An evaluation of soft soil models based on trial embankments
[Link] & [Link]
Institute of Geotechnical Engineering, University of Stuttgart, Germany
[Link]
PLAXIS B.V., Delft, The Netherlands
ABSTRACT: Two constitutive models are introduced briefly and used for back-analysis of two test embankments in order to assess their performance. The Soft-Soil model based on the modified Cam-Clay model, is
employed as a reference model. The Soft-Soil-Creep model is an extension which includes time and strain
rate effects. It simply needs one additional input parameter compared to the Soft-Soil model. Therefore the
performance of these two models can be compared relatively easily. The first embankment considered is the
well-known Boston trial embankment. For this embankment, however, measurements of settlements and horizontal displacements have not been continued long enough to include secondary compression. Consequently,
for the second 2D FE-analysis an embankment with a long period of secondary settlements has been considered.
This is an embankment from the Ska Edeby test site in Sweden.
mary compression, it is clear that creep plays an important part in problems involving large primary settlements, i.e. soft soils. This is for instance the case
for the Boston and Ska Edeby trial embankment on
soft soils considered here.
1 INTRODUCTION
In recent years more and more constructions like
buildings and embankments have been built on soft
soils. Under soft soils the authors consider almost
normally consolidated clays, clayey silts and peats.
The special feature of these materials is their high
compressibility. Considering the tangential oedometer stiffness moduli at a reference vertical pressure of
, Janbu (1985) reports for normally consolidated clays to . For normally consolidated sands he determined values in the range of
to . Hence, in oedometer tests normally
consolidated clays behave ten times softer then normally consolidated sands.
Another feature of soft soils is the linear stressdependency of their oedometer stiffness. When plotting a stress-stiffness curve one gets a line of the form
. That behaviour leads to the well-known
logarithmic compression law
The construction of those embankments and their
subsequent behaviour were simulated involving the
deformation and pore-pressure. Two constitutive laws
are used to show the effects of taking creep into account or not. One is the Soft-Soil model (SS-model),
which is based on the modified Cam-Clay model,
using the Mohr-Coulomb (MC) criterion to describe
failure. The other, the Soft-Soil-Creep model (SSCmodel), is an extension to account for creep.
2 MODEL BASICS
The above mentioned two constitutive models are
briefly introduced in this section. They are both implemented in the FE-Code PLAXIS.
(1)
2.1 Soft-Soil model
where compression is regarded positive both for
stresses as well as strains. These considerations are
valid for normal consolidated stress states and do not
include secondary compression.
But all soils exhibit some creep. Hence primary
compression is always followed by a certain amount
of secondary compression. Assuming that the secondary compression is a small percentage of the pri-
The SS-model is based on the modified Cam-Clay
model. A logarithmic relationship between the volumetric strain and the mean effective stress
is assumed. Because of the usage of the volumetric strain
instead of the void ratio, Eq. (2) involves the modified compression index instead of the index as
used by Burland (1965). For virgin isotropic compres1
sion it yields
te
(2)
ppeq
2.2 Soft-Soil-Creep model
Buisman (1936) was probably the first to propose a
constitutive law for creep after observing that soft soil
settlements cannot be fully explained by classic consolidation theory. In the framework of 1D-secondary
compression other researchers like Bjerrum (1967),
and Garlanger (1972) should be mentioned. More
mathematical lines of research on 3D-creep were followed by, for example, Sekiguchi (1977), Adachi and
Oka (1982) and Borja and Kavaznjian (1985).
Butterfield (1979) proposes a creep equation of the
form
(4)
(8)
where the deformation during consolidation is given
by the expression
. The modified creep index
describes the secondary compression per logarithmic
time increment. In this equation is the logarithmic strain. The superscript is applied because the
logarithmic strain measurement was originally used
by Hencky. In case of large strains both Butterfield
(1979) and Den Haan (1994) showed that logarithmic
strain supersedes the traditional engineering strain.
Please note, that the time
is not the consolidation
time
. The time
is not a material parameter either
as it not only depends on the consolidation but also
on the geometry of the tested sample. Janbu (1969)
developed a construction for evaluating the parameter
and the time
from experimental data. This and
the difference between
and
is shown in Fig. 2.
(6)
This stress
is a function of the plastic strain.
1
ilu
b fa
m
o
ul
Figure 1. Yield surfaces of the SS-model in p-q-plane
(5)
)
(n
[Link] j
where
is related to the actual stress state and
is
the equivalent pre-consolidation stress, see Fig. 1.
ine
re l
(3)
The subscript ur is used to specify that the parameters are related to unloading and reloading. In the
SS-model and
are used as input parameters to
compute the elastic strains.
For triaxial stress states the yield function of the
SS-model is defined as
p eq
The parameter is the modified swelling index
which determines soil behaviour during unloading
and reloading. This behaviour is assumed to be elastic
and is described by Hookes law of elasticity. Eq. (3)
implies the following linear stress dependency of the
tangent stiffness modulus.
hr-C
Mo
ca
iti
Cr
For isotropic unloading/reloading situations the elastic volume strain is formulated as
ta
ls
)
S
(n C
e
in
(7)
The yield function (Eq. (5)) can be described as ellipses in the
- -plane. The tops of the ellipses are
located on a line with the inclination . In the modified Cam-Clay model (Burland 1965, 1967) the line represents the critical state line, which describes
the stress states at post peak failure. It should be
noted that in the SS-model the MC-criterion with the
strength parameters and is used to describe the
failure. Both the MC-line and the -line are given
the same shift of away from the origin. This
is illustrated in Fig. 1 and is taken into account in (Eq.
(6)).
The total yield contour shown in Fig. 1 by the bold
lines is the boundary of the elastic area. The MCfailure line is fixed, but the cap (ellipse with
) may
increase due to primary compression.
tc
ln t
ec
.
1/e
tc
.
(a)
tc
m*
t = tc - t
(b)
Figure 2. Consolidation and creep behaviour in standard Oedometer tests
The strain rate as derived from Eq. (8) reads
or inversely
depends entirely on the amount of the creep strain being accumulated by time. Eq. (10) can then be written
as
(9)
For isotropic stress states one finds the same modified
creep index . Therefore one can include the secondary compression part and combine Eqs. (2) and
(3) to get the total volumetric strain as
where
pp0
(10)
ev
(12)
(13)
to
or
(14)
is very small.
(15)
Having derived a expression for
it is now possible
to formulate the differential creep equation.
(16)
where
can be eliminated by means of Eq. (13)
to obtain
(17)
where
is defined as in Eq. (12). Entering Eq. (15)
into Eq. (17) one gets
evccr = (l -k ) ln (ppc/pp0 )
e = m ln (tc +t/tc )
cr
vac
With respect to the difference
So Eq. (14) can be simplified as
tained. Eq. (13) leads for
evc = k ln (p/p0 )
l
(11)
Assuming that in conventional testing the load is increased stepwise and each load step is maintained for
a constant period of
, where is precisely
one day, the so-called IC-line with
is then ob-
ln p
e
Combining Eqs. (10) with (11) one obtains
ppc pp p
where is the total volumetric strain due to an increase in mean effective stress from
to
in a time
period of
. The total volumetric strain is divided
into an elastic and a visco-plastic creep part, denoted
by and respectively. The visco-plastic part can be
separated into a part during consolidation and a part
after consolidation. This is given by the subscripts
and . The subscript has also been added to the
elastic strain to indicate that this is also related to consolidation. Fig. 3 clarifies these relations.
One should note that the isotropic consolidation
line (IC-line) is not reached after the end of consolidation, but after some creep has occurred. In fact the
IC-line is the line obtained while applying one load
step per day. This is based on the assumption that consolidation occurs in less than one day.
Eq. (10) is valid for a constant mean effective
stress, but for transient or continuous loading problems it is necessary to formulate a constitutive law
in differential form. Moreover there is still the question of how to express the time
analytically. Therefore the basic idea is adopted that all inelastic strains
are time dependent. Additionally following Bjerrum
(1967) it is supposed that the pre-consolidation stress
p0
(18)
Assuming the same yield function as in the SS-model
(Eq. (5)) the strain rate can then be defined as
IC-line
(1-day load step curve)
Figure 3. Logarithmic relationship between volumetric strain
and mean stress including creep
(19)
where
In Tab. 2 the other parameters for the sub-layers of
BBC are plotted. All parameters were determined according to the detailed subsoil description of Ladd,
Whittle and Legaspi (1994). The chosen ratio between the modified soil parameters is and
.
(20)
. Hence the SSC-model
and is defined as
is an extension of the SS-model that takes creep into
account.
DATA
3 ANALYSIS AND RESULTS
In this paragraph, the results of a back-analysis of
two embankments are presented. Note that for both
applied models the input parameters are always the
same. The only difference is that the modified creep
index is not used in the SS-model. Therefore creep
is not taken into account in this model.
Settlements [cm]
20
20
40
40
60
60
20
20
40
40
20
20
Till
Figure 4. Vertical displacements at different depths under the
embankment
DATA
SSC
SS
10
620 Days
0
-10
Elevation [m]
SS
Time [years]
Time [years]
3.1 Boston Trial Embankment
An extension of the Interstate highway I-95 north of
Boston was built in 1965. A portion of the embankment was instrumented with piezometers, settlement
rods and inclinometers. The embankment has a height
of with a crest width of and a footing
width of . It is founded on a thick sand-layer,
which has an underlying thick deposit of Boston
Blue Clay (BBC). The BBC layer is divided into
sub-layers with different compressibility parameters
and over-consolidation ratios (). Under the embankment the top peat layer is replaced by fill. The
parameters used for the calculation are given in Tab.
1 for sand, fill and peat respectively. The constitutive
law applied to the sand and fill is the Hardening Soil
model (Brinkgreve and Vermeer 1998). The peat was
computed with the MC-model. The following parameters are the same for all sub-layers of BBC.
SSC
Fill
-20
-30
-40
Table 1. Soil parameters for sand, fill and peat
sand
fill
peat
18
40
120
18.5
0.5
0.2
1
37
0
1
0.426
0.45
20
40
120
19
0.5
0.2
5
35
0
1
0.398
0.398
-165
-65
-165
-65
-165
Excess Pore Pressures (kPa)
11,8
0,208
0.3
5
25
0
1
0.577
0.577
10
2000 Days
0
-10
Elevation [m]
resp.
resp.
-65
-20
-30
-40
-65
-165
-65
-165
-65
-165
Excess Pore Pressures (kPa)
Figure 5. Pore-pressure distribution in different sections
layer
Table 2. Soil parameters for Boston Blue Clay
B1
B2
C1
C2
D
E1
E2
19
2
15.8
0.522
8.34
1.35
10.4
0.652
4.6 3.07
1.08 0.85
6.3
0.870
2.25 1.77
0.80 0.65
The inclination of the critical state line is significantly steeper than the MC-failure line to prevent the
overprediction of horizontal deformation. One gets
this inclination by choosing a -value, which yields
to !
. The overprediction of horizontal
displacements is well known in the framework of the
classical critical state mechanics. Moreover the value
for
is chosen to uncouple the elastic strain.
This is done to reduce the magnitude of the horizontal displacements as well.
The high -values (Tab. 2) indicate that creep
will not be so important because about half of the
BBC layer has an -value of at least two, decreasing at the bottom of the layer to a value near one. For
that type of soil the amount of creep will be small in
comparison to the rest of the settlements.
Figs. 4 to 6 show the calculation results and the
measurements. The solid lines indicate the measurements, the dashed lines are the results from the SSCmodel and the dotted lines are the analysis with the
SS-model. In Fig. 4 one can recognize that consolidation is not over after the measurement period of
days, because the settlement curves still have an inclination. That fact is supported by Fig. 5, one can see
that in the center of the BBC there is still a pore-waterpressure of around after days. The
SSC-model slightly overestimates in the most cases
the vertical displacements. Moreover the calculated
pore-water-pressures are slightly higher then those
measured. The horizontal displacements are overestimated as well. In most cases the SS-model matches
the measurements better, especially for the deep settlements. So it seems that if there are high values in the soft soils the SSC-model is not so useful.
7.1
1.304
1.44
0.60
E3
17.7
1.5
7.8
1.957
1.25 1.21 1.18
0.57 0.57 0.56
F1
F2
F3
5.9
1.522
1.16 1.14 1.11
0.56 0.56 0.55
ing parameters are for all 9 sub-layers the same.
The other parameters used for the calculations are
given in Tab. 3. All parameters were taken from the
literature (Larsson 1997). The ratio between the modified soil parameters is and .
The inclination of the critical state line is here chosen to match the !
-value correctly, as well.
DATA
SSC
SS
10
620 Days
0
Elevation [m]
-10
Sand
A
B
C
-20
D
E
-30
F
-40
Till
0,0
0,1
0,0
0,1
0,0
0,1
Horizontal Displacements (m)
DATA
SSC
SS
10
2000 Days
0
-10
Elevation [m]
3.2 Ska Edeby test embankment
The test embankment was built on the test site of Ska
Edeby about west of Stockholm in the year
1961. The embankment has a height of with a
crest width of and slopes . The fill was instrumented with piezometers and settlement markers
at different locations under and outside the fill. The
soil under the embankment consists of soft clay with
a thickness of on top of till or rock. The soft
clay layer is divided into sub-layers with different
compressibility parameters and s. The follow-
Sand
A
B
C
-20
D
E
-30
F
-40
Till
0,0
0,1
0,0
0,1
0,0
0,1
Horizontal Displacements (m)
Figure 6. Horizontal displacements in different sections
layer
A
B
C
D
E
F
G
H
Table 3. Soil parameters for Ska Edeby
14.2
14.4
16.1
14.1
2
1.2
0.106
0.091
0.083
Den Haan, E. J. (1994). Vertical Compression of Soils. Ph. D.
thesis, Delft University.
Garlanger, J. E. (1972). The consolidation of soils
exhibiting creep under constant effective stress.
Geotechnique 22(1), 7178.
Janbu, N. (1985). 25th rankine lecture: Soil models in offshore engineering. G e otechnique 35(3), 239281.
Ladd, C. C., A. J. Whittle, and D. E. Legaspi (1994). Stressdeformation behaviour of an embankement on boston
blue clay. In Vertical and Horizontal Deformations of
Foundations and Embankments Proceedings of Settlement 94, College Station Texas, pp. 17301759.
Larsson, R. (1997). Consolidation of soft soils. Report 29,
Swedish Geotechnical Institute.
Sekiguchi, H. (1977). Rheological characteristics of clays. In
Proceedings 9th International Conference on Soil Mechanics and Foundation Engineering, Volume 1, Tokyo,
pp. 289292.
Stolle, D. F. E., P. A. Vermeer, and P. G. Bonnier (1997).
A soft soil model and experiences with two integration
schemes. Numerical Models in Geomechanics, 123128.
Vermeer, P. A. and H. P. Neher (1999). A soft soil model that
accounts for creep. In R. B. J. Brinkgreve (Ed.), Proceedings of the International Symposium Beyond 2000 in
Computational Geotechnics, Amsterdam, pp. 249261.
Balkema.
Vermeer, P. A., D. F. E. Stolle, and P. G. Bonnier (1997). From
classical theory of secondary compression to modern
creep. In J.-[Link] (Ed.), Computer Methods and Advances in Geomechanics, Volume 4, Wuhan, pp. 2469
2478. Balkema.
0.076
1.0
0.069
As shown in Tab. 3 the soft clay is almost normally consolidated ( ). Only the upper
have a high value caused by a crust. Hence, a
large amount of creep is to be expected. Furthermore
it demonstrates that the settlements as well as the
pore-pressures and the horizontal displacements were
strongly underestimated by the SS-model whereas the
results from the SSC-model and the measured data
agree fairly well.
4 CONCLUSIONS
After a brief introduction of the SS-model and the
SSC-model two test embankments were used for
back-analysis. The results show that for overconsolidated soils the SSC-model has no advantages compared to the SS-model, because in such case creep
doesnt play an important role. However in normally
consolidated soft clays like in Ska Edeby creep becomes significant and the SSC-model exceeds the SSmodel. This should be kept in mind when simulating
the construction of any building on soft soils.
Time [years]
0,0
5,0
10,0
15,0
DATA
SSC
SS
20,0
-0,1
-0,3
Settlements [m]
-0,5
-0,7
REFERENCES
-0,9
Adachi, T. and F. Oka (1982). Constitutive equation for normally consolidated clays based on elasto-viscoplasticity.
Soils and Foundations 22, 5770.
-25,0 -20,0 -15,0 -10,0 -5,0
-0,02
0,0
Bjerrum, L. (1967). 7th rankine lecture: Engineering geology of norwegian normally-consolidated marine clays as
related to settlements of buildings. G e otechnique 17(1),
81118.
-4
-6
-8
-10
Burland, J. B. (1965). The yielding and dilation of clay (correspondence). Geotechnique 15(2), 211214.
-12
Burland, J. B. (1967). Deformation of Soft Clay. Ph. D. thesis,
Cambridge University.
-14
Elevation [m]
Brinkgreve, R. B. J. and P. A. Vermeer (Eds.) (1998). PLAXIS
Finite Element Code for Soil and Rock Analyses. Rotterdam: [Link].
Buismann, K. (1936). Results of long duration settlement
tests. In Proceedings 1st International Conference on
Soil Mechanics and Foundation Engineering, Volume 1,
Cambridge, Massachusetts, pp. 103107.
Excess Pore Pressure [kPa]
Butterfield, R. (1979). A natural compression law for
soils (an advance on e-log p)(technical note).
Geotechnique 29(4), 469480.
0,06
0,10
A
B
-2
Borja, R. I. and E. Kavaznjian (1985). A constitutive
model for the stress-strain-time behaviour of wet clays.
Geotechnique 35(3), 283298.
0,02
Horizontal Displacement [m]
Figure 7. Time-Settlement-Curve; Pore-pressure distribution after 10 years and horizontal displacements after 20 years
