Vous êtes sur la page 1sur 10

Downloaded from ascelibrary.org by Queen's University Libraries on 05/04/13. Copyright ASCE.

For personal use only; all rights reserved.

Geo-Congress 2013 ASCE 2013

333

Numerical Modeling of Wetting-Induced Settlement of Embankments


Yewei Zheng1, S.M. ASCE, Kianoosh Hatami2, M. ASCE, P.Eng. and Gerald A.
Miller3, M. ASCE, P.E.
1

Graduate Research Assistant, School of Civil Engineering and Environmental Science, University of

Oklahoma, 202 West Boyd Street, Room 334, Norman, OK 73019; zhengyewei@ou.edu
2

(Corresponding Author) Associate Professor, School of Civil Engineering and Environmental Science,

University of Oklahoma, 202 West Boyd Street, Room 334, Norman, OK 73019; kianoosh@ou.edu
3

Professor, School of Civil Engineering and Environmental Science, University of Oklahoma, 202 West

Boyd Street, Room 334, Norman, OK 73019; gamiller@ou.edu

ABSTRACT: Compacted fills of embankments are typically in an unsaturated state.


Water infiltration can weaken the embankments, resulting in excessive deformations
and/or embankment slope failure. In this study, the Barcelona Basic Model (BBM)
was implemented in the computer program FLAC (2D) to simulate the behavior of
unsaturated soils. The BBM implementation in FLAC was verified using analytical
results available in the literature. The verified BBM model in FLAC was then used to
simulate the wetting-induced response of an example unsaturated embankment due to
complete saturation as a worst-case scenario for the embankment design. It is shown
that the magnitudes of wetting-induced deformations in the embankment could result
in measurable amount of differential settlement in the embankment and even
embankment slope failure.
INTRODUCTION
Wetting-induced collapse settlement of embankments is a pervasive problem
around the world that leads to costly damages to infrastructure facilities including
roadways and bridges (e.g. the bridge approach slabs and parapet walls, distresses in
the pavement and embankment slope failure). Consequences of embankment collapse
settlement problems cost the taxpayers hundreds of millions of dollars annually in
repair and reconstruction of failed slopes and facilities.

Geo-Congress 2013

Downloaded from ascelibrary.org by Queen's University Libraries on 05/04/13. Copyright ASCE. For personal use only; all rights reserved.

Geo-Congress 2013 ASCE 2013

334

Lawton et al. (1992) determined that nearly all types of compacted soils can be
susceptible to collapse under proper conditions. The compacted fills of embankments
are typically in an unsaturated state. The wetting process of the embankment fill due to
water infiltration can occur as a result of rising of water table, rainfall and/or flooding,
resulting in wetting-induced deformations of embankments.
In this paper, the Barcelona Basic Model (BBM) is used as a constitutive model to
simulate the collapse behavior of unsaturated soils. The BBM is implemented in the
finite difference computer program FLAC (Itasca 2005) and is verified against
published analytical results. The wetting-induced response of an example unsaturated
embankment is then simulated using the verified BBM model in FLAC and the results
are discussed.
IMPLEMENTATION OF BARCELONA BASIC MODEL (BBM) IN FLAC
The Barcelona Basic Model (BBM) proposed by Alonso et al. (1990) is a
constitutive model for unsaturated soils. It is an extension of the Modified Cam-Clay
(MCC) model to include the influence of soil suction on its stiffness and strength.
The BBM includes two independent stress variables: the net normal stress (-ua)
and suction (ua-uw), where is the total stress, ua is the pore-air pressure, and uw is the
pore-water pressure. The implementation of BBM in FLAC is subject to certain
assumptions: the pore-air pressure is assumed to remain constant and equal to zero,
which is true for most practical engineering problems. Therefore, the net normal stress
is expressed as the total stress, and the soil suction is a variable that influences the soil
stiffness and strength.
In the BBM, both the net normal stress and suction can induce strain in the soil.
The incremental strains induced by the net normal stress are divided into two
components: elastic strain (d e) and plastic strain (d p). On the other hand,
incremental strains due to loss of suction (d s) are volumetric elastic strains which are
induced as a result of wetting. Hence, the total strain increment (d) in the soil can be
written as:

d = d e + d p + d s

(1)

Details of each strain component are described in the following sections.


Stress-Induced Elastic Strain
The volumetric and deviatoric elastic strain increments due to net stress are
respectively defined as:

Geo-Congress 2013

Downloaded from ascelibrary.org by Queen's University Libraries on 05/04/13. Copyright ASCE. For personal use only; all rights reserved.

Geo-Congress 2013 ASCE 2013

335

d ve =

1
dp
K

(2)

d qe =

1
dq
3G

(3)

where K and G are the soil bulk modulus and shear modulus, respectively.

Stress-Induced Plastic Strain


The yield surface in the p - q plane, f, is defined as:
f = q 2 M 2 ( p + ps )( p0 p ) = 0

(4)

where

ps = ks

(5)

p0
p0* [ (0) ]/[ ( s ) ]
=
(
)
pc
pc

(6)

(s) = (0)[(1 r ) exp( s) + r ]

(7)

In equations 4 through 7, p and q are net mean stress and deviatoric stress, respectively,
M is the slope of the critical state line, s is the matric suction, ps is the soil cohesion at
suction s, po is the preconsolidation pressure at suction s, p*o is the preconsolidation
pressure for saturated condition, pc is a reference pressure, (0) is the slope of the
normally consolidated line (NCL) for saturated condition, (s) is the slope of the NCL
at suction s, is the slope of the swelling line, and r and are parameters that control
the soil stiffness as a result of change in suction.
In the BBM used in this study, an associate flow rule was used to calculate the
plastic volumetric strain increment (dvp). Once the yielding occurs, the evolution of
the preconsolidation pressure is controlled by a hardening law defined as:
dp0*
1+ e
=
d vp
*
p0 (0)

(8)

Geo-Congress 2013

Geo-Congress 2013 ASCE 2013

336

Downloaded from ascelibrary.org by Queen's University Libraries on 05/04/13. Copyright ASCE. For personal use only; all rights reserved.

Suction-Induced Strain
In a wetting-induced deformation problem, the wetting path is in fact a suction
unloading process. At any point during the analysis, the current suction value is
assumed to be smaller than or equal to the maximum suction s0 ever experienced by
the soil in the past. Hence, the soil remains in the elastic region in the p - s plane during
the course of wetting. Rutqvist et al. (2011) determined the suction-induced strain by
adding an equivalent mean stress increment to the net mean stress. This method was
adopted in this study to determine suction-induced strains in the BBM.
The suction-induced volumetric strain is calculated as (Rutqvist et al. 2011):
d vs =

1
ds
Ks

(9)

where Ks is the equivalent suction bulk modulus defined as:

Ks =

(1 + e)( s + patm )

(10)

in which patm is the atmospheric pressure and s is a compressibility parameter related


to the suction-induced elastic strain. The equivalent mean stress increment is:
dp s = Kd vs =

K
ds
Ks

(11)

The equations for suction-induced volumetric strain were implemented in FLAC


through computational subroutines by adding the equivalent mean stress increment dp
s
to normal stresses xx, yy and zz.

VERIFICATION OF THE BBM IN FLAC


The BBM implementation in FLAC was verified against the data from the
published literature. In studying wetting-induced deformation problems, the wetting
paths are the primary focus of the analysis. Alonso et al. (1990) reported three
well-defined wetting paths AB (1), CD (2) and EF (3) in a BBM model prediction case
study (Case 1) as shown in Figure 1. An ideal reference soil was subjected to isotropic
compression in a triaxial stress space in Case 1. All three stress paths shown start from
Point A which represents an initial net mean stress p = 0.15 MPa and suction s = 0.2
MPa. The wetting processes take place at three different confining pressures: 0.15,
0.35 and 0.6 MPa, resulting in wetting-induced volumetric deformation in the soil.

Geo-Congress 2013

Geo-Congress 2013 ASCE 2013

337

0.25

Suction (MPa)

0.20

0.15

E
2

0.10
0.05
0.00

B
0.1

D
0.2

0.3

F
0.4

0.5

0.6

0.7

Net Mean Stress (MPa)

FIG. 1. Stress Paths in Case 1 Reported by Alonso et al. (1990)


2.0
B

Specific Voulme

Downloaded from ascelibrary.org by Queen's University Libraries on 05/04/13. Copyright ASCE. For personal use only; all rights reserved.

Model parameters of the reference soil as reported by Alonso et al. (1990) are
summarized in Table 1. An axisymmetric single element model was used to simulate
the triaxial tests on the soil. The isotropic compressive loading was simulated in a
strain-controlled condition in all directions. The wetting path was simulated by
decreasing the soil suction from 0.2 MPa to zero at constant net mean stress.

1.9

2,3

C
1

1.8

3
E

1,2

1.7

FLAC Results
Alonso et al. (1990)

1.6
0.1

0.2

0.3

0.4

0.5

0.6

0.7

Net Mean Stress (MPa)

FIG. 2. Comparison of Predicted FLAC Results with the Analytical Results


Reported by Alonso et al. (1990)
Figure 2 shows an overall satisfactory agreement between the predicted FLAC
results and Alonsos analytical data (BBM) with respect to both magnitudes and trends.
However, the FLAC results slightly overpredict the collapse volumetric deformation.
Results in Figure 2 also indicate that some swelling occurs as a result of wetting at low

Geo-Congress 2013

Geo-Congress 2013 ASCE 2013

338

stress levels whereas collapse occurs at higher stress levels.

Downloaded from ascelibrary.org by Queen's University Libraries on 05/04/13. Copyright ASCE. For personal use only; all rights reserved.

TABLE 1. BBM Model Parameters used in FLAC Simulations


p0*
Model
G

s
k
r
M
(MPa)
Parameters (MPa)
RSa
10
1
0.2 0.02
0.2
0.008 0.6 0.75
b
CK
3.3
0.82 0.14 0.015 0.055 0.01 1.24 0.26
a
RS: Reference soil (Alonso et al. 1990) used for model verification
b
CK: Compacted kaolin (Josa et al. 1988) used in numerical example

(MPa-1)
12.5
16.4

NUMERICAL EXAMPLE
After the implementation of the BBM in FLAC was verified, the model was used to
study wetting-induced deformations of an example embankment model. A compacted
kaolin soil with well-documented properties (Josa 1988) was selected as the
compacted fill of the example embankment. The compacted kaolin was low to
moderately plastic (wP = 26.9%, wL = 38.7%; USCS Classification: ML). The BBM
model parameters for the compacted kaolin were calibrated by Alonso et al. (1990) as
summarized in Table 1.

Model Grid, Boundary Conditions and Modeling Procedure


The example embankment was 10 m high with a 1V:2H slope. The foundation soil
was assumed to be the same as the embankment fill with a thickness of 5 m in a
saturated condition. Figure 3 shows the finite difference grid of the embankment
geometry in the example. The side boundaries of the model were fixed in the
horizontal direction but allowed for vertical displacement of the embankment. The
bottom boundary was fixed in both horizontal and vertical directions.
The embankment soil was modeled using the BBM with an initial suction value of
200 kPa for which the soil at the foundation level of the embankment was slightly
overconsolidated. A Mohr-Coulomb model was used for the foundation soil with a
saturated friction angle of 21.1. In the first stage of analysis, the foundation soil was
brought to equilibrium under gravitational stresses. The 10 m embankment fill with an
initial suction value of 200 kPa was then constructed on the foundation soil by 5 lifts
and the deformations of the entire model under gravity loading were calculated.
Finally, the wetting-induced deformation of the embankment was calculated by
decreasing the soil suction to zero to represent a completely saturated condition as the
worst-case scenario for the embankment design.

Geo-Congress 2013

Downloaded from ascelibrary.org by Queen's University Libraries on 05/04/13. Copyright ASCE. For personal use only; all rights reserved.

Geo-Congress 2013 ASCE 2013

339

FIG. 3. Finite Difference (FLAC) Grid of the Example


Embankment-Foundation Model
Results and Discussion
Figures 4 and 5 show displacement vectors in the embankment-foundation model
at equilibriums under self-weight loading and full saturation stages, respectively.
Under self-weight loading, the settlement at the top of the embankment at its centerline
is 51 mm. After the embankment becomes fully saturated, the settlement at the top
(centerline) reaches 154 mm. The wetting-induced deformation in this paper is defined
as the deformation due to wetting beyond the magnitude attained after self-weight
equilibrium. Hence, the predicted wetting-induced settlement at the top of the
embankment at its centerline is 103 mm. Results in Figure 5 indicate that not only the
magnitudes of the displacement vectors increase as a result of wetting, but also the
displacement vectors of the embankment slope tend to move outward. The horizontal
displacement at the toe of the embankment slope reaches 64 mm as compared to a
mere 10 mm under self-weight loading. Excessive lateral deformation of the slope
could result in slope failure. In the example shown in Figure 5, the wetting-induced
horizontal displacement of the embankment slope is fairly large, which indicates a
high probability of embankment slope failure.

FIG. 4. Displacement Vectors in the Embankment under Self-Weight Load (m)

Geo-Congress 2013

Downloaded from ascelibrary.org by Queen's University Libraries on 05/04/13. Copyright ASCE. For personal use only; all rights reserved.

Geo-Congress 2013 ASCE 2013

340

FIG. 5. Displacement Vectors in the Embankment after Full Saturation (m)


Figure 6 shows velocity vectors and a contour map of shear strain rates within the
embankment following full saturation. The velocity vectors in the embankment also
indicate the tendency of the slope to move outward. The magnitudes of shear strain
rate within the embankment show a coalescing potential slip surface, which together
with a zone of larger strains at the top of the embankment, indicates that the
embankment slope is on the verge of failure.

FIG. 6. Shear Strain Rate and Velocity Vectors in the Embankment after Full
Saturation
Figure 7 shows the settlement of the embankment at the top at different stages of the
analysis. The embankment surface at the center settles approximately 50 mm under
self-weight loading. After the embankment becomes fully saturated, the settlement
reaches approximately 150 mm. The magnitude of the wetting-induced settlement of
the embankment surface is hence approximately 100 mm. Results in Figure 7 also
indicate that after complete saturation of the embankment, the settlement of the
embankment surface at its centerline is approximately 50 mm larger than that at the
edge, resulting in a measurable amount of differential settlement at the top of the
embankment fill. This amount of differential settlement at the top of the embankment
could lead to significant damages to the structures and pavements constructed on the
embankment.

Geo-Congress 2013

Geo-Congress 2013 ASCE 2013

341

Distance to Embankment Centerline (m)


0

10

Settlement (mm)

100
150
200
Self-Weight
Wetting-Induced
Full Saturation

250
300

FIG. 7. Settlement of the Embankment Surface


10

Embankment Height (m)

Downloaded from ascelibrary.org by Queen's University Libraries on 05/04/13. Copyright ASCE. For personal use only; all rights reserved.

50

8
6
4
Self-Weight
Wetting-Induced
Full Saturation

2
0
0

100

200

300

400

Settlement (mm)

FIG. 8. Settlement of the Embankment Centerline along Height


Figure 8 shows the settlement of the embankment centerline along its height at
different stages of the analysis. These results suggest that the maximum values of the
fully-saturated and self-weight settlement in the embankment both occur at
approximately 6 m above the foundation level. In addition, the magnitude of
wetting-induced settlement increases significantly over the bottom 6 m of the
embankment but stays essentially constant along the remaining top portion of the
embankment (only 5 mm of reduction in magnitude over the top 4 m as compared to
the maximum value). A typical characteristic of unsaturated soils when subjected to
wetting is that they exhibit swelling behavior at low stress levels and collapse at high
stress levels. The magnitude of the net mean stress at 6 m above the foundation level
approximately equals to the preconsolidation pressure of the soil in saturated condition

Geo-Congress 2013

Geo-Congress 2013 ASCE 2013

342

as shown in Table 1. In the elastic region, the soil experiences swelling with relatively
small deformations whereas the collapse settlement occurs in the soil at higher
overburden pressures after yielding occurs.

Downloaded from ascelibrary.org by Queen's University Libraries on 05/04/13. Copyright ASCE. For personal use only; all rights reserved.

CONCLUSIONS
The Barcelona Basic Model (BBM) describing the elasto-plastic behavior of
unsaturated soils was implemented in FLAC and was verified against analytical results
reported in the literature. The implemented BBM in FLAC showed a promising
capability for use in studying the behavior of unsaturated soils subjected to wetting.
The verified BBM model was used to study the wetting-induced response of an
example embankment constructed using a clayey soil. A fully saturated condition was
assumed as a worst-case scenario for embankment design. The results showed that the
top surface of a 10 m-high embankment could settle by approximately 100 mm after
full saturation. In addition, wetting-induced deformation could also result in a
measurable amount of differential settlement in the embankment fill due to larger net
mean stresses toward the center of the embankment. This differential settlement could
lead to significant damage in the structures and pavements constructed on the
embankment. Furthermore, it was shown that excessive horizontal deformation at the
toe as a result of wetting could lead to slope failure in the embankment.

REFERENCES
Alonso, E. E., Gens, A. and Josa, A. (1990). A constitutive model for partially
saturated soils. Geotechnique, Vol. 40 (3): 405-430.
Itasca Consulting Group, Inc. (2005). Fast Lagrangian Analysis of Continua (2D)
Version 5.0. Itasca Consulting Group, Inc., Minneapolis.
Josa, A. (1988). Un modelo eslastoplastico para suelos no saturados. Ph.D. Thesis,
Universitat Politencnica de Catalunya, Barcelona.
Lawton, E. C., Fragaszy, R. J. and Hetherington, M. D. (1992). Review of
wetting-induced collapse in compacted soil. Journal of Geotechnical
Engineering, Vol. 118 (9): 1376-1394.
Rutqvist, J., Ijiri, Y. and Yamamoto, H. (2011). Implementation of the Barcelona
Basic Model into TOUGH-FLAC for simulations of the geomechanical behavior
of unsaturated soils. Computers & Geosciences, Vol. 37: 751-762.

Geo-Congress 2013