Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

International Conference on Civil Engineering

Architecture & Urban Sustainable Development


18 &19 December 2013, Tabriz , Iran

Effect of Soil Constitutive Models on


Determination of Deformations Caused by
Deep-Excavations
1. Seyed Mohammad Iman Bathaeian, Department of Civil Engineering, Islamic Azad
University – Central Tehran Branch, iman.bathaeian@gmail.com
2. Morteza Jiryaei Sharahi, Department of Civil Engineering, Qom University of Technology,
jiryaei@qut.ac.ir

Abstract

This study presents a detailed analysis of a well-documented deep


excavation case history under undrained conditions with three different
soil models, i.e. Mohr Coulomb, the soft soil (SS) model and the
hardening soil (HS) model. Results from the application of MC model
indicate over prediction in the initial stages of excavation and under
prediction in the final stages, both for wall deflection and ground
settlement. The SS model generally over predicts wall deflection and
ground settlement, moreover, the peak of settlement curves is further
from the wall and also the pattern of wall deflection in the first stages is
not cantilever. The HS model under predicts both wall deflection and
ground settlement in the final stages and over predicts ground settlement
in the initial stages, however, in the initial stages the wall deflection is
predicted with acceptable accuracy and almost the same deflection
pattern as the field observations. Furthermore, the peak of the settlement
curves is similar to the field measurements. With a change in Pref
parameter of the HS model in proportion to the effective stress in depth,
the results become closer to those measured in the field. Therefore, it can
be inferred that among the three MC, SS and HS models. The hardening
soil model can predict excavation-induced movements more accurately.
Furthermore, it can be understood that limiting soil behavior to a
hypothetic yield surface is the main weakness of most models.
Keywords: Deep excavation, MC model, SS model, HS model, Finite element method (FEM)
International Conference on Civil Engineering
Architecture & Urban Sustainable Development
18 &19 December 2013, Tabriz , Iran

1. Introduction
Deep excavations are often used in urban areas. Ground deformations due to such
excavations can be extremely hazardous for the adjacent buildings. Therefore, it is
important to accurately predict these movements. Many studies have proved the
reliability of FEM in predicting excavation deformations; however, accurate
estimation of excavations is highly dependent on the application of the appropriate
soil model and the model parameters. The three selected soil models in this
investigation are implemented in a FEM model; however, none of the above
mentioned models can accurately predict the deformations caused by the deep
excavation.

In this paper we analyze a well-documented excavation case history in Taipei. The


soil parameters required in the three models are determined by soil laboratory tests.
Before presenting the results of each model, a summary of soil constitutive model is
provided so as to better clarify the model’s strengths and weaknesses in prediction of
ground deformation. Due to change in stress level during the excavation, model’s
performance is compared in the initial stages and the final stages of excavation. It
should be noted that the predictions of excavation responses are so-called “Class C”
predictions, made after the event. In other words, these are back analyses of a well-
documented case history.

2. TNEC Case History

2.1 Construction details

The TNEC structure is an 18 story building with five level basements. The depth of
the excavation is 19.7 m; the thickness and depth of the diaphragm wall are 90cm and
35m respectively. The ground water table is at the depth of 2.0 m below the ground
surface. The excavation was completed in seven stages. Figure 1 shows the sequence
of excavation and stratigraphic condition [7].
International Conference on Civil Engineering
Architecture & Urban Sustainable Development
18 &19 December 2013, Tabriz , Iran

Fig.1 Profile of excavation sequence and subsurface soil layers [7]

2.2 Instrumentation plan

The excavation induced wall-deflection and ground settlement were observed with
five inclinometers, three extensometers and a number of settlement points along the
main observation section. The arrangement of instrumentation in the main
observation section is shown in Figure 2 [4]. The reader is referred to Ou et al [5] for
additional details.
International Conference on Civil Engineering
Architecture & Urban Sustainable Development
18 &19 December 2013, Tabriz , Iran

Fig 2 Plan view of the TNEC case and instrumentation plan [4]

2.3 Subsurface soil conditions

Figure 3 shows the variation of water content and in-situ void ratio with depth
obtained from [5] along with the test data conducted recently [6]. Figure 4 shows the
values of compression index and swelling index obtained from [5]. Figure 5(a) shows
the variation of OCR with depth. As shown in the figure, the clay at depths of below
12m can be treated as normally consolidated clay and above GL-12m
overconsolidated. Figure 5(b) shows the variation of undrained shear strength
obtained from the UU test, the field vane shear test and CK0U-AC test and CK0U-AE
test [6, 7].
International Conference on Civil Engineering
Architecture & Urban Sustainable Development
18 &19 December 2013, Tabriz , Iran

Fig. 3 The variation of (a) water content and (b) initial void ration with depth [5]

Fig. 4 The variation of (a) compression index and (b) swelling index with depth [5]
International Conference on Civil Engineering
Architecture & Urban Sustainable Development
18 &19 December 2013, Tabriz , Iran

Fig. 5 The variation of (a) OCR and (b) undrained shear test with depth [5]

2.3 Excavation behavior

Field observations of the TNEC case show cantilever deflection of the diaphragm
wall in the first stages and as the excavation completes, the wall shows more inward
movement with the maximum wall deflections (106 mm) near the bottom of the
excavation. The ratio of maximum deflection to excavation depth is about 0.54%,
Moreover, concave-type settlement patterns occurred at all stages except for stage 1.
The maximums surface settlement is 78mm, and the distance from the wall where this
maximum surface settlement occurred is about 11m, with a ratio of distance to
excavation depth of about 0.5 according to literature [4].

3. Parameters and FEM considerations

3.1 Lateral supports and sandy layer parameters

The structural components of TNEC excavation were diaphragm wall, slabs and steel
strut systems. The modulus of steel struts and bending moment of diaphragm wall are
reduced due to the cracking in the wall and repeated use of the steel struts (a
International Conference on Civil Engineering
Architecture & Urban Sustainable Development
18 &19 December 2013, Tabriz , Iran

reduction of 20% for the struts and 30% for the diaphragm wall as suggested by Ou
and Shiau [11]). The summary of the input parameters are described in tables 1 and 2
[8].

Table 1 Input parameters of diaphragm wall

Parameters Name Value Unit


Normal Stiffness EA 22140000 kN/m
Flexural Rigidity EI 1494450 kN/m2/m
Equivalent Thickness d 0.9 m
Weight w 21.6 kN/m/m
Poisson’s Ratio υ 0.15 -

Table 2 Input parameters of struts and slab

Structure Stiffness EA (kN) Spacing (m) Preload Force (kN/strut)


Slabs 3690000 1 -
H300 1954810 8 784.8
H400 3569021 3 1177

Since the clayey layers mainly affect the behavior of excavation and the effect of soil
models on these layers is the main goal of this study, sand layers are simulated as a
drained material, using the Mohr-Coulomb model with the input parameters
demonstrated in table 3 [7].

Table 3 Input parameters of drained material for the Mohr-Coulomb model

Depth Ndesign γ Φ' E υ


(m) (kN/m3) (kPa)
5.6 – 8 7 18.93 30 68351 0.3
33 – 35 23 19.62 33 265473 0.3
37.5 – 45 26 19.62 35 300247 0.3

3.2 Mesh boundary

Adequate mesh boundaries must be determined because the FEM mesh may
significantly affect the computation of stress and strain in the soil and the retaining
wall. As shown in figure 5, the left boundary is half of the excavation width, while
the right boundary is established based on the previous studies [2, 3], in which the
influence zone behind the wall is approximated to be four times the excavation
depth. The excavation depth in the TNEC case is 19.7 m, so a distance of 100m from
the wall to the right boundary is considered to be appropriate.
International Conference on Civil Engineering
Architecture & Urban Sustainable Development
18 &19 December 2013, Tabriz , Iran

3.3 Clayey layers consideration

In order to define the stiffness and cohesion of the soil with regard to depth in HS and
SS models, for the computations conducted by these two soil models, the clayey
layers are defined with clusters 1m deep. This consideration is well demonstrated in
figure 6.

Fig. 6 Finite element mesh and clayey layers consideration

3.4 Plane strain assumption

The analysis was assumed to be 2D and plane strain which means considering an
infinitely long retaining wall. As a result, field observations may not be directly
comparable to the results of the analysis with plane strain assumption. Therefore, the
concept of the “plane strain ratio” as defined by Ou et al. [10] was applied so as to
check the applicability of plane strain conditions. The ratio is about 0.9 for the main
section of the TNEC case [4], which implies that the assumption of plane strain is
largely valid for this case.
International Conference on Civil Engineering
Architecture & Urban Sustainable Development
18 &19 December 2013, Tabriz , Iran

4. Analysis
4.1 Mohr-Coulomb model

4.1.1 A brief overview of the Mohr Coulomb model

In this model plasticity is associated with the development of irreversible strains. The
occurrence of plasticity in this model depends on a yield function named ƒ, which is a
function of stress and strain. The Mohr-Coulomb model is a perfectly plastic model
which means that the yield function is fixed and is a function of model parameters.

The Mohr-Coulomb yield function is an extension of Coulomb’s friction law to


general states of stress, for further information you can refer to [1, 12]. Two
important plastic parameters φ and c representing friction angle and cohesion are
respectively in the yield functions, which give a hexagonal cone in principal stress
space as shown in figure 7.

Fig. 7 The Mohr-Coulomb yield surface in principal stress space (c=0) [1]

Moreover, six plastic potential functions are also introduced with a new plasticity
parameter ψ which is required to model positive plastic volumetric strain.

For stress states within the yield surface, the behavior is elastic and obeys Hook’s
law for isotropic linear elasticity. Therefore, two elasticity parameters Young’s
modulus E and Poisson’s ratio υ are required, for further information you can
refer to [13].
International Conference on Civil Engineering
Architecture & Urban Sustainable Development
18 &19 December 2013, Tabriz , Iran

4.1.2 Soil parameters for the Mohr-Coulomb model

For the clayey layers, the following parameters are used for analysis.

Table 4 Input parameters for undrained materials for Mohr-Coulomb model

Depth γt Φ' S/σ ʹv Eur/ S υ


0 – 5.6 18.3 33 0.32 600 0.2
8 – 33 18.9 30 0.34 950 0.2
35 – 37.5 18.2 32 0.36 950 0.2

Passion’s ratio, for undrained materials is chosen to be equal to 0.2. For other
stiffness parameters, Eref and Eoed can be obtained according to equations 1 and 2
[14].

Eref = Eur / 3 (1)

Eoed = 0.7 Eref (2)

4.1.3 Comparison and discussion of the Mohr-Coulomb model result

Figures 8 and 9, illustrate a comparison between field measurement and the


results from analysis with Mohr-Coulomb model.

Distance from the middle of excavation (m)

Fig. 8 Comparison between field settlement and analysis with Mohr-Coulomb model
International Conference on Civil Engineering
Architecture & Urban Sustainable Development
18 &19 December 2013, Tabriz , Iran

Fig. 9 Comparison between field deflection and analysis with Mohr-Coulomb model

As can be seen in figures 8 and 9, both for ground settlement and wall deflection,
stages 1,2 and 3 of excavation are over predicted by Mohr-Coulomb model, stage 4 of
excavation is close to the field measurements and stages 6 and 7 are under predicted
by the model. Furthermore, the distance at which the maximum ground settlement
occurs, is further predicted by the model as the excavation proceeds. However, the
pattern of wall deflection is well predicted by the model. In the initial stages, wall
deflection is predicted to be cantilever and in the latter stages, inward bending of the
wall is predicted by the model.

4.2 Soft soil model


4.2.1 A brief overview of soft soil model

In soft soil model, a logarithmic relation between the volumetric strain, εv , and the
mean effective stress, P', is assumed. Therefore, stiffness is dependent on stress
level. Furthermore, distinction between primary loading and unloading-reloading is
considered. Two modified parameters, λ* and κ*, are introduced in this model.
Equations 3 and 4 describe their relationship to Cam-Clay parameters.

(3)

λ
λ
International Conference on Civil Engineering
Architecture & Urban Sustainable Development
18 &19 December 2013, Tabriz , Iran

The soil unloading and reloading is assumed to be elastic as described by Hooke’s


law. Failure behavior is also according to the Mohr-Coulomb criterion. For further
information about this model you can refer to [1].

4.2.2 Soil parameters for the Soft Soil model

Strength parameters, c and φ, are the same as those introduced for the Mohr-Coulomb
model. However, for the stiffness parameters, Cc, Cs and eintitial ,are used as alternatives.
These values are derived by interpolation from figures 3 and 4.

4.2.3 Comparison and discussion of the Soft Soil model result

Figures 10 and 11; illustrate a comparison between field measurement and the results
from analysis with Soft Soil model.

Distance from the middle of excavation (m)

Fig. 10 Comparison between field settlement and analysis with Soft Soil model
International Conference on Civil Engineering
Architecture & Urban Sustainable Development
18 &19 December 2013, Tabriz , Iran

Fig. 11 Comparison between field deformations and analysis with Soft Soil model

As can be seen in figures 10 and 11, soft soil model has over predicted both wall
deflection and ground settlement for all stages. Furthermore, the distance at which the
maximum ground settlement occurs is predicted to be further from the wall. Wall
deflection at all stages is predicted to occur with an inward bending which is in
contrast to what the field measurements show.

4.3 Hardening Soil model


4.3.1 A brief overview of Hardening Soil model

The existing double-stiffness models such as Cam-Clay model [16] and Duncan-
Chang model [17], cannot consistently distinguish between loading and unloading. In
addition, such models are not suitable for collapse computations in fully plastic range.
These restrictions are overcome in Hardening Soil model firstly by using the theory
of plasticity rather than the theory of elasticity as used in Duncan-Chang model,
secondly by including soil dilatancy and thirdly by introducing a yield cap. The yield
surface of the Hardening Soil model is not fixed in principal stress space, but it can
expand due to plastic straining. Distinction is made between two main types of
hardening, namely shear hardening and compression hardening. For further
information about this model, readers are referred to [1, 18].
International Conference on Civil Engineering
Architecture & Urban Sustainable Development
18 &19 December 2013, Tabriz , Iran

4.3.2 Soil parameters for the Hardening Soil model

Two models are simulated using the Hardening Soil model, in both models; strength
parameters are as used in Mohr-Coulomb model. However, reference stiffness
parameters are calculated using the value of σ in the middle of each layer and
substituting it in equation 5. Equations 1 and 2 are applied to determine Eur and Eoed.

φ σ φ
( )
φ φ
In the first model, the default setting of the model for Pref =100 kN/m2 is used,
however, in the second model, Pref is calculated based on the increase of vertical
effective stress with depth in soil layers.

4.3.3 Comparison and discussion of the Hardening Soil model result

Figure 12 and 13; illustrate a comparison between field measurement and the results
from analysis with hardening soil model, using the default Pref=100 kN/m2.

Distance from the middle of excavation (m)

Fig. 12 Comparison between field deformations and analysis with hardening soil
model using the default Pref=100 kN/m2
International Conference on Civil Engineering
Architecture & Urban Sustainable Development
18 &19 December 2013, Tabriz , Iran

Fig. 13 Comparison between field deformations and analysis with hardening soil
model using the default Pref=100 kN/m2

As can be seen in figures 12 and 13, for ground settlement, in the initial stages 1, 2
and 3, the Hardening Soil model has over predicted the settlement, stage 4 is the
closest to the field measurements and the final stages are under predicted by the
model. The distance from the wall at which the maximum settlement occurs is
predicted with accurate accuracy and the results are predicted with better
approximation as we get further from the retaining wall. For wall deflection, the
initial stages 1, 2 and 3 are well predicted; however, for the other stages we have
underprediction. The pattern of wall deformation also matches with the field
measurements.

Figures 14 and 15; illustrate a comparison between field measurements and the
results from analysis with hardeing soil model, calculating Pref with depth.
International Conference on Civil Engineering
Architecture & Urban Sustainable Development
18 &19 December 2013, Tabriz , Iran

Distance from the middle of excavation (m)

Fig. 14 Comparison between field deformations and analysis with hardening soil
model with a modification in Pref

Fig. 15 Comparison between field deformations and analysis with hardening soil
model with a modification in Pref

As can be seen in figure 14, the same as figures 12, for ground settlement, we have
overprediction in the initial stages and underprediction in the final stages, however,
with a change in Pref , it can be clearly seen that the results from hardening soil model
have become closer to the field measurements, especially in the initial stages. As can
be seen in figure 15, for wall deflection, this improvement in prediction can be better
seen, as the results for stages 1, 2 and 3, especially at the depths that the maximum
deflections occur, have become closer to the field measurements in comparison with
figure 13.
International Conference on Civil Engineering
Architecture & Urban Sustainable Development
18 &19 December 2013, Tabriz , Iran

5. Discussion

As observed in the text, for MC and HS models we have overprediction in the final
stages and the results of the final stages are further from the field measurements. This
phenomenon can be justified by the fact that in both of these models the yield surface
is symmetric to the
(σ1 = σ2 = σ3) line; however, the real soil behavior is symmetric to K0 line. In the
initial stages of excavation unloading forces are small and consequently the stress
state is inside of the prescribed yield surface and also inside of the real soil behavior;.

However, when the excavation progresses deeper and the unloading forces become
large enough to cause the stress state of the soil to be in the plastic state to form larger
deformations, the models still predict the stress state inside of the yield surface.
Hence, predicted deformations in the final stages are smaller than those of the field
measurements.

For this case study the soil is considered to be overconsolidated and subjected to
unloading; however, SS model is developed for primary loading of near normally-
consolidated soils, therefore it is expected for this model to overpredict deformations
in almost all stages of the excavation.

6. Conclusion

The TNEC case history is a well-documented excavation which many researchers and
scientists have used for verification of their analysis. In this study, three Plaxis soil
models are used in order to model this excavation. Among the models used, the
Hardening Soil model has the closest approximation, both in size and shape of
deformations. A modification in Pref parameter of this model yields also better
predictions of the excavation.

5. Reference

1. Materials model manual (plaxis)


2. Kung GTC, Hsiao ECL, Juang CH. Evaluation of a simplified small-strain
soil model for estimation of excavation-induced movements. Can Geotech J
2007;44:726-36
3. Yu-Geng Tang, Gordon Tung-Chin Kung. Investigating the effect of soil
models on deformations caused by braced excavations through an inverse-
analysis technique. Computers and Geotechnics J 2010;37:769-780
International Conference on Civil Engineering
Architecture & Urban Sustainable Development
18 &19 December 2013, Tabriz , Iran

4. Gordon Tung-Ching Kung, Chang-Yu Ou, C.Hsein Juang. Modeling small-


strain behavior of Taipei clays for finite element analysis of braced
excavations. Computers and Geotechnics J 2009;36:304-319
5. Ou CY, Liao JT, Lin HD. Performance of diaphragm wall using top-down
method. Geotech Geoenviron Eng J 1998;12(3):271-283
6. Teng, F.C. 2010; Personal file
7. Aswin Lim, Chang-Yu Ou, Pio-Gp Hsieh. Evaluation of clay constitutive
models for analysis of deep excavation under undrained conditiona.
GeoEngineering J 2010;5(No.1)9-20
8. Dang Houu Phouc. Excavation behavior and adjacent building response using
user-defined soil models in Plaxis NTUST 2009;Master dissertation
9. Clough GW, O’Rourke TD. Construction induced movements of insitu walls.
Design and performance of earth retaining structures, vol.25. Geotechnical
Special Publication;1990.p.439-70
10. Ou CY, Chiou DC, Wu TS. Three-dimensional finite element analysis of deep
excavations. J Geotech Eng;1996:122(5):337-45
11. Ou CY, Shiau BY. Analysis of the corner effect on the excavation behavior.
Can Geotech J 1998;35(3):532-540
12. Niele Saabye Ottosen, Matti Ristinmaa. The mechanics of constitutive
modeling. Elsevier 2005;First edition:165-171
13. Niele Saabye Ottosen, Matti Ristinmaa. The mechanics of constitutive
modeling. Elsevier 2005;First edition:90-93
14. Calvello, M. and Finno, R. Selecting parameters to optimize in model
calibration by inverse analysis. Computers and Geotechnics J 2004;31:410-
424
15. Burland, J.B. The yielding and dilation of clay. Géotechnique J 1965;15:211-
214
16. Hashiguchi, K. Fundamental requirements and formulations of elastoplastic
constitutive equations with tangential plasticity. Plasticity J1993;9:525-549
17. Duncan, J.M. Chang, C.Y. Nonlinear analysis of stress and strain in soil. Soil
Mech. Found. Div. ASCE 1970;96:1629-1653
18. Schanz, T. Vermeer, P.A. Bonnier, P.G. The hardening soil model:
Formulation and Verification. Beyond 2000 in Computational
Geotechnics1999
International Conference on Civil Engineering
Architecture & Urban Sustainable Development
18 &19 December 2013, Tabriz , Iran

You might also like