Behavior of Flexible Pavements on Expansive Soils

Adel Djellali,1 Abdelhafid Ounis2, Behrooz Saghafi3,

Received: 2012.08.22

Accepted: 2012.12.11

Abstract
This investigation discusses the findings on analysis results of a pavement structure constructed on a relatively expansive material consisting mainly of brown clayey silt in Tebessa, Algeria. Cores were taken to laboratory from a
piece of roadwork. Remolded samples were taken from subgrade of a deteriorated road. The soils were generally
found to be of medium plasticity and of a medium to high expansion potential. Oedometer test for free swell indicated
a constant volume pressure that produced stress in excess of 350 kN/m2. The following involves description of the
behavioral characteristics of expansive soil as it relates to flexible pavements. Computer code Plaxis 8.2 was used for
numerical simulation to check the behavior of structure pavement with three models in pavement structure, and free
expansion test was performed to calibrate the soil subgrade with Soft-Soil Model. The results showed that the combined model Mohr-Coulomb in the pavement structure and the Soft-Soil in subgrade accord perfectly with deformations taken on the existing pavement. To stabilize the swelling character of subgrade, an adequate surcharge pressure
is used according to the procedure described in the paper. It showed that after 9 months, the deformations have been
stabilized at -0.007 m without any cracks in the pavement.

Keywords: Flexible pavements, expansive subgrades, soil calibration, Oedometer test, soil behavior, finite element
method

Corresponding author: E-mail: adel1830@yahoo.fr

1- Civil Engineering and Hydraulic Department, University of Mohamed Khider, Biskra, Algeria.
2- Civil Engineering and Hydraulic Department, University of Mohamed Khider, Biskra, Algeria.
3- Tarahan Parseh Transportation Research Institute,Tehran, Iran.

International Journal of Transportation Engineering,

Vol.1/ No.1/ Winter 2012

Behavior of Flexible Pavements on Expansive Soils

1. Introduction

stresses can be created when volume change of a material occurs. The stresses reflect in the form of cracking, heaving and settlement of highway pavements.
Therefore, a significant increase in the costs of routine
maintenance, rehabilitation and even reconstruction of
the deteriorated pavements will be forced to the road
authorities [Hyunwook and William, 2009]. The cracking phenomenon can occur through volumetric changes
under changing moisture conditions in expansive subgrades. These volumetric deformations usually result
in differential movements of flexible pavements resting
on the expansive subgrade. Consequently, structural
damages could happen if no special measures have already been taken during the design process [Ayman,
2007].
The method used for pavement design in Algeria is
known as the catalog structure and is based on the
French Method that uses the elastic Burmisters model
for a multi-layer, semi-infinite structure. It assumes a
semi-analytical, stress-based field where deformations
are calculated for when pavement is subjected to very
heavy traffic. However, in the case of flexible pavement over expansive soils, subjected to high gradients
of volume change, the method does not take into account such an effect in predicting pavement behavior.
Literature reveals that several locations in Tebessa, Algeria, are made of expansive soils causing pavement
deteriorations. As a part of road network maintenance,
rehabilitation methods have been developed specifically for the flexible pavements of a National Highway
(N10), which has suffered from severe degradation in
its structural integrity. The geotechnical records of N10
show that it was constructed on expansive subgrades.
Totally, 31 soil samples were taken, 10 of which were
cored and 21 were from wells. The results of laboratory test classified these soils of having medium to high
expansion potentials. Plaxis 8.2 software package with
its linear elastic, Mohr Coulomb and soft soil models
were used to predict soil variations and the surcharge
pressure was used to stabilize the heave of the pavement structure.

Pavements on expansive soils are common in many

parts of the world, particularly in semi-arid regions.
Such soils are generally unsaturated and contain a
large amount of clay. Expansion of the clayey soils,
containing smectites or illites in varying quantities,
is the root cause of numerous distresses in buildings
and large structures [Baheddi et al., 2007]. In the presence of water after a dry state, these soils undergo a
significant volume change. The volume change of expansive soils is primarily due to the hydration of the
clay minerals or more precisely, the adsorption of water
molecules to the exterior and interior surfaces of clay
mineral to balance the inherent change deficiency of
the particle [Snethen et al., 1975]. Conversely, if the
soil is parched, it undergoes a reduction in volume,
i.e. shrinkage takes place, which leads to the development of network cracks of polygonal shapes. These are
manifested by intense degradation of pavements and
surrounding grounds. The losses due to extensive damage to highways running over expansive soil subgrades
are estimated to be in billions of dollars all over the
world [Prasad et al., 2010; Jones and Holtz, 1973]. Various remedial measures like soil replacement [Snethen
et al., 1975], pre-wetting [SubbaRao and Satyadas,
1980], moisture control [Marienfeld and Baker, 1999]
and lime stabilization [Thompson and Robnett, 1976]
have been practiced with varying degrees of success.
However, these techniques suffer from certain limitations with respect to their adaptability , like longer time
periods required for pre-wetting the highly plastic clays
[Steinberg, 1977; Felt, 1953], difficulties in constructing the ideal moisture barriers [Snethen et al., 1975],
pulverization and mixing problems in case of lime stabilization [Ramana Murty, 1998], high cost for hauling
suitable re-fill materials for soil replacement purposes
[Chen, 1988; Snethen et al., 1975], geogrids reinforcement [Gupta et al., 2008], polymer grid reinforcement
[Miura et al., 2003], geosynthetic reinforcement [Zornberg and Gupta, 2009] and polymer grid [Miura et al.,
2003].
The problem is further exacerbated when the subgrade is expansible. Even if the pavement is correctly
designed, the swelling character of the subgrade goes
to distort all predictions. It is well known that larger
International Journal of Transportation Engineering,
Vol.1/ No.1/ Winter 2012

2. Location of the Study Area

Flexible pavement on expansive soils is generally common in Tebessa, Algeria. The city consists of a col-

Adel Djellali, Abdelhafid Ounis, Behrooz Saghafi

lapsed basin surrounded by mountains, with an average
altitude of 800 to 1600 m above mean sea level. The
city is bounded from north by the city of Souk-Ahras,
from south by ElOued, from east by the Tunisian border and from west by two cities Oum El-Bouaghi and
Khenchella, with an area of 21,000 km2. The 4.5 km
long highway commences at the intersection of N10
and N82 (El Kouif Road) and ends at the intersection
of the N16 (El Malabiod road) and road of Bekkaria,
with an average altitude between 814 and 842 meters
above mean sea level (Figure 1).

3. Classification and Soil Profile at the Test

Site

LL and plastic limit PL showed that these soils were

inorganic clay with medium to high plasticity. Another classification with [Dakshanamanthy and Raman,
1973] based on same parameter of Casagrande chart
for the expansion potential showed that the expansion
potential of the soils was medium to high (Figure 2).
In addition to the routine characterization testing, the
X-ray diffraction (XRD) technique was used to obtain semi-quantitative mineralogical composition and
chemical analysis [Saad and Aiban, 2006]. PANalytical
XPert PROX-Ray Diffractometer (Type MPD) was
used to obtain the data. These XRD micrographs vividly confirmed that soils were marly clays with 64% of
calcite and 35% of aluminosilicate (Figure 3).

As a part of enhancing the mechanical behavior of Na4. Calibration of Oedometer Test for Free
tional Highway N10, a soil-testing program was set up
Swell
to test 10 trials of the samples cored 6 m deep and 21
on expansive
is generally common
in Tebessa,forAlgeria.
wells of 2 to Flexible
3 m deep pavement
using a shovel.
The purpose soils One-dimensional
tests are instrumentals
predictingThe
city
consists
of
a
collapsed
basin
surrounded
by
mountains,
with
an
average
altitude
of 800 to
was to establish the geological profile of the site and
the compressibility, collapse and expansion potential
1600that
m above
mean
seaintact
level.andThe
city is bounded
by the
city
of Souk-Ahras,
to ensure
there were
enough
disturbed
of soilsfrom
[Saad north
and Aiban,
2006].
A geotechnical
invessamples
forsouth
laboratory
testing. Visual
analyses
tigation
company
permission
extracting
from
by El'Oued,
from
east ofbywells
the Tunisian
border
andissued
fromthewest
by twoofcities
OumEland Bouaghi
core samples
revealed
the presence
of marly
clay,of 21,000
undisturbed
samples
depths
of 0.3 commences
3.0 m and
4.5from
km the
long
highway
and
Khenchella,
with
an area
km2. The
clayey
silt and
marl. Tableof
1 shows
the geotechnical
3.0 6.0
m for
oedometer
for free swell.of the N16
at the
intersection
N10 and
N82 (El Kouif road)
and
ends
at thetest
intersection
characteristics
of
the
samples.
Classification
by
Casa(El Malabiod road) and road of Bekkaria, with an average altitude between 814 and 842
Test results were used in a calibrated soft soil model
grande
chart (Atterberg
Limits)
liquid limit
meters
above mean
sea based
levelon(Fig.1).

DETERIORATED
ROAD

FigureFigure
1. Satellite
view of Tebessa
indication
the modeledofarea
Google
Map
Service) map).
1.Satellite
view ofwith
Tebessa
withofindication
the(After:
modelled
area
(Google
International Journal of Transportation Engineering,
3. Classification and soil profile at3the test
site
Vol.1/ No.1/ Winter 2012

As a part of enhancing the mechanical behavior of National Highway N10, a soil

testing program was set up to test 10 trials of the samples cored 6 m deep and 21 wells of 2 to

Behavior of Flexible Pavements on Expansive Soils

Table 1. Basic geotechnical and swelling characteristics of the samples.

Table
swelling characteristics
of the
the samples
samples.
Table 1.
1. Basic
Basic geotechnical
geotechnical and
and swelling
characteristics of
Sample
Sample
N
N

Depth (m)
Depth (m)

elements
elements
<0.08mm
<0.08mm
%
%

1
12
23
3
4
4
5
56
67
78
89
9
10
10
11
11
12
12
13
13
14
14
15
15
16
16
17
17
18
18
19
19
20
20
21
21
22
22
23
23
24
24
25
25
26
26
27
27
28
28
29
29
30
30
31
31

1.18-2.00
1.18-2.00
1.30-2.00
1.30-2.00
0.25-2.00
0.25-2.00
0.30-2.00
0.30-2.00
0.30-2.10
0.30-2.10
0.30-2.00
0.30-2.00
0.20-2.00
0.20-2.00
0.20-3.00
0.20-3.00
0.30-3.00
0.30-3.00
0.20-1.50
0.20-1.50
0.70-2.30
0.70-2.30
0.25-1.50
0.25-1.50
0.25-2.00
0.25-2.00
0.30-2.00
0.30-2.00
0.30-2.00
0.30-2.00
0.30-3.00
0.30-3.00
0.20-3.00
0.20-3.00
0.40-2.00
0.40-2.00
0.40-3.00
0.40-3.00
0.70-1.40
0.70-1.40
0.68-1.00
0.68-1.00
0.50-6.00
0.50-6.00
0.70-2.50
0.70-2.50
0.50-3.00
0.50-3.00
0.40-4.00
0.40-4.00
0.50-5.00
0.50-5.00
0.60-5.50
0.60-5.50
0.40-2.40
0.40-2.40
0.40-3.00
0.40-3.00
0.50-5.00
0.50-5.00
0.70-6.00
0.70-6.00

93.4
93.4
92.4
92.4
92.4
92.4
98
98
92
92
93.6
93.6
97.6
97.6
97.6
97.6
97.2
97.2
96.4
96.4
97.6
97.6
95.6
95.6
97.2
97.2
96.8
96.8
97.4
97.4
97.2
97.2
96.8
96.8
96.8
96.8
96.6
96.6
45.18
45.18
98.25
98.25
91.6
91.6
97.8
97.8
98
98
93.8
93.8
92.4
92.4
93.4
93.4
92.6
92.6
93.8
93.8
98.6
98.6
98.6
98.6

Moisture
Moisture
content,
content,
W%
W%

Dry
Dry
density
d
density
kN/m33d
kN/m

Wet
Wet
density
h
density
kN/m33h
kN/m

Liquid
Liquid
limit,
limit,
LL%
LL%

Plasticity
Plasticity
index,
index,
PI%
PI%

MB
3
MB
cm
/g
cm3/g

Caco3
Caco3
%
%

Swell
Swell
pressure
pressure
kN/m2
kN/m2

19.36
1.56
1.82
63
38
7.3
48.38
19.36
1.56
1.82
63
38
7.3
48.38
-19.26
1.56
1.84
64
36
7.8
51.38
19.26
1.56
1.84
64
36
7.8
51.38
-18.36
1.56
1.86
64
37
7.8
48.36
18.36
1.56
1.86
64
37
7.8
48.36
14.26
1.52
1.82
62
36
7.2
49.38
14.26
1.52
1.82
62
36
7.2
49.38
12.89
1.70
1.86
52
32
7.1
59.38
12.89
1.70
1.86
52
32
7.1
59.38
-8.37
1.49
1.64
34
16
2.0
73.55
8.37
1.49
1.64
34
16
2.0
73.55
-12.82
1.71
1.88
50
30
7.1
60.26
12.82
1.71
1.88
50
30
7.1
60.26
-12.71
1.70
1.89
51
31
7.0
60.00
12.71
1.70
1.89
51
31
7.0
60.00
-12.84
1.72
1.87
48
32
6.9
59.38
12.84
1.72
1.87
48
32
6.9
59.38
-12.6
1.70
1.86
52
32
7.18
46.28
12.6
1.70
1.86
52
32
7.18
46.28
-18.33
1.47
1.75
65
39
7.17
46.15
addition
theroutine
routine
characterization
testing,
theX-ray
X-ray
diffraction(X
(X
InInaddition
totothe
characterization
testing,
the
diffraction
18.33
1.47
1.75
65
39
7.17
46.15
-12.76
1.70
1.86
50
30
7.2
59.36
12.76
1.70
1.86
50
30
7.2
59.36
technique
was
used
to
obtain
semi-quantitative
mineralogical
composition
and
technique
composition
and chc
12.36
1.71 was used
1.88 to obtain
51 semi-quantitative
32
7.0mineralogical
58.86
12.36
1.71
1.88
51
32
7.0
58.86
-12.38
1.72
1.90
50
31
7.26
59.42
analysis
(Saadand
and
Aiban,2006).
2006).We
We
haveused
used
PANalytical
XPert
PROX-Ray
analysis
(Saad
Aiban,
have
PANalytical
XPert
PROX-Ray
12.38
1.72
1.90
50
31
7.26
59.42
-13.34
1.72
1.91
52
33
6.82
64.36
Diffractometer
(Type
MPD)
to
obtain
the
data.
These
XRD
micrographs
vividlyc
Diffractometer
(Type
MPD)
to
obtain
the
data.
These
XRD
micrographs
vividly
13.34
1.72
1.91
52
33
6.82
64.36
12.89
1.73
1.99
58
36
6.84
47.26
12.89
1.73
1.99
58
36
6.84
47.26
that
soils
are
marly
clays
with
64%
of
calcite
and
35%
of
aluminosilicate
(Figs.3
15.85
1.70
1.98
59
38
6.83
45.38
that soils
are marly
clays with
64% of38calcite6.83
and 35%
of aluminosilicate
(Figs.
15.85
1.70
1.98
59
45.38
-14.25
1.72
1.96
58
37
6.26
46.34
14.25
1.72
1.96
58
37
6.26
46.34
-12.65
1.71
1.93
47
28
5.67
44.62
12.65
1.71
47
28
5.67
44.62
-14.5
1.72
11.93
.95
34
14
4.5
61.48
14.5
1.72
11.97
.95
34
14
4.5
61.48
-18.6
1.66
52
30
6.5
44.44
18.6
1.66
1.97
52
30
6.5
44.44
27.66
1.48
1.89
50
31
4.11
49.61
200
27.66
1.48
1.89
50
31
4.11
49.61
200
17.53
1.71
2.01
56
36
6.3
53.48
240
17.53
1.71
2.01
56
36
6.3
53.48
240
17.24
1.70
2.01
53
31
3.95
63.33
280
17.24
1.70
2.01
53
31
3.95
63.33
280
19.39
1.63
1.99
59
42
7.8
42.76
350
19.39
1.63
1.99
59
42
7.8
42.76
350
26.60
1.47
1.86
58
37
7.14
45.59
300
26.60
1.47
1.86
58
37
7.14
45.59
300
19.83
1.70
2.03
41
40
7.3
48.78
340
19.83
1.70
2.03
41
40
7.3
48.78
340
19.84
1.56
1.86
51
33
7.8
40.00
320
19.84
1.56
1.86
51
33
7.8
40.00
320
175
27.72
1.56
1.95
52
33
5.9
62.20
175
27.72
1.56
1.95
52
33
5.9
62.20
240
17.74
1.58
1.85
46
30
6.0
50.00
240
17.74
1.58
1.85
46
30
6.0
50.00
23.06
1.60
2.04
53
35
6.1
53.97
240
23.06
1.60
2.04
53
35
6.1
53.97
240

Figure3.(a).XRD
3.(a).XRDresult
resultofofsample
samplen15
n15(bleu:
(bleu:calcite,
calcite,red:
red:
Figure
silica,green:
green:alumina)
alumina)
silica,

Figure3.(b).
3.(b).XRD
XRDresult
resultofofsample
samplen2
n
Figure
silica,green:
green:alumina
alumi
silica,

Calibrationofofoedometer
oedometertest
testfor
forfree
freeswell
swell
4.4.Calibration

One-dimensionaltests
testsare
areinstrumentals
instrumentalsfor
forpredicting
predictingthe
thecompressibility
compressibilit
One-dimensional
andexpansion
expansionpotential
potentialofofsoils
soils(Saad
(Saadand
andAiban,
Aiban,2006).
2006).AAgeotechnical
geotechnicalinvestiga
investig
and
companyissued
issuedthe
thepermission
permissionofofextracting
extractingundisturbed
undisturbedsamples
samplesfrom
fromthe
thedepths
depth
company
3.0mmand
and3.0
3.06.0
6.0mmfor
foroedometer
oedometertest
testfor
forfree
freeswell.
swell.
3.0

Theresults
resultsofoftesting
testingwere
wereused
usedininaacalibrated
calibratedsoft
softsoil
soilmodel
modelprogrammed
programmedinincom
co
The
cellisissimulated
simulate
as Plaxis 8.2. With respect to the geometry of the test, oedometer cell
Figure
2.Classification
of subgrade
soils
of Tebessa,
based
Dakshanamanthy
and
Raman
(1973)
plane
strain
model based
by
elements
containing
15
nodes.
For
10trials
trials
swellingpres
pre
programmed in
computer
code Plaxis
8.2.
With
respect
was
toonreach
as
closely as
possible
toFor
the10
real
paramplane
strain
elements
containing
15
nodes.
ofofswelling
Figure 2.Classification of subgrade
soils ofmodel
Tebessa,by
on Dakshanamanthy
and
Raman
(1973)
intention
was
to
reach
as
closely
as
possible
to
the
real
parameters
seen
in
site.
to the geometry of the test, oedometer cell
is simulated
eters
seen in site.
Initial values
to the soil paintention
was to reach
as closely
as possible
to thegiven
real parameters
seen in site. InIn
33
33
3
3
=18.3kN/m
kN/m,, ,sat
=21.1kN/m
kN/mand
andpermeabili
permeabil
given
the15
soilparameters
weren=18.3
4parameters
sat=21.1
kN/m
and
given
totothe
soil
n=18.3
using a plane strain model by elements
containing
rameters were
were
kN/m
=21.1
4
m/day
in
undrained
conditions.
Shear
strength
parameters
were
taken
directly
fro
m/day
in undrained conditions.
strength
parameters
taken directly from
nodes. For 10 trials of swelling pressure,
the intention
permeability Shear
of 0.001
m/day in
no drained were
conditions.
2
2
=8and
test(direct
(directshear
sheartest)
test)which
whichreleased
released=8
andC=75
C=75kN/m
kN/m with
withaadilatancy
dilatancyang
an
test
Oedometer
parameters
are
taken
directly
from
the
oedometer
test
results,
the
mo
International Journal of Transportation Oedometer
Engineering, parameters

are taken directly from the oedometer test results, the mod
4

Vol.1/ No.1/ Winter 2012

compressionindex
index
Cc/2.3(1+e)=0.036and
andthe
themodified
modifiedswelling
swellingindex
index

compression
Cc/2.3(1+e)=0.036
(1+e)==0.035.
0.035.Void
Voidratio
ratiowas
wase=0.47.
e=0.47.
(1+e)
Figure 2.Classification of subgrade
soils8.2.
of Tebessa,
on Dakshanamanthy
andof
Raman
(1973)
as Plaxis
With based
respect
to the geometry
the test,
oedometer

In addition to the routine characterization testing, the X-ray diffraction (XRD)

technique was used to obtain semi-quantitative mineralogical composition and chemical
analysis (Saad and Aiban, 2006). We have used PANalytical XPert PROX-Ray
Diffractometer (Type MPD) to obtain the data. These XRD micrographs vividly confirmed
that soils are marly clays
with
64% of
calcite andOunis,
35% of
aluminosilicate
Adel
Djellali,
Abdelhafid
Behrooz
Saghafi (Figs. 3).

XRD result of sample n15 (bleu: calcite, red:

Figure 3.(b). XRD result of sample n29 (bleu: calcite, red:
silica,
green:
alumina)
silica, green: alumina)
3.(b). XRD result of sample n29 (bleu: calcite, red:
Figure 3(b). XRD result of sample
silica, green: alumina)
n29 (blue: calcite, red: silica,
libration of oedometer test for free swell
green: alumina)

ll

One-dimensional tests are instrumentals for predicting the compressibility, collapse

pansion
potential
of soils (Saad
and Aiban, 2006). A geotechnical investigation
dicting the
compressibility,
collapse
any
the permission
of extracting undisturbed samples from the depths of 0.3
). Aissued
geotechnical
investigation
andsamples
3.0 6.0
m for
testfor free swell.
ed
from
theoedometer
depths of 0.3

sults of testing were used in a calibrated soft soil model programmed in computer code
Figure 3.(a).XRD
result of sample
calcite, red:
Figure 3.(b).using
XRD a
result of sample n29 (bleu: calcite, red:
xis
8.2. With
respect
to
geometry
of then15
test,(bleu:
oedometer
cell is simulated
model
programmed
in the
computer
silica, code
green:
alumina)
silica, green: alumina)
4.
Calibration
of
oedometer
test
for
free
swell
Figure
3.(a).XRD
result of
n15 (bleu:
calcite,
Figure pressure,
3.(b). XRDthe
result of sample n29 (bleu: calcite, red:
strain model
by
containing
15 nodes.
For 10
trialsred:
of swelling
oedometer
cell
iselements
simulated
using
a sample
silica,
green:toalumina)
silica, green: alumina)
on
was
to
reach
as
closely
as
possible
the
real
parameters
seen
in
site.
Initial
values
strength
parameters
vealed a value of 240 kN/m2 instead of 340 kN/m2
or 10 trials ofShear
swelling
pressure,
the were taken directly from
3
3
kN/m
, satinstrumentals
=21.1
kN/m
of 0.001
to
the soil parameters
were
One-dimensional
for permeability
predicting
the
compressibility,
collapse andwere
n=18.3
the C.U
test
(direct
sheartests
test) are
which
released
=8and
obtained from
the
real test. Similar deformations
parameters
seen
in site.
Initial
values
3
4.
Calibration
of
oedometer
test
for
free
swell
in
undrained
conditions.
Shear
strength
parameters
were
taken
directly
from
the
C.U
21.1 kN/m and
ofwith
0.001
observed
in the first four
loading levels.
All these
andpermeability
C=75 kN/m2
a of
dilatancy
angle of
expansion
potential
soils [Saad
and=0.
Aiban, 2006].
A geotechnical
investigation
company
2
=8
irectwere
shear
test)
which
released
and
C=75
kN/m
with
a
dilatancy
angle
of
=0.
ers
taken
directly
from
the
C.U
Oedometer
p
arameters
are
taken
directly
from
the
oevalidate
that
the
free
expansion
test
overestimates
issued the
permission of tests
extracting
undisturbedfor
samples
fromthethe
depths of 0.3 collapse
3.0 m and the
One-dimensional
are instrumentals
predicting
compressibility,
2
meter
parameters
are
taken
directly
from
the
oedometer
test
results,
the
modified
test
results,
modified test
compression
index
expansion pressure in the case of medium to highly
dilatancy
angle
of the
=0.
kN/m with adometer
3.0
6.0
m for
oedometer
free and
swell.
andexpansion
potential
of soilsfor
(Saad
Aiban, 2006).
A geotechnical investigation
ession test
index
Cc/2.3(1+e)=0.036
and
2C
andthe
themodified
modified swelling
swell- index
expansive
soils,
since the free-swelling test can cause
ometer
results,
the modified
samples
company
g/2.3
issued
the
permission
of
extracting
undisturbed
from the depths of 0.3

=dified
0.035.
Void
ratio
was
e=0.47.
ingindex
index
(1+e)
=
0.035.
Void
ratio
was
changes
in
structure
expansion before it returns
swelling
2C
/2.3

g
3.0 mresults
and 3.0
6.0used
m for
testsoft
forsoil
freemodel
swell. programmedduring
Test
were
in oedometer
a calibrated
in computer
code Plaxis
e=0.47.
to zero strain state, where this state was also seen by
8.2.essential
With respect
tomodel.
the geometry
ofnumber
the test,ofoedometer
cellhave
is simulated
using a plane strain
ese parameters
forare
the
Athelarge
calculations
been
Allare
these
parameters
essential
forused
model.
A large soft
Bultel
[Bultel,
2011].
The
results
of
testing
were
in
a
calibrated
soil
model
programmed
computer
code
model
byofelements
containing
15 nodes.
For 10and
trials
of bounds.
swellingThe
pressure,inthe
intention
was
ed by
changing
the
value
the
parameters
between
their
lower
upper
arge
number
of
calculations
have
been
number
of
calculations
have
been
followed
by
changas Plaxis 8.2. With respect to the geometry of the test, oedometer cell is simulated using a
to
reach
closely
as elements
possible
to the
real
parameters
seen
in site.
values
giventheto the
havetheir
been
summarized
in
Table
2.The
een
lower
upper
bounds.
ing and
the
value
ofas
the
parameters
between their
lower
and15
plane
strain
model
by
containing
nodes.
For
10 trials
of Initial
swelling
pressure,
5. Numerical
Modeling
3
3
=18.3
kN/m
,
=21.1
kN/m
and
permeability
of
0.001
m/day
in no
soil
parameters
sat in to the5.1.
as
n closely
upper
bounds.
The
have
been summarized
intention
was results
towere
reach
as possible
realObjective
parameters seen in site. Initial values
3
3
kN/m were
, IIn
andfrom
permeability
given
toconditions.
the soil parameters
were n=18.3
Table
2.
drained
Shear strength
parameters
takentokN/m
directly
the C.U of
test0.001
sat=21.1
order
cope with
the complications
of(direct
describing
2the
Figure
4
shows
the
results
of
curves
of
expansion
pressure;
the
variation
of H/H versus
m/day
in
undrained
conditions.
Shear
strength
parameters
were
taken
directly
from
the
C.U
Figure
4
shows
the
results
of
the
curves
of
expansion
with
a dilatancy
angle of
=0.
shear test) which released =8 and C=75 the
kN/m
swelling
behavior
of the expansive
soil,
research2been compared with the numerical curve, which has been
log
(
).
The
real
curve
(RC)
has

pressure;
the variation
of H/H
log ( =8
). The and C=75
test (direct
test)
which
released
kN/m
with aalternative
dilatancy
angle
=0.
ers
developed
approaches
[Banu et al.,
Oedometer
parameters
areversus
taken
directly
from
thehave
oedometer
test
results,
theofmodified
shear
resulted
from
simulation.
The calibration
results the
allow
for a good simulation of both the
real curve
(RC) has
been compared
with
the
numerical
Oedometer
are
taken
directly
from
the
oedometer
test
results,
modified
parameters

elements aims
2009].
One
of
them
being
the
use
of
finite
andstiffness
the modified
2Cg/2.3
compression index Cc/2.3(1+e)=0.036
initial small strain
and of the swelling
large strain index
behavior. The
user-defined function of

curve,
which has been
resulted
from simulation. Theand the
compression
index
swelling
index 2C
Cc/2.3(1+e)=0.036
g/2.3
themodified
following
purposes:
Plaxis
was
used
in
order
to
set
the
pore
water
pressure
distribution
so that it simulates the
5 was
(1+e)
Void
e=0.47.of both
calibration
results
allow
for ratio
a good
simulation
(1+e) =
= 0.035.
0.035.
Void
ratio
was
e=0.47.

H/H

1.the
Tocalibration
get a deeper
understanding as to why permanent
expansion pressure for
process.
the initial small strain stiffness and of the large strain
deformations happen in different parts of flexible paveAll these
these
parameters
A
number
of
have
All
parameters
are essential
the
model.ments
A large
large
number
of calculations
calculations
havebeen
been
behavior.
The user-defined
function
of Plaxisfor
wasthe
usedmodel.
resting
on expansive
soils.
7.00
followed
by
changing
the
value
of
the
parameters
between
their
lower
and
upper
bounds.
The
in order
to set by
the changing
pore water the
pressure
so
followed
valuedistribution
of the parameters
their
lower
andforupper
bounds.
The
2.between
To find the
best
method
modelling
the stresses.
6.00
have
been
summarized
in
Table
2.
that results
itresults
simulates
the
expansion
pressure
for
the
calibrahave been summarized in Table
2.
Since the stress state depends on the soil model, several
5.00
tion process.
kinds of material models have been tested to estimate
4.00
NC
The numerical simulation of freeTable
expansion
test re- of parameter for calibration
2. Combination
3.00

Table 2. Combination of parameter for calibration.

RC

2.00

Parameter
Cc
Cg
C (kN/m2)

Upper
Lower
Number
Selected
bound 1.00
bound
of steps
value
0.00
0.100
0.800
70
0.290
1
10
100
1000
5
Log() kPa
0.020
0.100
45
0.072
of free swell test for the 45
sample at the depth of 3 82
m (RC: real curve; NC: numerical curve)
10 Figure 4. Simulation100
1
25
25
4.4
2

The numerical simulation of free expansion test revealed a value of 240 kN/m instead of 340
real test. Similar
deformations
were observed
in the first four
kN/m2 obtained from the
International
Journal
of Transportation
Engineering,
loading levels. All
test
overestimates
the expansion
5 these validate that the free expansion
Vol.1/ No.1/ Winter 2012
6 of medium to highly expansive soils,
pressure in the case
since the free-swelling test can
cause changes in structure during expansion before it returns to zero strain state, where this
state was also seen by Bultel [ Bultel, 2011].

(). The real curve (RC) has been compared with the numerical curve, which has been
resulted from simulation. The calibration results allow for a good simulation of both the
initial small strain stiffness and of the large strain behavior. The user-defined function of
Plaxis was used in order to set the pore water pressure distribution so that it simulates the
expansion pressure for the calibration process.
Behavior of Flexible Pavements on Expansive Soils

7.00
6.00

H/H

5.00
4.00

NC

3.00

RC

2.00
1.00
0.00
1

10

100
1000
Log() kPa
Figure
4.Simulation
of of
free
sampleatatthe
thedepth
depth
m (RC:
curve;
NC: numerical
Figure
4. Simulation
freeswell
swelltest
testfor
for the
the sample
of of
3 m3 (RC:
realreal
curve;
NC: numerical
curve) curve)

ment structure.
which one would give the most reliable stress distribuThe numerical simulation of free expansion test revealed a value of 240 kN/m2 instead
tion. This stress-strain
analysis
created
a
basis
platform
of 340 kN/m2 obtained from the real test. Similar deformations were observed in the first four
for future development of the models in permanent de5.2. Modeling
loading levels. All these validate that the free expansion
test overestimates the expansion
formation.
Modeling
was tothe
simulate
an existingtest
flexible
pressure in the case of medium to highly expansive soils, since
free-swelling
can pave3.To find the stabilization methods to overcome the cyment
structure
that
had
suffered
several
damages
cause changes in structure during expansion before it returns to the zero strain state, where afcling variations of expansive subgrade.
ter one year of its construction. This road had been
this state was also seen by Bultel (2001).
Plaxis version 8.2 was used to model the behavior for
constructed on an expansive subgrade classified as
this research. Pavement structures have rarely been
brown clayey silt. The total thickness of the pavement
5. Numerical modeling
analyzed with finite element programs under dynamic
was 0.76 m. Clayey silt subgrade was covered with a
5.1.One
Objective
loading.
reason for this is the fact that traffic load0.20 m layer of calcareous tufa as the improved subing is much more complicated than static loading norgrade, 0.20 m crushed gravel as subbase course, 0.20
to cope
with the
complications
ofmdescribing
the swelling behavior of the
mally appliedIninorder
geotechnical
problems.
Another
reason
crushed gravel as base course and 0.06 m asphalt on
expansive
soil,models
researchers
developed
approaches (Banu et al., 2009). One of
is that
the material
in finite have
element
programs alternative
top. The distresses observed on the pavement can be
them
being
the
use
of
finite
elements,
the
main
purpose
of finite element simulation is to:
have mainly been developed for static loadings not for
summarized as mild transverse and longitudinal cracks,
1. Get
a deeper
understanding
why permanent deformations happen in
repetitive cyclic loading.
Dynamic
analysis
needed to be as to mid-block
cracking, alligator cracking along the shouldifferent
parts
of
flexible
pavements
resting on expansive soils;
tested with a repetitive half-sin loading; but it was found
ders, consolidation rutting for 0.06 m deep and finally
2. Find
the best
to model the stresses. Since the stress state depends on
that the dynamic module
of Plaxis
8.2 method
was not suitable
average subsidence. A form of cracking can be seen in
the soil model, several kinds of material models have been tested to estimate
for modeling the traffic loading (Leena and Rainer,
Figure 5. Oedometer test for free swell showed that the
which soil model would give the most reliable stress distribution. This stress2004). Given the aforementioned reasons, the analysis
vertical deformation (uplift) has a maximum height of
strain analysis created a basis platform for future development
of models in
has to be simplified. The first part of the expansion pres0.021 m with 350 kN/m2 of expansion pressure. The
permanent
deformation;
sure analysis was to study the volumetric changes of the
groundwater level fluctuated between 0.00 and 0.50 m.
subgrade by the test oedometer for free swell. With this
No change in water pore pressure was considered for
6
analysis, the subgrade calibration was based on physical
hydraulic analyses.
and mechanical parameters such as cohesion (C), fricAt first, passage of just one truck was simulated. Then,
tion angle (), density and oedometer parameters of
loading from two trucks at the same time with dual
Cc, Cg and Pc. To approach as closely as possible to
wheel of 0.60m wide was studied. Modeling was done
the actual state, a large-scale numerical model, which
by a static ax-symmetric analysis and the element mesh
includes a flexible pavement structure based on an
consisted of triangular elements each with 15 nodes.
expansive subgrade, was established. The structure
The input parameters of the structure are shown in
was subjected to the loads from two trucks with dual
Table 3. To simulate the variation of stresses dependwheel load of 650 kN/m2 per axle. Then a complete
ent on the Youngs modulus, layers were divided into
analysis of stress-strain was performed on the pavesub-layers with the same strength parameters but with
International Journal of Transportation Engineering,
Vol.1/ No.1/ Winter 2012

summarized as mild transverse and longitudinal cracks, mid-block cracking, alligator

cracking along the shoulders, consolidation rutting for 0.06 m deep and finally average
subsidence. A form of cracking can be seen in Fig. 5. Oedometer test for free swell showed
that the vertical deformation (uplift) has a maximum height of 0.021 m with 350 kN/m2 of
expansion pressure. The groundwater level fluctuated between 0.00 and 0.50 m. No change in
water pore pressure was
considered
for hydraulic
analyses.
Adel
Djellali, Abdelhafid
Ounis,
Behrooz Saghafi

Figure 5.5.Cracking
Cracking distress
structure.
Figure
distressononpavement
pavement
structure.
static failure line. This assumption is not entirely valid
different elasticity moduli. Use of plain-strain analysis,
where the loading would have been continuous linear
7 in the case of traffic load, but it gives a good estimate
about sensitivity of materials for permanent deformaloading, could result in an overestimation of stresses
tion. The static failure criterion, where it is widely
and responses [Leena and Rainer, 2004]. To model the
adopted in geotechnical and in pavement materials, is
loading area of the dual wheel, the total load was transwith the failure criterion of Mohr-Coulomb [Leena and
ferred to a circular load with a known mean contact
widely2004].
adopted
in geotechnical
andresults
in pavement
materials, is
Rainer,
However,
the experimental
show
pressure. The detail of the structure and boundary conwidely adopted
in geotechnical
and in pavement
materials, i
Mohr-Coulomb
(Leenaofand
Rainer,
2004). However, the exp
that
the strength envelopes
almost
all geo-materials
ditions are illustrated in Figure 6.
Mohr-Coulomb (Leena and Rainer, 2004). However, the exp
strength
envelopes
of almost
geo-materials
have
the nature
of nonlinearity
in theall
n
stress space. have the natur
For this model, the attention was focused on the stressstrength
envelopes
ofcriterion
almost isalla geo-materials
have the natu
In
addition,
linear
failure
special
case of is a special c
es and resilient deformations. The modeling initiated
stress space. In addition, linear failure criterion
At first, passage of just one truck was simulated.
Then,
loading
from
two
trucks
at
the
stress space.[Lianheng
In addition,
failure
criterion
c
failure
et al.,linear
2010].
The behavior
of is a special
from values derived from laboratory tests. Results of
et al., criteria
2010). Thewas
behavior
of aembankment
connects
the failu
same time with dual wheel of 0.60 m wide was studied.
Modeling
done
by
static
axet al., 2010).
The behavior
embankment
connects the fail
embankment
thefailure
failure of
ratio
to thecan
deviatoric
measurements and calculations of resilient deformaratio,
in thisconnects
case
the
be written as follows
symmetric analysis and the element mesh consisted
of triangular
elements
eachratio
withR15
ratio,ratio,
in this
case
thethefailure
canbebe
written as follow
stress
in this
case
failure ratio
ratio RRcan
writtions were compared with each of other soil parameters
nodes. The input parameters of the structure are shown in Table 3. To simulate the variation
ten as follows:
and were modified to provide a distribution of stresses
of stresses dependant on the Young's modulus, layers were divided
; into sub-layers with the
and strains as close as possible to reality, i.e. the cali(1)
;
same strength parameters but with different elasticity modulus. Use
of plain-strain analysis,
brated soil that was mentioned previously.
where the loading would have been continuous linear loading, could result in an
The magnitude and development of permanent deforwhere
overestimation of stresses and responses (Leena and
Rainer, 2004). To and
model the loading
where
and
mations depends on static stress state of the material
area of the dual wheel, the total load was transferred to a circular load with a known mean
[lender, 2008], and how far the stress state is from the
R= failure ratio

contact pressure. The detail of the structure and boundary

conditions
are illustrated in Fig. 6.
R
the
failure ratio
R
the
failure
ratio
q of the model.
deviatoric stress, kN/m22
Table 3. Input parameters
q
deviatoric stress,
stress, when
kN/mp = 0
Table 3. Input parameters
model.
q0of the deviatoric
q0 Subbase
deviatoric Improved
stress,2 when pSubgrade
=0
Base course C
Cohesion,
kN/m
2tufa
Material
Asphalt crushed gravel CcrushedCohesion,
gravel subgrade
swelling
kN/m
M
the slope of the failure line clay
in p-q space
2in p-q space
M
the
slope
of
the
failure
line
p
hydrostatic pressure,
kN/m2000
Thickness, mm
60
200
200
200
2
p
hydrostatic
pressure,
kN/m108
2
Youngs Modulus, MN/m
5400
300200
14090
friction angle.70
0.35
friction angle.
Poissons ratio
0.35
0.35
0.25
0.35
Unit weight, kN/m3
25
21.2
22.0
20
18
Since the subgrade has a swelling
character, it was m
Cohesion, kN/m2
30
20Since the subgrade
9
10
-has a swelling
character, it was m
model 44
takes into account
43
36 the following
25 parameters: stress de
-Friction angle ()
model takes into account the following parameters: stress de
compression
behavior),
primary loading
13
14
6 distinction between
0
-Dilatation angle ()
compression
behavior),
distinction
between
primary loading
K0
1
0.32
0.30swelling behavior),
0.4
0.8 for pre-consolidation
with the
memory
s
with the swelling behavior), memory for pre-consolidation s
according
to the
Mohr-Coulomb
criterion
(Brinkgreve, 2002
International
Journal
of Transportation
Engineering,
according to the Mohr-Coulomb criterion (Brinkgreve, 2002
7 that
No.1/ Winter
2012 volumetric st
there is a logarithmic Vol.1/
relationship
between
that there is a logarithmic relationship between volumetric s
stress (p'), where the virgin compression can be formulated a
stress (p'), where the virgin compression can be formulated

Youngs Modulus, MN/m

5400
300200
Poissons ratio
0.35
0.35
Unit weight, kN/m3
25
21.2
Cohesion, kN/m2
30
-43
-Friction angle ()
13
-Dilatation angle ()
K0
0.32
Behavior of1Flexible Pavements

14090
0.35
22.0
20
44
14
on 0.30
Expansive

70
0.25
20
9
36
6
0.4
Soils

108
0.35
18
10
25
0
0.8

widely adopted in geotechnical and in pavement materials, is with the failure criterion of
Mohr-Coulomb (Leena and Rainer, 2004). However, the experimental results show that the
widely
adopted
in geotechnical
and
in pavement
materials,
is with
the failure criterion
of
otechnical and in
pavement
materials,
is withallthe
failure
criterion
ofthe nature
strength
envelopes
of almost
geo-materials
have
of nonlinearity
in the n
Mohr-Coulomb
(Leena
and
Rainer,
2004).
However,
the experimental
results
show
that the
na and Rainer, 2004).
However,
the experimental
results
show
that
the
stress
space.
In addition,
linear
failure
criterion
is
a special
case of failure
criteria
(Lianheng

strength
envelopes
of
almost
all
geo-materials
nonlinearity
in the nstress
f almost all geo-materials
have
nature
of
inconnects
thehave
nthe
et
al., 2010).
Thethe
behavior
ofnonlinearity
embankment
thenature
failureofratio
to the deviatoric
stress
space.
In
addition,
linear
failure
criterion
is aasspecial
case of failure criteria (Lianheng
ion, linear failure
criterion
a special
case
of
failure
criteria
(Lianheng
ratio,
in
thisiscase
the failure
ratio
R can
be written
follows:
et
al.,
2010).
The
behavior
of
embankment
connects
the
failure
ratio to the deviatoric stress
havior of embankment connects the failure ratio to the deviatoric stress
ratio,
this case
the failure ratio R can be written as follows:
failure ratio R can
beinwritten
; as follows:
(1)
where

and

(1)

(1)

FigureFigure
6. Numerical
model:model:
detailsdetails
and theand
boundary
conditions
6. Numerical
the boundary
conditions

2
qwhere
stress, kN/m
6. Analysis of the Results
and
R= deviatoric
the
failure
ratio
For
this
model,
the
attention
was focused on the stresses and resilient deformations. The
2
qq0= deviatoric
stress,
when
p
=
0
The modeling results have been presented in Figures
deviatoricinitiated
stress, from
kN/mvalues derived from the laboratory
modeling
tests. The results of measurements
2
C=
Cohesion,
kN/m
7 and
8. Three
material
models,
linear elastic model,
the failure
ratioof when
deviatoric
stress,
p = 0deformations were
qR0
io
and
calculations
the resilient
compared
with
each of
other soil
2 space
2 in p-q
2
M
=
the
slope
of
the
failure
line
Mohr-Coulomb
and
Soft-Soil
model,
were
applied.
parameters
theykN/m
were modified to provide a distribution of stresses and strains as
close
as Linq
deviatoric and
stress,
C
Cohesion,
kN/m
ess, kN/m
2
p
=
hydrostatic
pressure,
kN/m
model was previously.
the first model for calculations,
possible
the
reality,
i.e.line
the=calibrated
soil that ear
haselastic
been mentioned
q0
deviatoric
stress,
when
p
the
slopetoof
the
failure
in0 p-q space
ess, when p = 0M
2 development
2 of permanent deformations
2
=
friction
angle.
The
magnitude
and
depend
on
staticstructure
stress state
of
which
look
into
the
pavement
and subgrade
in
C
Cohesion, kN/m
p
hydrostatic
pressure, kN/m
/m
Since
the
subgrade
has
a
swelling
character,
it
was
its entirety.
In thethe
modeling,
the modulus
of each layer
the
2008),
how far
the stress
state is from
static failure
line. This
Mp-q space
the material
slopeangle.
of(lender,
the failure
lineand
in p-q
space
he failure line in
friction
2 in thetakes
2 simulated
by
Soft-Soil
model.
This
model
was
tested
as
described
earlier
and
then
fixed
to some
assumption
is
not
entirely
valid
case
of
traffic
load,
but
it
gives
a
good
estimate
about
hydrostatic pressure, kN/m
ressure, kN/m p
into
account
the
following
parameters:
stress
depractical
values.
Mohr-Coulomb
model
the same
sensitivity
of
materialshas
forapermanent
deformation.
failure
where
itused
is This
Since
theangle.
subgrade
swelling character,
it The
wasstatic
modeled
bycriterion,
Soft-Soil
model.

friction
.
pendent stiffness (logarithmic compression bedeformation
parameters
in
linear
elastic
model,
model takes into account the following parameters:
stress dependent stiffness (logarithmicalong the
8 subgrade
havior), distinction between primary loading and
and pavement structure. Finally, the pavement

and

Since behavior),
theitsubgrade
has aby
swelling
character,
itloading
was modeled
by Soft-Soil
This
grade has a swelling
character,
was modeled
Soft-Soil
model. This
compression
distinction
between
primary
and unloading
cyclemodel.
(compatible
unloading cycle (compatible with the swelling bestructure
was
modeled
by
Mohr-Coulomb
model
and
model
intostress
account
thememory
following
stress dependent
stiffness
(logarithmic
ount the following
parameters:
dependent
stiffness
(logarithmic
with
thetakes
swelling
behavior),
forparameters:
pre-consolidation
stress
and
failure
behavior
havior),
and
memory
for pre-consolidation
stress
subgrade with Soft-Soil model. The stress states of the
compression
behavior),
between
primary
loading
and
cycle
(compatible
r), distinction between
primary
loadingdistinction
and unloading
cycle
(compatible
according
the
Mohr-Coulomb
criterion
(Brinkgreve,
2002).
Theunloading
Soft-Soil
model
assumes
and
failure to
behavior
according to the
Mohr-Cousubgrade
in three
models
and the
analysis
of the strucwith
the
swelling
behavior),
memory
for
pre-consolidation
stress
and
failure
behavior
havior), memorylomb
for
pre-consolidation
stress
and
failure
behavior
criterion
2002]. The Soft-Soil
that there
is a [Brinkgreve,
logarithmic relationship
between volumetric
strain
(
)
and
mean
effective
ture, as illustrated inv Figure 7, indicate where the plastic,
according
the
criterion
(Brinkgreve,
2002).
assumes
hr-Coulomb criterion
(Brinkgreve,
2002).
Soft-Soil
model
model
assumes
thatMohr-Coulomb
there
is a The
logarithmic
relationtension-cut-off
andSoft-Soil
cap points model
are situated
in structure
stress
(p'), to
where
the
virgin
compression
can
beassumes
formulated
as: The
that
there isvolumetric
a logarithmic
relationship
between
volumetric
strain
(
)
and
mean
effective
hmic relationship
between
strain
(v) and
effective
and mean
mean
efship
between
volumetric
strain
v
with the same calculations.
fective
stress
(p), the
where
the virgin
compression
Figure 7 shows
the plastic points in subgrade and pavestresscan
(p'),
virgin
compression
can
be
formulated
as:
virgin compression
bewhere
formulated
as:
(2)
( )
can be formulated as:

ment according to the three models. For the linear elastic model, it shows no presence
of plastic points accord(2)
( )compression (2)
( )
where * is the modified
index and ing
is the
initial
volumetric
strain.
to the Mohr-Coulomb model. It is observed that the
where: * is the modified compression index and v0 is
subgrade
is largely affected by tensile cut-off points, in
*
During
unloading
and
reloading
a different
(line)
is
followed,strain.
which
can be
where
is the is
modified
compression
index
and upper
ispath
thepart
initial
volumetric
fied compression
index
and
the
initial
volumetric
strain.
the
initialisotropic
volumetric
strain.
of the
pavement and
shoulders, with the apformulated
as: unloading and reloading a different
During
isotropic
pearance of plastic Coulomb points in transition zone
During
isotropic
unloading
reloadingwhich
a different
oading and reloading
a
different
path
(line) and
is followed,
can bepath (line) is followed, which can be
path (line) is followed, which can be formulated as:
between the pavement and shoulder and in contact with
formulated as:
(3)
traffic loads. With combined
(3)model, the cap points are
( )
concentrated in the subgrade under the shoulder and
where: k**is the modified (
swelling
index [Brinkgreve,
(3) tensile stresses (tension
)
width of the pavement,
(3)
( )
where is the modified swelling index (Brinkgreve,middle
2002).
2002].
cut-off) in shoulders and in surface course, the plastic
(2)

where
*is theof
modified
swelling index (Brinkgreve, 2002).
ied swelling index
(Brinkgreve,
2002).
6.
Analysis
the results

e results

International Journal of Transportation Engineering,

8
Vol.1/
No.1/
Winter
The
modeling
6. Analysis
of2012
the results
resultshave been presented

in Fig. 7 and 8. Three material models were

applied, such as linear elastic model, Mohr-Coulomb and Soft-Soil model. Linear elastic
Thethe
modeling
been
presented
in Fig.
7 and
8. Three material
model
was
calculations,
which
look
into
the pavement
structuremodels
and were
g results have been
presented
infirst
Fig.model
7results
andfor
8. have
Three
material
models
were

Finally, the pavement structure was modeled by Mohr-Coulomb model and subgrade with
Soft-Soil model. The stress states of the subgrade in three models and the analysis of the
structure, as illustrated in Fig.7, indicate where the plastic, tension-cut-off and cup points are
Adelwith
Djellali,
Abdelhafid
Ounis, Behrooz Saghafi
situated in structure
the same
calculations.

Linear elastic model

Mohr-Coulomb model

10

Soft-Soil model
Figure 7. Plastic, tension-cut-off and Cap points
Figure 7. Plastic, tension-cut-off and Cap points

Fig. 7 shows the plastic points in subgrade and pavement according to the three
International Journal of Transportation Engineering,
models. For the linear elastic model, it shows
for Mohr9 no presence of plastic points,Vol.1/
No.1/ Winter 2012
Coulomb model. It is observed that the subgrade is largely affected by tensile cut-off points,
in upper part of the pavement and shoulders, with the appearance of plastic Coulomb points
in transition zone between the pavement and shoulder and in contact with traffic loads. For

Behavior of Flexible Pavements on Expansive Soils

Coulombs points in contact between wheel and pavement and in the transition zone between the pavement
and shoulder are clearly seen.
Figure 8 illustrates total displacements in the subgrade and pavement by three models after two cycles of loading. The Figure depicts that according
to the first two models, i.e. linear elastic model
where the total displacement is 1.65x10-3 m and
Mohr-Coulomb model with maximum displacements of 1.79x10 -3 m, displacements in the pavement occurred only in the contact between wheels
and surface course while they do not affect the subgrade. On the other hand, it was evidenced in the
combined model results that the movements had
concentrated largely in subgrade beneath the transi-

tion zone between the pavement and shoulders and

along the left side of the road where the pavement
is slightly deformed with maximum displacements
of 44.26x10-3 m.
Figure 9 represents the curves of stress path for subgrade and surface pavement. It reveals that the subgrade (of expansive clay) has elastoplastic hardening
behavior, and the pavement has an elastoplastic nonlinear behavior with a various moduli of deformation
along different parts of pavement.
Figure 10 represent the vertical displacement of subgrade and pavement structure. It shows that movements of swelling subgrade affect different parts of
pavement, where points A, B, C and D are located in
the pavement structure as shown in Figure 6.

a- Linear elastic model

b- Mohr-Coulomb model

c- Soft-Soil model
Figure
Figure8.8.Total
Totaldisplacements
displacements

International Journal of Transportation Engineering,

Fig. 8 illustrates total displacements in 10
the subgrade and pavement by three models
Vol.1/ No.1/ Winter 2012

after two cycles of loading. The Figure depicts that according to the first two models, i.e.
linear elastic where the total displacements are 1.65 x 10-3 and Mohr-Coulomb model with
maximum displacements of 1.79 x 10-3 m, displacements in the pavement occurred only in

Adel Djellali, Abdelhafid Ounis, Behrooz Saghafi

To overcome the distress of pavement over expansiv

heaves, the subgrade is loaded by a surcharge pressure equal
Figure 9. Stress path; left: subgrade, right: pavement
structure.is to make equilibrium between the pavement struc
principle
Figure
Stress
path;
left:
subgrade,
right:
pavement
structure.
Figure
9.9.
Stress
path;
left:
subgrade,
right:
pavement
structure.
eliminating
the
upward swelling pressure. This can be done
subgrade (Fig. 11). The surcharge pressure is designed withi
replacing a part of the current subgrade with a non-expansiv
embankment so much that the desired surcharge pressure is
equilibrium as a result of which the swelling pressure is elim
pavement will diminish. The subgrade soil is over-excavated
depth where the moisture content remains virtually constant
change of the soil to be negligible over time.

Figuredisplacement
10. Vertical displacement
pavement
Figure 10. Vertical
of pavement of
and
subgradeand subgrade

Pp

7. Stabilization by Surcharge Pressure

ment so much that the desired surcharge pressure is

Shoulders
Figure
Vertical
displacement
of obtained.
pavement
and
9 represents
the curves
of10.
stress
for subgrade
and surface
pavement,
and itequilibrium as a result
Thesubgrade
goal
is to obtain
OneFig.
of the
following methods
provides
thepath
calculated
which theand
swelling
pressure
reveals
that the
subgrade
(of
expansive
has
elasto
plastic
hardening
behavior,
and
theis eliminated
surcharge
pressure:
constructing
anclay)
inertof
embankFig. 9 (1)
represents
the curves
stress
path
forofsubgrade
surface
pavement,
and it and cyclic
variation
in
pavement
will
diminish.
pavement
an that
elasto
plastic
nonlinear
behavior
with
a
various
moduli
of
deformation
along
ment reveals
tohas
a calculated
height
and
(2)
replacing
the
exthe subgrade (of expansive clay) has elasto plastic hardening behavior, andThe
thesubgrade soil
is over-excavated
in theofseasonal
zone to a depth
different
of apavement.
Fig.
10 with
represents
the behavior
vertical displacement
subgrade
and active along
isting parts
soil to
calculated
depth
inert material.
pavement
has an elasto
plastic
nonlinear
with
a variousofmoduli
deformation
PS
where
the
moisture
content
remains
virtually
constant
Expansive
Subgrade
pavement
structure.
It
shows
that
movements
of
swelling
subgrade
affect
different
parts
of
Combination
of
the
two
methods
can
also
be
utilized
different parts of pavement. Fig. 10 represents the vertical displacement of subgrade and
over time,
leadingasthe
volume
change of the soil to be
as well.
pavement,
where points
A, B,ItCshows
and Dthat
are movements
located in the
structure
shown
in Fig.
pavement
structure.
of pavement
swelling
subgrade
affect
different
parts of
Figure
11. The principle of the stabilization o
negligible
over
time.
To
overcome
the
distress
of
pavement
over
expansive
6

pavement, where points A, B, C and D are located in the pavement structure as shown in Fig.

The swelling pressure was determined by the oedomsubgrades

6 and to limit their heaves, the subgrade is
Thetest,
swelling
was
determined
the oedometer test
7. Stabilization
by surcharge
pressure
eter
and afterpressure
simulation,
it output
the valueby
of 240
loaded by a surcharge
pressure equal
to the swelling
2
2
. The
must be the
the value
of 240 kN/m
. The surcharge
pressure
mustsurcharge
be the samepressure
in orkN/m
pressure.
The principle is toby
make
equilibriumpressure
between
7. Stabilization
surcharge
One of the following methods provides the calculatedder
surcharge
pressure:
(1)
settlements.
of surcharge
pressure
to avoid theCalculation
height settlements.
Calculation
of sur- was based on
the pavement structure and the subgrade by eliminatconstructing an inert
embankment
to amethods
calculated
height the
andcalculated
(2) replacing
the existing
soil (1)
to
One
of
the
following
provides
surcharge
pressure:
pressure
was based on athe
method introduced
ing the upward
pressure.
This
be done by of thecharge
(1954).
When
fill by
properties, non-t
a calculated
depth swelling
withaninert
material;
thecan
combination
two and
methods
cancomputing
also the
be existingsurcharge
constructing
inert
embankment
to a11).
calculated
height
(2) replacing
soilsurcharge
to
Terzaghi
(1954).
When
computing
the
fills
applying
a
surcharge
on
the
subgrade
(Figure
The
considered
with
the purpose
of stabilizing the swelling press
utilized. a calculated depth with inert material; the combination
of the two
methods
can
also be were considered
properties,
non-traffic
load
conditions
surcharge pressure is designed within a two-stage proutilized.
with the purpose of stabilizing the swelling pressure.
cedure, replacing a part of the current subgrade with a
(4)
non-expansive material and heightening the 13
embank13

11

International
Journal of Transportation Engineering,

Vol.1/ No.1/ Winter 2012

Then

HLseasonal active zone to a

pavement will diminish. The subgrade soil is over-excavated in the
depth where the moisture content remains virtually
PS constant over time leading the volume
Expansive Subgrade
change of the soil to be negligible over time.
PS
Expansive Subgrade
PS
Expansive Subgrade

Figure 11. The principle of the stabilization of an expansive subgrade

Behavior of Flexible Pavements onFigure
Expansive
Soils
11. The
principle of the stabilization of an expansive su
Figure
11. The principle
the stabilization
an after
expansive
subgrade it output
The swelling pressure was
determined
by theofoedometer
test, of
and
simulation,

The
swelling
pressure
by
Pp =toP
p was
S the oedometer
pressure
must
bePthe
samedetermined
in order
avoid
the height test, and after sim
the value of 240 kN/m2. The surcharge
2
The swelling pressure was determined
by
the
oedometer
test,
and
after
simulation,
output
be the same in order
the value of 240 kN/m . The surcharge pressureitmust
settlements. Calculation of
2 surcharge pressure was based on a method introduced by Terzaghi
pressure
must
be
the
same
in
order
to
avoid
the
height
the value of 240Shoulders
kN/m . The surcharge
settlements.
Calculation
of surcharge
pressure
was based on a method intr
(1954). When computing the surcharge
fill properties,
non-traffic
load conditions
was
settlements. Calculation of surcharge
pressure
was
based
on
a
method
introduced
by Terzaghi
H
L
(1954).
computing
considered with the purpose of stabilizing
theWhen
swelling
pressure.the surcharge fill properties, non-traffic load con
(1954). When computing the surcharge
fill properties,
conditions
was
considered
with thenon-traffic
purpose ofload
stabilizing
the swelling
pressure.
considered with the purpose of stabilizing the swelling pressure.
PS
(4)
Expansive Subgrade
(4)

Figure
11.11.
TheThe
principle
of theofstabilization
of an expansive
subgrade
(5)subgrade
Figure
principle
the stabilization
of an expansive

heave of the subgrade had been 0.023 m as the maxi

(5) it decreased

(5)
The
swelling
pressure
was
determined
by
the
oedometer
test, expansion.
and afterLater,
simulation,
it output
mum recorded
to -0.005 m
Then
2
Then
Then pressure must
at the be
endthe
of the
first in
month
of to
theavoid
treatment.
nine
same
order
theAfter
height
the value of 240 kN/m . The surcharge
Then
(6)

(6)
months,
almost stabilized
at -0.007 m
settlements.
Calculation of surcharge pressure was
basedtheonsettlements
a methodare
introduced
by Terzaghi
2

where PP is
the
applied
vertical
pressure
(kN/m
),
with
no
cracks
in
pavement.
Moreover,
after

(1954).
When computing the surcharge
fill properties,
non-traffic load conditions
was three years
2
3

(6)
),
HL
is
the
the
average
total
density
of
the
fill
(kN/m
average total
density
of the
fillm due the secwhere PPis the applied vertical pressure (kN/m ), a the
the maximum
settlement
reaches
-0.009
considered
with the purpose of stabilizing
the swelling
pressure.
2
3
applied
vertical
pressure
(kN/m
), in
total den
where
selected
(m) and fill
S isthickness
the area
of(m)
thePfill
(kN/m
), fill
HLthickness
is the selected
and
is
the
area
of
the
fill
soil
which
is
PisSthe
a the
ondary
consolidation
with
no
deformation
the average
pave2
3
is
the
applied
vertical
pressure
(kN/m
),

the
average
total
density
of
the
fill
where
P
(kN/m
),
H
is
the
selected
fill
thickness
(m)
and
S
is
the
area
of
the
fill so
P
a
soil
which
is
assumed
as
square
of
each
side
1
m.
L
assumed 3as square of each side 1 m.
ment structure.
(kN/m
),
H
is
the
selected
fill
thickness
(m)
and
S
is
the
area
of
the
fill
soil
which
is
assumed
L
Calculation resulted
in a total thickness of the
surchargeas square of each side 1 m.
(4)layer is
assumed
square
of aeach
side
1 including
m. of the
Calculation
resulted
in
total
thickness
layer of 1.1 m. The fill
layer
of 1.1 as
m.
The fill layer
is designed
0.40surcharge
7. Conclusions
Calculation
in a total
thickness
of the
surcharge
layer of 1.1 m. T
m layer ofincluding
ballast to limit
designed
0.40themcapillarity
layer of ascension,
ballast
to0.10
limit resulted
theThis
capillarity
ascension,
0.10
m layer
of behavior
study is focused
on determination
of the

(5)
Calculation
resulted
in
a
total
thickness
of
the
surcharge
layer
of
1.1
m.
The
fill
layer
is
designed
0.40
mpavements
layer oftufa
ballast
toinfluence
limit the
capillarity ascension
m layersand
of crushed
sand as anti-contamination
layer,including
crushed
as anti-contamination
layer,
0.40m
of layer
of calcareous
as the
embankment,
of flexible
under
of the evodesigned
including
0.40
m
layer
of
ballast
to
limit
the
capillarity
ascension,
0.10
m
layer
of
startedgravel
intufa
Julyasas
2007,
when
the0.20
precipitation
isasphalt
rarecharacter
andon
thethe
temperature
0.40m
mlayer
of operation
layer
calcareous
embankment,
crushed
sand
as lutionary
anti-contamination
layer,
0.40m
of
layer Algeof calcareous tuf
of top.
swelling
soils
of to
Tebessa,
0.20
ofofcrushed
base
course
and 0.06m
Thereaches
earthmoving
Then
sand The
asgravel
anti-contamination
layer,
0.40m
of
layer
of
calcareous
tufa
as
embankment,
purpose
wascourse
to reduce
the variation
of
moisture
content.
During
the
construction
mcrushed
layer of45C.
crushed
as base
and0.20
0.06
m
asm layer of crushed
gravel as
base course
and 0.06m
asphalt
on the to
ria. The following
conclusions
are drawn
based on
the
of the
fill
and
after gravel
five years
of
service,
the
variation
of
heave
of
pavement
versus
the
time
phalt
top.of
The
earthmoving
operation
started
in
0.20on
mthe
layer
crushed
as
base
course
and
0.06m
asphalt
on
the
top.
The
earthmoving
14
analysis results:

(6)
beentheplotted
in Fig. 12.
Theand
Figure
shows that,1.prior
the fill, soils
the heave
the subgrade
July 2007,has
when
precipitation
is rare
the tem14 are marly clays
Theto
subgrade
of theofstudy
area
had
been
0.023
m
as
the
maximum
recorded
expansion.
Later,
it
decreases
to
-0.005
m at the
perature reaches to 45C. The purpose was to reduce the 14 2 with 64% of calcite and 35% of alumino-silicate
havthefirst
applied
vertical
pressureAfter
(kN/m
), a the the
average
totalare
density
where PPis
of the
month
of the
settlements
almostof the fill
variation ofend
content.
During
thetreatment.
construction
of nine months,
3moisture
ing
medium
to
high
Atterberg
Limits,
medium
to
high
(kN/m ), HL is atthe
selected
fillnothickness
(m) and S
is the area
the years
fill soil
is
-0.007
with
in pavement.
Moreover,
afterofthree
the which
maximum
the fill andstabilized
after five years
of m
service,
thecracks
variation
of
swell potential and high swelling pressure of about 350
assumed
as
square
of
each
side
1
m.
settlementversus
reaches
m due
secondary
with no deformation in the
heave of pavement
the-0.009
time has
beenthe
plotted
in consolidation
kN/m2.
pavement
structure.
Figure
12. The Figure
showsinthat,
priorthickness
to the fill, of
the the surcharge
2. The calculations
indicate
thatfill
with
linear
Calculation
resulted
a total
layer of 1.1
m. The
layer
is elastic

designed including 0.40 m layer of ballast to limit the capillarity ascension, 0.10 m layer of
0.025
crushed sand as anti-contamination
layer, 0.40m of layer of calcareous tufa as embankment,
0.02
0.20 m layer of crushed gravel as base course and 0.06m asphalt on the top. The earthmoving
Haeve (m)

0.015

14

0.01
0.005
0
-0.005
-0.01
-0.015

10

20

30

40

50

60

70

Time(month)

Fig 12. Variation of heave of pavement versus the time

Fig 12. Variation of heave of pavement versus the time

International Journal of Transportation Engineering,
7. Conclusions
Vol.1/ No.1/ Winter 2012

12

This study is focused on determination of the behavior of flexible pavements under

the influence of the evolutionary character of swelling soils of Tebessa, Algeria. Based on the
results reported in this study, we can concluded that the subgrade soils of the study area are

Adel Djellali, Abdelhafid Ounis, Behrooz Saghafi

model, the deformations concentrated exclusively in
the pavement, which limits the damage only in these
areas with zero plastic deformations.
3. The Mohr-Coulomb model reveals that the deformations are concentrated exclusively in the pavement
situated in direct contact with traffic loads, with the appearance of plastic tension cut-off points in the surface
course, where the cracks propagate in these areas due
to the shrinkage of the clayey soil.
4. The results of simulation accord perfectly with the
distresses checked on the existing pavement by the
combination of the models: Mohr-Coulomb in the
pavement structure and the Soft-Soil in subgrade.
5. The combined model shows that the majority of
displacements are in the shoulder side of the subgrade
where the alligator cracking could be seen. It indicates
that the source of displacements is in the transitional
zone between the pavement and its shoulder, with
consolidation rutting value of 44.26x10-3 m instead
6.0x10-3 m measured on pavement.
6. The subgrade has demonstrated a hardening elastoplastic behavior while the pavement has had a nonlinear elastoplastic behavior.
7. These observations accord fairly with Laroche
(1973), who perfectly described the behavior of flexible pavements over expansive soils.
8. Stabilization of cyclic movements of expansive subgrade is neutralized by applying surcharge equal to the
swelling pressure. This is done by applying surcharge
fill pressure equal to the swelling pressure.

55, pp. 169-184

- Banu, S., Mustafa, A., Mustafa, V. and Fikret, K.
(2009) Prediction of swelling pressures of expansive
soils using artificial neural networks, Elsevier Scientific Publishing, Advances in Engineering Software, Vol.
41, pp. 647655
- Brinkgreve, R.B.J. (2002) PLAXIS Users manual version 8.2, Delft University of Technology and
PLAXIS b.v, Netherlands
- Bultel, F. (2001) Prise en compte du gonflement des
terrains pour le dimensionnement des revtements des
tunnels. Thse de Doctorat, cole Nationale des Ponts
et Chausses, Paris
- Chen, F. H. (1988) Foundations on expansive soils.
2nd ed. ELSEVIER Science Publishing, New York.
- Dakshanamanthy, V. and Raman, V. (1973) A simple
method of identifying an expansive soil, Japanese Society of Soil Mechanics and Foundation Engineering,
Soils and Foundations, Vol. 13, pp. 97-104
- Duhamel, D., Nguyen, V., Chabot, A. and Tamagny, P.
(2003) Modlisation de chausses viscolastiques. In:
16me Congrs Franais de Mcanique, Nice, pp. 1-5
- Felt, E. J. (1953) Influence of vegetation on soil
moisture content and resulting soil volume changes.
In: Proceedings of 3rd Conference on SMFE, Zurich,
Vol. 1

8. Acknowledgement
The authors gratefully acknowledge the support given
by the team of laboratory of public work of Tebessa for
their support and help to complete the laboratory tests
and data collection.

- Gupta, R., McCartney, J. S., Nogueira, C. L. and

Zornberg, J.G. (2008) Moisture migration in geogrid
reinforced expansive subgrades, In: The First Pan
American Geosynthetics Conference & Exhibition,
Cancun, Mexico, pp. 242-251

9. References
- Ayman, A. (2007) Numerical simulation of a trial
wall on expansive soil in Sudan. Plaxis Bulletin issue
21, Netherland. Available from : http://www.plaxis.nl/

- Hyunwook, K. and William, G. (2009) Finite element cohesive fracture modeling of airport pavements
at low temperatures, Elsevier Scientific Publishing,
Cold Regions Science and Technology, Vol. 57, pp.
123-130

- Baheddi, M., Djafarov, M. and Charif, A. (2007)

Design method of flexible continuous footing on
swelling clayey soils. Building Research Journal, Vol.

13

International Journal of Transportation Engineering,

Vol.1/ No.1/ Winter 2012

Behavior of Flexible Pavements on Expansive Soils

- Ramana Murty, V. (1998) Study on swell pressure
and the method of controlling swell of expansive soil,
PhD Thesis, University of Kakatiya, Warangal.

- Isik, Y. (2006) Indirect estimation of the swelling

percent and a new classification of soils depending on
liquid limit and cation exchange capacity, Elsevier
Scientific Publishing, Engineering Geology, Vol. 85,
pp. 295-301

- Saad, A. and Aiban. A. (2006) Compressibility and

swelling characteristics of Al-Khobar Palygorskite
Eastern Saudi Arabia, Elsevier Scientific Publishing,
Engineering Geology, Vol. 87, pp. 205-219

- Jones, D. E. and Holtz, W. G. (1973) Expansive soilsthe hidden disaster , Civil Engineering : American Society of Civil Engineers, Vol. 43, No. 8, pp. 49-51

- Snethen, D. R., Townsend, F. C., Johnson, L. D., Patrick, D. M. and Vedros, S. P. J. ( 1975) Review of engineering experiences with expansive soils in highway
subgrades, US Army Engineer Water Ways Experiment Station, FHWA, USA.

- Laroche, C. (1973) tude des sols pour la route Waza

Maltam, sur argile gonflante, Revue Gnrale des
Routes et Arodromes , No. 486, France.

- Steinberg, M. L. (1977) Ponding an expansive clay cut:

evaluations and zones of activity, TRR-641, TRB, pp. 61-66

- Leena, K. and Rainer, L. (2004) Modeling of the

stress state and deformations of APT tests. In: Proceeding of the 2nd International Conference on Accelerated Pavement Testing, Finland.

- Subaraob, K. S. and Satyadas, G. C. (1980) Prewetted and overloaded expansive soil, Proceedings of
Geotech. 80, Bombay, pp. 29-32

- Lianheng, Z., Liang, L., Feng, Y. and Jing, Z. (2010)

Evaluating the effects of the magnitude of pseudostatic force on reinforced slopes based on the nonlinear
M-C failure criterion, Electronic Journal of Geotechnical Engineering, Vol. 15, pp. 1017-1052

- Terzaghi, K. (1954) Anchored bulkheads, Transactions American Society of Civil Engineers, Vol. 119,
pp. 1243-1280.

- Marienfeld, M. L. and Baker, T. L. (1999) Paving fabric interlayer as a pavement moisture barrier,
Transportation Research Circular, E-C006, ISSN 00978515

- Thompson, M. R. and Robnett, Q. L. (1976) Pressure injected lime for treatment of swelling soils. In:
One of the 4 reports Prepared for the 54th Annual meeting of the TRB, TRR-568, pp. 24-34

- Miura, N., Sakai, A., Taesiri, Y., Yamanouchi, T. and

Yasuhara, K. (2003) Polymer grid reinforced pavement on soft clay grounds, Elsevier Scientific Publishing, Resources, Geotextiles and Geomembranes, Vol.
9, pp. 99-123

- lender, B. (2008) Prediction of permanent deformation of pavements unbounded layers based on cyclic
triaxial tests, American Journal of Applied Sciences,
Vol. 5, pp. 22-28
- Zornberg, J. G., Gupya, R. and Prozzi, J. (2009) Geogrid reinforcement of pavements over expansive clay
subgrades, In: Proceedings of the 17th International
Conference on Soil Mechanics and Geotechnical Engineering, Alexandria, Egypt, pp. 765- 768

- Prasad, D. S. V., Kumar, M. and Prasadaraju, G.V. R.

(2010) Behavior of reinforced sub bases on expansive
soil subgrade, Global Journal of Researches in Engineering, Vol. 10, pp. 225-235

International Journal of Transportation Engineering,

Vol.1/ No.1/ Winter 2012

14