pro-
increase as illustrated in Fig. 1. The increase in the duce a soft clay bed with uniform properties. Several
confining pressure depends on the geometric proper- undisturbed samples were collected from the bed af-
ties of the geocell and the material properties of the ter consolidation to determine the CBR value and
geosynthetic used to make the geocell. The follow- the vane shear strength of the clay soil. The unit
ing well known hyperbolic equation was used to weight, moisture content, vane shear strength and
quantify the effect of the confining pressure on the CBR value of the clay foundation were maintained
Young’s modulus. almost the same for all the model tests by carefully
controlling the amount of water added and using
 R (1 - sin  )(s 1 - s 3 ) 
2 n
s3 
E t  1 - f  K Pa   (3) uniform mixing procedures.
 2c cos   2s 3 sin    Pa  On this clay bed a layer of geocells was formed
with different pocket sizes and geogrids. After the
The different parameters in the above equation were formation of the geocell layer, pockets of the geo-
determined as described by Duncan and Chang cells were filled with soil. The soil within each cell
was compacted uniformly by tamping with a steel
(1970). It was found that the parameter “n” was
rod having an enlarged base. The unit weight of in-
nearly equal to 0.70 for all the samples and the mod- fill soil was maintained at 17 kN/m3 for all the tests.
ulus parameter “K” was found to depend on the The compaction quality of this layer was verified by
number of geocells and the modulus of the geotex- testing undisturbed core samples collected from at
tile used to fabricate the geocell. The Kr of the geo- least six individual cells. Above the geocell layer, a
symmetrical half of the embankment was construct- the interfaces between different materials were
ed using clayey sand in 50 mm lifts. The number of simulated using 4-node joint elements. Typical finite
blows on each layer was adjusted to achieve a uni- element mesh consisted of about 1500 nodal points,
form average unit weight of 19 kN/m3 in all the 2900 3-node triangular elements and 50 4-node joint
tests. elements at the interfaces.
The embankment was of height 400 mm with The average Young’s modulus (E) of the clayey
crest width of 700 mm. The side slope was 2H:1V. sand in embankment was obtained as 15,000 kPa
The surcharge pressure on embankment crest was from the slopes of the unloading part of the triaxial
applied through a pre-calibrated proving ring using a compression tests performed at different confining
hydraulic jack supported against a reaction frame. pressures. The cohesion and friction angle of this
The uniform distribution of the surcharge pressure soil were 12 kPa and 34 respectively. The Poisson’s
was achieved by applying the loads through rigid I- ratio () of this soil was assumed as 0.30 which is a
sections running over the full width of the test tank. reasonable value for this type of soil. Because of the
A thick steel plate and an expanded polystyrene soft consistency of the foundation soil layer, no tri-
sheet were placed between the I-sections and the axial tests were possible on that soil. Hence, its
embankment crest to ensure uniform load distribu- Young’s modulus (E) value could only be back-
tion. analysed from the results of tests on unreinforced
The tests were performed with two different embankments. By trial and error the E and  values
pocket sizes with equivalent diameters (d) of 113 were obtained as 200 kPa and 0.45 by matching the
and 226 mm. The heights of geocell layers (h) were experimental data. The predicted surface heave was
varied from 100 mm to 250 mm in increments of 50 quite sensitive to the  value of the foundation soil.
mm. The aspect ratio (h/d) of geocell pockets in the A value of 0.49 has given rise to excessive heave
tests varied from 0.44 to 2.21. In all the tests, the ge- predictions and hence its value was reduced to 0.45
ocell layer was formed over the full length of the test by trial and error. The cohesive strength of soft clay
tank except in one test in which it was truncated at bed from vane shear tests was obtained as 10 kPa.
the toe of the embankment. Both the embankment soil and the foundation soil
The load was applied in small increments. Each were modeled as elastic-perfectly plastic with Mohr-
load increment was kept constant until the displace- Coulomb flow rule. The dilation angle was assumed
ments under that load increment reached steady as zero.
state. Typically the steady state was obtained within The geocell layer was modeled using the hyper-
10 minutes after applying the load. The vertical and bolic model as described earlier. The modulus pa-
lateral deformations in the model embankment were rameter of the unreinforced soil (Ku) was found to be
measured using dial gauges placed at different loca- 382. The same for geocell encased soil were esti-
tions of the test tank. The details of the embankment mated from the earlier equations and are reported in
tests are reported by Krishnaswamy et al. (2000). Table 1.

Table 1 Young’s modulus parameters for different


6 FINITE ELEMENT ANALYSIS OF MODEL geocell layers
EMBANKMENT TESTS Type of M Kr equivalent Apparent
geocell diameter cohesion
The finite element analysis of the above test em- layer of pocket (kPa)
bankments were performed using plane-strain ideali- (m)
sation. The geocell layer was represented using hy- UX 267 870 0.226 51
perbolic model using the properties estimated from BX 183 842 0.226 42
the previously described model. The incremental fi- BX 183 842 0.113 74
nite element equation was formed with out-of-
balance forces on the RHS and the tangent stiffness NP-1 75 781 0.226 24
matrix on the LHS to simulate the full non-linear NP-2 95 796 0.226 29
behavior of soil as closely as possible.
The soft foundation soil and the embankment The shear strength properties of the interfaces were
soil were modelled using elastic-perfectly plastic determined from large-scale direct shear box tests
type models. The geocell layer was simulated using between the different materials. The shear stiffness
hyperbolic model with equivalent shear strength and was reduced to a small value after the shear stress
stiffness values. The dilation angles for all the soils exceeds the shear strength of the interface. The
were assumed to be zero. In the finite element pro- normal stiffness was kept constant at 106 kN/m/m/m
cedure, the excess stresses beyond the yield surface so as to maintain the continuity at the interface.
were corrected back along the flow direction and The uniform settlement at the crest of the em-
hence it is possible to simulate the dilation behavior bankment was simulated in the analysis by applying
of the soil also, Rajagopal and Bathurst (1995). All equal vertical displacement to all the nodes on the
crest of the embankment. The pressure developed ments within that region of mesh. The different as-
corresponding to the applied displacement was de- pect ratios of geocell pockets were modeled by using
termined as the ratio between the sum of the reaction different apparent cohesion values for the layer as il-
forces and the embankment crest width. lustrated in Table 2. The excellent finite element
predictions for different pocket diameters for BX
geocells are shown in Figure 4.
7 RESULTS surcharge pressure
surcharge (kPa)
pressure (kPa)
0 20 40 60 80 100 120 140
The typical comparison between the measured and 0
predicted lateral deformations in unreinforced and
reinforced embankment (height = 100 mm with BX
geocells made of BX grid) slopes is shown in Fig-
ures 2 and 3. The H1, H2 and H3 are the measure-

lateral deformation (mm)


ments at three levels of the embankment. 5
pocket size
surcharge pressure (kPa) 113 mm

0 20 40 60 80
0
10 pocket-size
226 mm
5
H1
lateral deformation (mm)

10 unreinforced
15
15 H2 Fig. 4. Effect of pocket size on embankment behav-
experimental ior
20
It is interesting to note from the above figure that the
equivalent model developed is able to account for
25 finite element the effect of the diameter of the geocell pocket on
predicted
the overall response i.e. the effect of the aspect ratio
30 can also be accounted for using this equivalent mod-
H3
el.
35
Fig. 2. Lateral deformations in unreinforced slope 8 CONCLUSIONS
surchargepressure
surcharge pressure (kPa)
(kPa)
0 20 40 60 80 100 The equivalent model developed here is able to rep-
0 licate the 3-dimensional behavior of geocells using
2-d FE models. Further verification with other test
data is being performed.
2

9 REFERENCES
lateral deformation (mm)

4
Duncan, J.M. and Chang, C.Y., ‘Non-linear analysis
of stresses and strains in soils’, J. of Soil Mech.
6 and Foundns. Div., ASCE, 1629-1653, 1970.
Krishnaswamy, N.R., Rajagopal, K. and Madhavi
Latha, G., ‘Model Studies on Geocell Supported
8 experimental Embankments Constructed Over a Soft Clay
Foundation’, Geotech. Testing J., ASTM, 23, No.
2, 45-54,2000.
10 finite element Rajagopal, K. and Bathurst, R.J., ‘Behaviour of geo-
predicted synthetic reinforced soil retaining walls using the
finite element method’, Computers and Geotech.,
12 17, 279-299, 1995.
Fig. 3. Lateral deformations in reinforced slope Rajagopal, K., Krishnaswamy, N.R. and Madhavi
Latha, G., ‘Behaviour of sand confined in single
The different thicknesses of the geocell layers was and multiple geocells’, Geotextiles and Geomem-
modeled by assigning the relevant properties to ele- branes, 17, 171-184, 1999.

View publication stats

You might also like