Determination of parameters for HS and SS model for

Transdanubian clay
E. Koch
Széchenyi István University, Győr, Hungary,[email protected]

ABSTRACT: In recent years, in the field of geotechnical design, software based on FEM has come to the front. Ad-
vanced computer programs make it possible to use advanced soil models besides the most current elastic-plastic Mohr-
Coulomb model. By using these computer programs, nonlinear behavior of the soil can be described more realistic,
even in the case of more complicated load events. Among the constitutive models incorporated in the commercial soft-
ware the Hardening Soil Model (HS), Hardening Soil model with small-strain stiffness (HSsmall) and Soft Soil Model
(SS) are the most promising ones. Observations and experience gained in tunnel construction, deep open excavation or
preloaded embankment prove that with these soil models reality can be followed more accurately, especially in cases
where unloading and reloading are present.
In order to produce the software input parameters, in case of more complicated soil model, more demanding laboratory
tests are needed. In case of HS model one of the basic demands is to determine the power for stress-level dependency of
stiffness (m), tangent stiffness for primary oedometer loading ( Eoedref ) and unloading/reloading stiffness ( Eurref ). For
SS model determination of the modified compression index (λ*) and modified swelling index (κ*) are esential.
The paper focuses on the behavior of transdanubian clay common in Hungary. Sampling, laboratory investigations and
evaluation aimed to determine the input parameters for the HS and SS are described. Results based on a number of oe-
dometric tests accomplished with unloading and reloading proved to be adequate for the computational purposes. The
paper does not focus on the HSsmall model due to the difficulty of determination of the input parameters from oedome-
ter test.

Keywords: Clay, Hardening Soil Model, Soft Soil Model

established correlations for the computational model

1. Introduction parameters. The aim of this research was to determine
reliable relationships to serve as a design aid during the
Due to the increasing demands of construction preliminary phase of a project.
projects, civil engineers must cope with the challenges
of unloading/reloading behavior. The typical cases are 2. Investigated soil properties
deep excavations in an urban area or applying and
removing preload to an embankment. Movements The samples originated from different regions of
induced by excavations demand more accurate analysis Hungary. Liquid limit and plasticity index of the tested
of soil-structure interaction, especially in urban areas. In soils are plotted on the Casagrande chart (Fig.1).
such cases, strains and displacements cannot be descried The points are above the A-line and diverge from the
accurately by the conventional constitutive model: the A-line with increasing liquid limit values. The equation
elasto-plastic model with Mohr-Coulomb failure of the best-fit line is
criteria. Excavations and reloading can be analyzed only = 0,86 ∙ ( − 15,5) (1)
by a constitutive model capable of describing nonlinear The correlation coefficient is r = 0.95. Based on the
behavior and the hardening process [1]. Advanced Casagrande chart, the tested samples were mainly medi-
geotechnical software like PLAXIS or MIDAS now um and high plasticity clay but with some silt too.
make it possible to use more complex soil models. The 70
U line
two most promising new models are: the Hardening Soil 60
Model (HS); and the Soft Soil Model (SS). A line
Plasticity index IP [%]

50
Observations and experience gained in tunnel
construction, deep excavations, and preloaded 40
embankments show that these soil models describe field 30
behavior more accurately. This is especially true where
20
there are complex loading and unloading sequences in
construction [2]. 10
Geotechnicans using these computational tools often 0
face the problem of not having adequate or accurate 0 20 40 60 80 100
data for preliminary geotechnical estimations or Liquid limit wL ]%]
recommendations for structural design. To overcome Figure 1. Correlation of IP - wL
this difficulty, the laboratory at Széchenyi István Table 1. summarizes the main parameters of the tests.
University has performed approximately two hundred The plasticity index varies between 9 % and 60 % with
oedometer tests using unloading and reloading during an average value of 21 %. The consistency index is be-
the last few years. We have analyzed the test results and
tween 0.2 and 1.8, with an average value of 1.0. The Application of the HS model requires the parameters
average oedometer modulus due to primary compres- m, Eoedref modulus and Eurref modulus [3].
sion is 11 MPa, and the average elastic unload- In order to determine the input parameters, the stress-
ing/reloading modulus is 54 MPa. Note that the unload- strain behavior as a power curve as suggested by Janbu
ing/reloading is about five times the primary [4] was applied as indicated in Eq.(5)
compression. This value fits into the recommendations
=A (5)
of the PLAXIS manual which states that the quotient
should be between three and five. The collected data where z is the vertical specific strain, z is the
have been evaluated statistically and correlations have vertical stress, A is the scale factor relating stress and
been evaluated, and these results are presented below. strain, B is the power factor relating stress and strain,
and p equals 100 kPa. The same shape as presented in
Table 1. Statistical parameters of the soil properties
aver- devi- medi- the HS model was yielded by this formula. Taking the
parameters min max derivative of Eq.(5), the oedometric tangent modulus
age ation an
wo % 12.14 34.47 22.10 3.92 22.78 was determined as a function of z and is expressed in
eo - 0.48 0.88 0.64 0.07 0.64 Eq.(6)
Ip % 9.00 59.90 20.70 9.69 18.0
= ∙
(6)
Ic - 0.18 1.80 0.96 0.26 0.94
z MPa 3.20 74.80 25.10 16.27 21.30
Eoed MPa 3.80 31.70 10.90 4.69 10.20 3.1. Parameter analysis for HS model
Eur MPa 10.30 157.5 53.70 27.92 49.69
A - 0.007 0.114 0.031 0.02 0.03
Fig. 2. shows the relation between A and B.
According to Smoltczyk [5], parameter B should depend
B - 0.241 0.966 0.572 0.15 0.56
on the type of soil and parameter A should depend on
Cc - 0.032 0.191 0.106 0.03 0.105
the state of the soils. Considering our test results, there
Cs - 0.004 0.037 0.018 0.01 0.019
is a very strong relation between A and B parameters.
* - 0.009 0.044 0.028 0.007 0.028
Based on all the tests, the correlation can be descried by
* - 0.001 0.009 0.005 0.002 0.005 ,
= 0.012 ∙ (7)
The correlation coefficient is r = 0.84, implying good
3. Parameters of Hardening Soil Model correlation. Based on Fig. 2, parameter B slightly
changes on the plasticity lines. The compression curve
The Hardening Soil model (HS) is an advanced can be descried by these A and B parameters.
model for simulating the behavior of different types of 0,14
0.14

soil, both soft and hard. A basic feature of the present IP < 15 A = 0,014·B-1,25 r = 0,88
0,12
0.12
HS model is the stress dependency of soil stiffness. The
A [-]

15 < IP < 20 A = 0,013·B-1,06 r = 0,79

HS model uses moduli both from oedometer tests and 0.10

0,10 20 < IP < 30 A = 0,011·B-1,61 r = 0,85

from triaxial tests. Oedometer test results demonstrate IP > 30 A = 0,011·B-1,67 r = 0,84
0.08
0,08
the dependence of stiffness on confining stress. The all tests A = 0,012·B-1,36 r = 0,84

increase of the oedometric modulus depends on the 0,06

0.06
mean hardening stress. This is described by Eq.(2)
0,04
0.04
∙ ∙
= (2) 0,02
0.02
∙ ∙
0,00
0.00
where the Eoed is a tangent stiffness modulus obtained 0.0
0,0 0.2
0,2 0.4
0,4 0.6
0,6 0.8
0,8 1.0
1,0
B [-]
1.2
1,2

from an oedometer test. Eoedref is a tangent stiffness at a Figure 2. A and B parameters from oedometer test results

vertical stress of − = = . The reference Based on previous research, correlations between

ref
pressure, p is typically 100 kPa. The amount of stress parameter A or B and other simple soil properties were
dependency is given by the power m. further investigated but, unfortunately, no relations were
Under triaxial conditions, the parameter E50 is the found. Fig. 3 shows the relationship between plasticity
confining stress-dependent stiffness modulus for index and parameter B. One can see that parameter B
primary loading and is given by Eq.(3) varies between 0.4 and 0.8 but slightly increases with
∙ ∙
plasticity index (IP). The average value of parameter
= (3) B0.5-0.55.
∙ ∙
ref Fig. 4. shows the relationship between consistency
where E50 is a reference stiffness modulus
corresponding to the reference confining pressure pref. index and parameter A. It seems that value of parameter
For unloading and reloading stress paths, another A slightly decreases with consistensy index (IC), but
stress-dependent stiffness modulus is used: values of parameter A0.015-0.04. The average value of
∙ ∙ parameter A0.03.
= (4)
∙ ∙ Fig. 5. shows the relationship between primary
ref
where Eur is the reference Young’s modulus for loading modulus and depth of the sample. It can be seen
unloading and reloading, corresponding to the reference that modulus inreases with depth as usual. The
pressure pref. equations are given for different plasticity, but for all
the tests Eq.(8) describes the relation in MPa.
 = 0.25 ∙ + 4.60 (8) unloading/reloading is shown on Fig. 6. It can be
The slopes of the regression lines vary slightly with established that, as plasticity increases, modulus ratio
plasticity. decreases, but the differences are not great.
1,2
1.2 The unloading/reloading modulus is ~4-5 times
1,0
1.0 higher than the primary modulus with a correlation
B [-]

coefficient of r = 0.85. Considering all the test results,

0,8
0.8 the ratio between primary compression and
0.6
0,6
unloading/reloading compression can be expressed by
IP < 15 = 4.95 ∙ (9)
0.4
0,4 15 < IP < 20 The result agrees with the recommendation of
0.2
0,2 20 < IP < 30 PLAXIS manual that is 3 to 5. Note that a ratio of 5
IP > 30 was also back-calculated from field measuremnts on
0.0
0,0 clayey soils in the southern part of Hungary.
0 10 20 30 40 50 60 70
200
Plasticity index IP [-] IP < 15

Unloading / reloading modulus Eur [MPa]

Figure 3. Relationship between IP and B
15 < IP < 20
0,12
0.12 160 20 < IP < 30
IP < 15 IP > 30
0.10
0,10 15 < IP < 20 120 all tests

20 < IP < 30 Eur = 5,41 · Eoed r = 0,87

A [-]

0,08
0.08
IP > 30 80 Eur = 4,98 · Eoed r = 0,88
0,06
0.06
Eur = 4,93 · Eoed r = 0,78
40 Eur = 4,29 · Eoed r = 0,84
0,04
0.04
Eur = 4,95 · Eoed r = 0,85
0,02
0.02 0
0 10 20 30 40 50
Primary loading modulus Eoed MPa
0,00
0.00
0,00
0.0 0.25 0,50
0.5 0.75 1,00
1.0 1.25 1,50
1.5 1.75 2,00
2.0 Figure 6. Relationship between Eoed and Eur
Consistency index Ic [-]
Figure 4. Relationship between I c and A 4. Parameters of Soft Soil Model
Similar equations were found for Hungarian Loess by
The Soft Soil model is suitable for materials that
Varga [6], and for Frankfurt Clay by Katzenbach [7].
exhibit high degrees of compressibility, such as
Silty specimens (Ip<15) showed only a slight normally consolidated clays, clayey silts and peat.
dependence on depth to their maximum at 40m.
Primary loading modulus Eoed [MPa]
Based on this, volumetric hardening is the dominant
0 10 20 30 40 50 feature that should be considered in constitutive
0
modelling. Of course, a shear strength criterion is also
10
IP < 15 Eoed = 0,20  z + 5,07 r=0,82 needed for these geomaterials and a Mohr Coulomb
15 < IP < 20 Eoed = 0,29  z + 4,98 r=0,88 yield surface is considered for this purpose [3].
20
20 < IP < 30 Eoed = 0,24  z + 5,08 r=0,87 The volumetric mechanism that captures the
Depth z [m]

30 IP > 30 Eoed = 0,23  z + 4,12 r=0,90 compressibility of the material is simulated by an

all tests Eoed = 0,25  z + 4,60 r=0,86 elliptical cap that is very similar to the Modified Cam
40
Clay model. Some features of the Soft Soil (SS) model
50 are: a) stress-dependent stiffness (logarithmic
compression behaviour); b) distinction between primary
60
loading and unloading-reloading; c) memory for pre-
70 consolidation stress; d) failure behaviour according to
the Mohr-Coulomb criterion [3].
80
In the SS model, a logarithmic relation between the
Figure 5. Relation between depth and Eoed
volumetric strain v, and the mean effective stress p‘, is
Note that dependency on depth due to the effect of assumed. It can be formulated as Eq.(10)
mechanical state (effective stress and the volumetric ∗ ∙
− =− ∙ (10)
deformation from origin). For normally consolidated ∙
soils it usually corresponds to the depth, therefore it is a The minimum value of p’ is set equal to a unit stress.
suitable correlation and in practice we use E=f(z) type The parameter * is the modified compression index,
equations due to their simplicity. From a more scientific which determines the compressibility of the material in
point of view, the stress-strain relation should be primary loading. During isotropic unloading and
modeled by a nonlinear relation and the soil reloading a different path (line) is followed, which can
characterisation should be independent from the be expressed as
confining stress (ie depth) such as the modified ∗ ∙
− =− ∙ ∙
(11)
compression index * in theSoft Soil model. *
The relation between oedometer moduli due to The parameter  is the modified swelling index,
primary compression andmodulidue to which determines the compressibility of the material in
unloading and subsequent reloading. The relationship
between the unloading/reloading elastic modulus, 0,20
0.20
IP < 15
Poisson’s ratio and modified swelling index is 15 < IP < 20 Plasztikus index

characterized by the following relation 0.15

0,15 20 < IP < 30
IP

IP < 15

Compression index Cc [-]

∙ 15 < IP < 20

( )
= = ∗ (12) IP > 30 20 < IP < 30
IP > 30
all tests összes adat

where  ur is the Poisson’s ratio for unloading- 0.10

0,10
Cc = 0,382 · eo  0,141 r=0,81
reloading [3].
Cc = 0,386 · eo  0,137 r=0,78
This model is based on input parameters from 0.05
0,05 Cc = 0,236 · eo  0,050 r=0,80
oedometer tests. The stiffness parameter is not strictly
Cc = 0,267 · eo  0,053 r=0,79
the oedometer modulus, but the compression modulus.
Cc = 0,339 · eo  0,110 r=0,79
This modulus relates the volumetric strain to the log of 0.00
0,00
0.4
0,4 0.5
0,5 0.6
0,6 0.7
0,7 0.8
0,8 0.9
0,9 1.0
1,0
mean effective stress. If volumetric strain is plotted
Void ratio eo [-]
against ln p’ (mean effective stress), parameters * and Figure 9. Relationship between C c and e0
* can be obtained from the graph, as it is shown in Fig.
7.[3]. In related literature, several recommendations can be
found for the relationship between compression index
and water content or void ratio [9]. Some of them are
collected in Table 2 according to Das [10]. Eq. (13) and
(14) are in agreement with the equations given in Table
2.
Table 2. Recommendations for compression index
Skempton moulded clay Cc = 0.007 · (wL – 7)
[1944] Chicago-clay Cc = 0.013 · wo
Hough
inorganic cohesion soils Cc = 0.3 · (eo – 0.27)
[1957]
Nishida clay Cc = 1.15· (eo – 0.27)
Figure 7. Logarithmic relation between volumetric strain and mean (1956) low plasticity soil Cc = 0.75· (eo – 0.50)
stress [3]
Fig. 10. shows the correlation between modified
International literature offers a wide range of compression index (*) and water content (w0). The
correlation between Cc and Cs index. Based on our test relationship is weaker than it is shown on Fig. 8. and the
results for clayey soils, the average value of the different plasticity lines show less consisten behavior.
compression index is Cc≈0.1, which corresponds to For all data, the relationship can be described by
recommendations in Hungarian literature [8]. ∗
= 0.0015 − ( − 3.3) (14)
0,050
0.05
4.1. Parameter analysis for SS model IP < 15
15 < IP < 20
Modified compression index  * [-]

0.04
0,040
20 < IP < 30
The best relationship occured between the
IP > 30
compression index and the original water content as
0.03 all tests
0,030
shown in Fig. 8. Based on all test results, the
relationship between water content and compression
l * = 0,0016 · wo  0,009 r=0,85
0.02
index can be described by 0,020
l * = 0,0019 · wo  0,013 r=0,81
= 0.007 − ( − 5.9) (13)
l * = 0,0010 · wo + 0,005 r=0,81
with a correlation coefficient r = 0.85. 0.01
0,010
0,20
0.20 l * = 0,0011 · wo + 0,006 r=0,77

0.00 l * = 0,0015 · wo  0,005 r=0,81

0,000
10 15 20 25 30 35 40
Compression index Cc [-]

0.15
0,15 Water content wo [%]
*
Figure 10. Relationship between  and w0

0.10
0,10 Fig. 11. shows the strong relationship between
IP < 15 Cc = 0,007 · wo  0,049 r=0,87
swelling index (Cs) and compression index (Cc) with
15 < IP < 20 Cc = 0,008 · wo  0,066 r=0,84
the compression index 6 times higher than swelling
0.05
0,05 20 < IP < 30 Cc = 0,005 · wo  0,001 r=0,86 index. This is in good agreement with the international
IP > 30 Cc = 0,006 · wo  0,008 r=0,85 literature. The swelling index can be expressed by
all tests Cc = 0,007 · wo  0,041 r=0,85 = ∙ 0.175 = /5.7 (15)
0.00
0,00
10 15 20 25 30 35 40 45 50
Based on the graph, correlations for all types of soils
Water content wo [%] are very similar. The correlation coefficient is r=0.84.
Figure 8. Relationship between CC and w0 Relationship between  and  was examined on
Fig. 12. The ratios are quite similar to those graphed on
Fig. 9. shows the correlation between compression Fig. 11., but the correlation coefficient is a bit smaller.
index and void ratio. The correlation coefficient is Based on all the data, the ratio between compression
weaker than for water content. The best-fit correlation index and swelling index can be derived from
can be described by ∗

= 0.34 − ( − 0.32) (14) ∗ = 5.7 (16)

 The ratio agrees with the recommendation from the 5. Conclusions
PLAXIS manual of 2.5 to 7 [3]. Based on some back
analyses, it is suggested that the SS advanved In order to produce the software input parameters for
constitutive model is suitable for designing embankment advanced soil models, more demanding laboratory tests
foundations on soft soil. are needed. The basic demands are to determine the
0,04
0.04
IP < 15
power for stress-level dependency of stiffness (m),
15 < IP < 20 tangent stiffness for primary oedometer loading (Eoedref )
0.03
0,03
0,035
20 < IP < 30
0,006125
and unloading/reloading stiffness (Eurref ) when using the
0,18 0,0315
Hardening Model. For the Soft Soil Model,the modified
Swelling index Cs [-]

IP > 30
all tests
compression index (λ*) and modified swelling index (κ*)
0,02
0.02
Cs = 0,167 · Cc r=0,82 are needed. This paper focused on the input parameters
Cs = 0,166 · Cc r=0,82 of Hardening Soil Model (HS) and Soft Soil Model (SS)
0,01
0.01
5,71428571
Cs = 0,182 · Cc r=0,81 for Transdanubian Clay. More than 150 oedometer test
Cs = 0,182 · Cc r=0,85 were analysed focusing on the determination of these
Cs = 0,175 · Cc r=0,84 parameters. Based on the data analysis, Table 3.
0,00
0.00
0.00
0,00 0,05
0.05 0,10
0.10 0,15
0.15 0,20
0.20 0,25
0.25
summarizes the equations established for Transdanubian
Compression index Cc [-] clay.
Figure 11. Relationship between C c and Cs
Table 3. Determined correlations
0,010
0.010 Hardening Soil Model Sof Soil Model
k* = 0,169 · l * r=0,80
,
= 0.012 ∙ = 0.007 − ( − 5.9)
Modified swelling index  [-]

k* = 0,166 · l * r=0,79
0,008
0.008 = 0.25 ∙ + 4.60 = 0.34 − ( − 0.32)
k* = 0,182 · l * r=0,78 ∗
= 4.95 ∙ = 0.0015 − ( − 3.3)
k* = 0,182 · l * r=0,82 ∗
0,006
0.006 / ∗ = 5.7
These parameters agree with the range of
0,004
0.004 recommended values from the PLAXIS manual, but
IP < 15
15 < IP < 20
give a more detailed description, especially for
0,002
0.002 20 < IP < 30 Transdanubian clay. These correlations can be applied
IP > 30 by geotechnical engineers in cases where only index
0,000
0.000
all tests
properties are available or when preliminary evaluations
0.0
0,00 0.01
0,01 0.02
0,02 0.03
0,03 0.04
0,04 0.05
0,05 are performed. They also add confidence to the analyst
Modified compression index  [-]
* * when performing computations using PLAXIS or other
Figure 12. Relationship between  and 
sophisticated programs.
Fig. 13. shows an example for the application of SS
model using PLAXIS 3D. An embankment with a 6. References
height of h=5.5m was constructed in central Hungary.
The top 3m of the subsoil was peat, resting on 6m of [1] Szepesházi R.: Mély munkagödrök mentén bekövetkező
mozgások (Dispalcements around deep excavations), Phd
soft clay. To reduce settlement, dynamic replacement kutatószeminárium, Miskolc, Hungary, 2007. (Magyar)
was executed. During contruction, systematic [2] Koch E.: Töltésalapozási eljárások modellezése (Modeling of
monitoring was carried out. After the execution of the embankment foundation), PhD thesis, Széchenyi István Universi-
highway section, a back-analysis was performed using ty, Győr, 2013. (Magyar)
[3] Brinkgreve R.B.J., Vermeer P.A. (2015): PLAXIS-Finite element
PLAXIS 3D and MC and SS material models. code for soil and rock analyses, Plaxis 3D. Manuals, Delft Uni-
Only index parameters of the subsoil were tested in versity of Technology  Plaxis bv, The Netherlands.
the laboratory. To determine the input parameters for [4] Janbu, N. (1963). Soil compressibility as detrmined by oedome-
the SS constitutive model, the above mentioned ter and triaxial tests. In Proc. ECSMFE, volume 1, pages 19–25,
Wiesbaden.
equations were applied. Result of the calculation with
[5] Smoltczyk, U. (1990): Grundbautaschenbuch, Ernst & Sohn,
conventional method, PLAXIS 3D modeling using MC Berlin, pp.143-174.
and SS material model and the result of the monitoring [6] Varga L. (1986): Geotechnika III, Tankönyvkiadó, Budapest,
can be seen on Fig. 13. Results from modeling with SS pp.35-40.
[7] Katzenbach R., Bachmann G., Gutberlet C. (2008): Soil-
model agreed very well with the results of the Structure interaction and ULS design of complex deep founda-
monitoring. tions, 6th International Conference on Case Histories in Ge-
10
height h (m)
embankment

otechnical Engineering, Arlington, VA, August 11-16, 2008,

5 OSP 5..
time t (day)
0
[8] Szepesházi R. (1986): Geotechnika II, Tankönyvkiadó, Buda-
0 50 100 150 200 250 300 350 400 pest, pp. 215.
-5
[9] Kempfert, H.G. (2006): Excavations and Foundations in Soft
-10
monitoring
Soils, Springer, The Netherlands, pp. 24-25.
-15
[10] Das, Braja,M.. (2006): Principles of Geotechnical Engineering,
settlement s (cm)

conventional method
Thomson, 6th edition, Toronto, pp. 327.
-20 PLAXIS 3D Mohr - Coulomb model

-25 PLAXIS 3D Soft - Soil model

-30

-35

-40

Figure 13. Settlement calculations and monitoring result