Tunnelling and Underground Space Technology 63 (2017) 118–132

Contents lists available at ScienceDirect

Tunnelling and Underground Space Technology

journal homepage: www.elsevier.com/locate/tust

Stress paths in deep excavations under undrained conditions and its

influence on deformation analysis
Aswin Lim a,b,⇑, Chang-Yu Ou c
a
Department of Civil and Construction Engineering, National Taiwan University of Science and Technology, 10672, Taiwan, ROC
b
Department of Civil Engineering, Universitas Katolik Parahyangan, 40141, Bandung, Indonesia
c
Department of Civil and Construction Engineering, National Taiwan University of Science and Technology, Taipei 10672, Taiwan, ROC

a r t i c l e i n f o a b s t r a c t

Article history: The objective of this study is to examine the stress state of soils during deep excavation, in relation to the
Received 23 October 2015 determination of appropriate soil parameters for deformation analysis of a deep excavation case using
Received in revised form 15 December 2016 the finite element method. Two well documented case histories of a deep excavation were utilized for
Accepted 30 December 2016
the validation of the analysis procedure and the selection of soil stiffness parameters. Results from the
Hardening Soil model showed that the out-of-plane stress has significance influences to the direction
of soil effective stress path. In addition, most of the soil inside and outside excavation zone is in the elastic
Keywords:
behavior. Even though the effective stress path of soils adjacent to the diaphragm wall have undergone
Deep excavation
Hardening soil model
yield, but the characteristics of those soils are still dominated by the elastic behavior. Hence, the unload-
Mohr-Coulomb model ing/reloading parameters are predominant in a deformation analysis of an excavation case. When the
Stress path undrained shear strength and unloading/reloading modulus were precisely specified, even the Mohr-
Unloading/reloading parameter Coulomb model could obtain good prediction of the wall deflections. Moreover, a hypothetical case
Deformation characteristic was employed to investigate the performance of the computed ground surface settlements. The result
Small strain characteristic showed that the computed ground settlement from Mohr-Coulomb model was close to the result from
the Hardening Soil Small model if the layer of soft soil is deep enough and a layer of small strain stiffness
zone is introduced at bottom of the model geometry.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction example the Nicoll Highway failure case (Whittle and Davies,
2006).
The necessity of underground space in dense urban areas, such Nowadays, many approaches can be used for design and analy-
as Taipei, Shanghai and Singapore is imperative. In the last decades sis of a deep excavation, such as the earth pressure method, the
there are clear trends using underground space in urban develop- numerical method, and the robust geotechnical design (RGD)
ment, for example, as part of buildings mainly for parking and method (Wang et al., 2014), a new design method that involves
commercial uses (a basement) or for underground metro system the theory of reliability and uncertainty in geotechnical engineer-
(Zhao and Künzli, 2016). The deep excavation method is usually ing. However, the numerical method, such as the finite element
performed to construct a basement or the cut-and-cover tunnel. method, seems still a popular method in design and analysis of a
In some cases, a deep excavation is located close to existing under- deep excavation, because not only it can simulate the stage con-
ground metro tunnels or adjacent buildings (Shi et al., 2015; Hsieh struction procedures of excavation, but also there are a lot of soil
et al., 2015; Hong et al., 2015; Chen et al., 2016). Obviously, the constitutive models can be adopted to model the soil behavior.
design and construction of a deep excavation should be carefully Due to many soil constitutive models available in the finite ele-
executed in order to avoid excessive wall deflections and ground ment method, it also been utilized by some researchers to study
surface settlements, or even the collapse of the retaining wall, for the soil behavior (Surarak et al., 2012; Ho and Hsieh, 2013).
Plenteous soil constitutive models have been developed to sim-
ulate the effective stress-strain-strength behavior of clay soil. Initi-
ated with the Mohr-Coulomb model, a classic but widely
⇑ Corresponding author at: Department of Civil and Construction Engineering,
performed in practical geotechnical engineering, until the exten-
National Taiwan University of Science and Technology, 10672, Taiwan, ROC.
sion of the Hypoplasticity Cam-Clay model with considering stiff-
E-mail addresses: [email protected], [email protected]
(A. Lim), [email protected] (C.-Y. Ou). ness anisotropy (Mašín, 2014), a sophisticated soil model with

http://dx.doi.org/10.1016/j.tust.2016.12.013
0886-7798/Ó 2017 Elsevier Ltd. All rights reserved.
 A. Lim, C.-Y. Ou / Tunnelling and Underground Space Technology 63 (2017) 118–132 119

twelve input soil parameters deduced from advanced laboratory The Hardening Soil model, abbreviated as the HS model, is a
testing apparatus such as a hollow cylinder apparatus, or the true second order model for soil in general (soft to stiff types of
MIT-E3 model (Whittle and Kavvadas, 1994), which requires fif- soil). The model involves frictional hardening characteristics to
teen input soil parameters obtained from at least five types of test- model the plastic shear strain in deviatoric loading, and cap hard-
ing such as 1D consolidation tests, K0-oedometer or K0 triaxial ening characteristics to model the plastic volumetric strain in the
tests, undrained triaxial shear tests (both axial compression and primary compression. The failure is defined by the Mohr-
axial extension conditions), shear wave velocity tests, and drained Coulomb failure criterion. The basic characteristics of the model
triaxial tests. are a Mohr–Coulomb failure with input parameters c, u and dila-
Although the MIT-E3 and/or the Hypolasticity-clay models tancy angle, w, stress-dependent stiffness according to a power
could provide reasonable prediction of an excavation deformation law defined by input parameter, m, plastic straining resulting from
behavior (Mašín, 2005; Whittle and Davies, 2006; Corral and primary deviatoric loading with an input parameter, Eref
50 , and plas-
Whittle, 2010; Mašín et al., 2011), but most of the engineers might tic straining from primary compression with an input parameter
adopt it reluctantly because the input parameters are not easily to
Eref
oed , elastic unloading/reloading is defined by input parameters
be justified from conventional soil tests. In geotechnical engineer-
ing practice, the Mohr-Coulomb model is yet a popular model to be Eref
ur and unloading/reloading Poisson’s ratio, tur . Fig. 1 displays

applied because it is simple and most of engineers have good expe- the shear yield surface and cap yield surface in the Hardening Soil
riences and confidence to utilize this model. One concern of this Model for soil with no cohesion (c0 ¼ 0). In this paper, the soil yield
model is the input parameters such as the Young’s modulus and is defined as the stress state of soil which is located in the shear
the shear strength of soil for analysis need to be adjusted based hardening zone. Meanwhile, the soil failure is defined as the stress
on a local experiences or from back analysis (Ng and Lings, 1995; state of soil which reaches to the Mohr-Coulomb failure line.
Whittle and Davies, 2006; Lim et al., 2010; Hwang et al., 2012). It should be noted that, the HS model is difficult to accurately
As the deformation characteristic of an excavation system is predict the drop in the deviator stress, which represents a strain
influenced by the stress state and stress history of the soil, under- softening response of soil behavior. Nevertheless, in terms of an
standing the stress path during an excavation process is essential. effective stress path, the typical shape of the normally consolidated
Powrie et al. (1998) investigated the stress-strain relations appro- clay stress paths, and their undrained shear strength, are handled
priate to diaphragm wall in clay by performing a series of triaxial very well by the HS model predictions (Surarak et al., 2012). In
test on speswhite Kaolin. The stress paths imposed in the triaxial other words, the HS model can represent real soil behavior as long
testing program intended to simulate those that would be experi- as the soil response is a strain hardening behavior.
enced in reality by soil elements at a depth of 12 m, behind and in The Hardening Soil Small-strain model, abbreviated as the HSS
front of an in situ retaining wall, during wall installation and main model, evolves from the HS model with the consideration of small
excavation. However, their results were limited to the stress path strain characteristics of soil. In the HSS model, two additional
in which no significant principal stress rotation occurred, as it parameters are required in addition to those in the HS model.
might be significant at the soil element near the wall toe or at The two additional parameters are the reference shear modulus
the excavation bottom. at very small strains ðGref
0 Þ and the shear strain at shear modulus
In this article, a proper investigation of the soil stress paths dur- equal to 0.7 shear modulus at very small strain ðc0:7 Þ. The Mohr-
ing excavation was performed based on finite element simulation Coulomb Model, abbreviated as the MC model, is an elastic per-
of the deep excavation case history by adopting the Hardening Soil fectly plastic model and in fact, a combination of the Hooke’s law
Model. It is a common practice in the deep excavation modelling and the generalized form of Mohr-Coulomb failure criterion. The
that the diaphragm wall is assumed to be wished-in-place and anal- model involves four input parameters, such as two pseudo-
ysis results are reasonable (Ng and Lings, 1995; Lim et al., 2010; elastic parameters from the Hooke’s law (Young’s modulus (E),
Zhang et al., 2015). Thus, the soil stress paths during construction and Poisson’s ratio (t)), and the two parameters from the Mohr-
of the diaphragm wall are not discussed here. Two well docu- Coulomb failure criterion (the friction angle (u), and cohesion
mented case histories were simulated to validate the proposed pro- intercept (c)). For the MC model, the effective stress undrained
cedure of parameter determination, especially the soil modulus. In analysis can be performed with combination of the effective shear
addition, a logic and systematic procedure of parameters determi- strength parameters (c ¼ c0 and u ¼ u0 ), referred to as the
nation, directly derived from the conventional laboratory tests Undrained A approach, or combination of the total strength param-
and in-situ tests results, was used to accommodate the Mohr-
Coulomb model effective stress approach. Besides, a hypothetical
case was performed to examine the profile of ground surface settle-
ments with considering a layer of small strain stiffness zone.

2. Finite element analysis and soil constitutive models

The PLAXIS v.2010 (Brinkgreve et al., 2010) computer software

was used to perform the finite element analysis with various soil
constitutive models. Based on the evaluation performed by Lim
et al. (2010), three soil constitutive models were selected, such
as the Hardening Soil Model (Schanz et al., 1999), the Hardening
Soil-Small model (Benz et al., 2009), and the Mohr-Coulomb model,
to simulate the clay soil behaviors with the effective stress
undrained analysis. Furthermore, the effective stress drained anal-
ysis with the Mohr-Coulomb model was selected to model the
sandy soil. Since those prominent models are well developed, thus
only essential description will be highlighted to ease discussion of Fig. 1. Shear yield surface and cap yield surface in the Hardening Soil Model
research results. (modified from Brinkgreve et al., 2010).
120 A. Lim, C.-Y. Ou / Tunnelling and Underground Space Technology 63 (2017) 118–132

eters (c ¼ Su and u ¼ 0), referred as the Undrained B approach,

where Su is the undrained shear strength of soil. For both types
of analysis, the Young’s modulus and the Poisson’s ratio should
be in terms of the effective stress rather than the total stress. In
addition, both types of analyses have been also introduced in the
PLAXIS.
In the analysis, the structural members, such as a diaphragm
wall, struts, and concrete slabs, were assumed to behave as
linear-elastic. The diaphragm wall was modelled with a plate ele-
ment, while the struts and the concrete slabs were simulated as
the fixed end anchor elements. The stiffness of all the structural
elements were reduced by 20% from the nominal value, consider-
ing that the stiffness of the concrete retaining wall reduces when
large bending moment of the diaphragm wall causes the occur-
rence of the crack in the concrete (Lim et al., 2010).

3. Finite element analysis of excavation cases

The HS model was utilized as the validation benchmark of input

parameters. Then, the Taipei National Enterprise Center (TNEC)
excavation case was used to examine the performance and the
rationality of those input soil parameters. In the TNEC case history,
numerous comprehensive laboratory tests, such as the small strain
tests, the triaxial CK0UC and CK0UE tests and the multiple
unloading-reloading triaxial tests have been conducted. Therefore,
all of the input parameters for analyses were well studied and sci-
entifically developed.

3.1. The TNEC excavation project

3.1.1. The TNEC project description

Fig. 2. Profile of the TNEC case excavation sequence and subsurface soil layers.
The TNEC structure was an 18-story building with five level
basements. The depth of the excavation was 19.7 m, with dimen-
sion of the diaphragm wall was 90 cm in thickness and 35 m in unloading/reloading Young’s modulus was mathematically derived
depth. The groundwater level was at a depth of 2 m below the based on a result of oedometer tests (Lim et al., 2010), as shown in
ground surface (GL 2.0 m). The excavation was completed in Eq. (1), which e is void ratio, p0 is mean effective stress, j = cs/ln10,
seven stages. Fig. 2 depicts the sequence of excavation and base- and cs is swelling index.
ment construction.
According to the site investigation, the stratigraphic condition 3ð1 þ eÞp0 ð1  2tur Þ
Eur ¼ ð1Þ
at the site can be described as follows (see Fig. 2): the first layer j
is a soft silty clay (CL), which ranges from the ground surface level
In order to be used as an input parameter in the HS model, Eur
(GL) 0.0 m to GL 5.6 m, whose NSPT-value is around 2–4. The sec-
ond layer, from GL 5.6 m to GL 8.0 m, a loose silty fine sand with should be converted to the Eref
ur as proposed by Schanz et al.

NSPT-values between 4 and 11 and /0 = 30°. The third layer, from GL (1999), as shown in Eq. (2).
8.0 m to GL 33.0 m, is again a soft silty clay (CL) that NSPT-value !m
is around 2–5 and the PI is within the range of 9–23, with an aver- c  cos u  r03 sin u
Eur ¼ Eref ð2Þ
age value of 17. This layer is the one which mostly affects excava-
ur
c  cos u þ pref sin u
tion behavior. The OCR value of this layer is around 1.23–1.4,
which is lightly over-consolidated and the behavior of soil is strain When Eref ref ref ref ref
ur is determined, then E50 ¼ 1=3Eur and Eoed ¼ 0:7E50 can be
hardening response (Teng, 2010). The fourth layer, ranging from GL estimated as suggested by Calvello and Finno (2004).
33.0 m to GL 35.0 m, is a medium dense silty fine sand with Teng (2010) conducted bender element tests on the Taipei Silty
NSPT -value between 22 and 24 and /0 = 33°. The fifth layer is a Clay soil samples. The result is shown in Fig. 3 where the abscissa
medium soft clay, ranges from GL 35.0 m to GL 37.5 m, and the ordinate represent the normalized mean effective stress
NSPT-value is 9–11. The sixth layer is a medium dense to dense silt and shear modulus at very small strain, respectively. In the analy-
or silty fine sand; ranging from GL 37.5 m to GL 46.0 m, ses, Fig. 3 was used to estimate the value of shear modulus at very
NSPT = 14–37 and /0 = 35°. Below the sixth layer is a dense small strains (G0 ) corresponding to the effective mean stress of
Chingmei gravel soil and its NSPT is above 100. The complete data, each soil layers. By following Eq. (3) (Brinkgreve et al., 2010), it
including soil properties and strength, were described by Ou et al. could be converted to Gref
0 as
(1998, 2000), Kung et al. (2009), and Lim et al. (2010).
!m
c  cos u  r03 sin u
G0 ¼ Gref ð3Þ
3.1.2. Determination of model parameters 0
c  cos u þ pref sin u
The input parameters of the clay layers for the HS and the HSS
models are shown in Table 1. The initial lateral stress state was cal- By following the research results by Tseng (2012), the value of c0:7
culated based on Schmidt (1966), which consider soil stress his- was 104 for Taipei Silty Clay. In addition, Tables 2 and 3 list the
sin u0
tory, that is K 0 ¼ ð1  sin u0 ÞðOCRÞ . In the analyses, an elastic structural properties used for analyses of the TNEC case.
 A. Lim, C.-Y. Ou / Tunnelling and Underground Space Technology 63 (2017) 118–132 121

Table 1
The Hardening soil and the Hardening Soil Small models input parameters for the TNEC case.

Soil layer Depth (m) ct (kN/m3) u0 (°) OCR K0 Eref Eref Eref
ur (kPa) Gref
50 (kPa) oed
(kPa) 0 (kPa)

1-a 0–2 18.25 30 4 1 7033 4923 21,100 69,464

1-b 2–4 18.25 30 4 1 6826 4779 20,479 53,009
1-c 4–5.6 18.25 30 4 1 6631 4642 19,894 48,574
3-a 8–12 18.15 29 1.4 0.59 8521 5964 25,562 55,535
3-b 12–16 18.15 29 1.25 0.56 9526 6668 28,578 52,985
3-c 16–20 18.15 29 1.25 0.56 9620 6734 28,860 49,670
3-d 20–24 18.15 29 1.25 0.56 9290 6503 27,870 47,099
3-e 24–27 18.15 29 1.23 0.55 9049 6335 27,148 45,491
3-f 27–30 18.15 29 1.23 0.55 9158 6410 27,474 44,129
3-h 30–33 18.15 29 1.23 0.55 8935 6255 26,806 42,928
5 35–37.5 19.13 30 1 0.5 10,016 7011 30,049 43,947

Note: m = 1; Rf = 0.9; tur = 0.2; c0:7 = 104.

E ¼ 2  b  Gð1 þ tÞ ð4Þ
where G is the shear modulus of soil; b is a reduction factor which
considers the difference between small strain and the strain at the
normal condition and the b value can be set equal to 0.5 (Ou, 2006).

3.1.3. Finite element mesh and geometry

Fig. 4 displays the finite element mesh and selected stress
points used for analysis, where the left boundary was set at the
center of excavation, considering the symmetry of the excavation,
the right boundary at a distance of 121.5 m from the excavation
center, which was beyond the excavation influence zone, usually
larger than 4 times of the excavation depth and the bottom at
the gravel soil level, i.e., 45 m below the ground surface (around
2 times of the excavation depth). The left and right vertical bound-
aries were restrained from the horizontal movement and the bot-
tom boundary was restrained from both the vertical and
horizontal movements. Five stress points were selected to investi-
gate the effective stress path during excavation. Three points were
located at the retained side (points N, O, and P) and two additional
points were placed inside the excavation zone (points Q and R).

3.1.4. Results and discussion

Fig. 3. Relationship between the small strain shear modulus and the mean effective The computed wall deflections and the ground surface settle-
stress of Taipei Silty Clay (Teng, 2010). ments of the HS model are plotted in Figs. 5 and 6, respectively.
The computed wall deflection at the final stage of excavation was
generally close to the field measurements though it was slightly
Table 2
Dimension and input of the diaphragm walls for analysis. larger at the depth of 20 m below the ground surface. The com-
puted ground surface settlements were still far from the field mea-
Case Ht (m) t (m) E (MPa) t surements. It might be due to the fact that the small strain
TNEC 35 0.9 21 0.15 characteristics of the soil were not considered in the HS model.
Formosa case 31 0.8 21 0.15
Three additional analyses were performed, namely 0.8 Eref
50 , 0.8
Note: Ht = depth of diaphragm wall; t = thickness of diaphragm wall; E = Young’s
Eref ref
oed , and 0.8 Eur . For example, when performing 0.8 Eref
50 , only value
modulus and t = Poisson’s ratio.
ref
of E50 was decreased to 80% of original value while Eref ref
oed and Eur
were kept the original value. As shown in Figs. 5 and 6, when
Meanwhile, according to Ou and Lai (1994), the Young’s modu- reducing Eref ref
50 and Eoed respectively, no significant wall deflections
lus of sand (drained material) can be obtained from the following and ground surface alteration are found as compared to the initial
equation: analysis. However, the computed wall deflections and ground sur-

Table 3
Dimension of lateral support input parameters for the analysis of the TNEC case.

Stage Type A (m2) t (m) s (m) E (GPa) t

1 Steel 0.012 – 8 210 0.15
2 Slab – 0.15 – 21 0.15
3 Slab – 0.15 – 21 0.15
4 Slab – 0.15 – 21 0.15
5 Slab – 0.15 – 21 0.15
6 Steel 0.0219 – 3.4 210 0.15

Note: A = cross section area; t = thickness; s = spacing; E = Young’s modulus and t = Poisson’s ratio.
122 A. Lim, C.-Y. Ou / Tunnelling and Underground Space Technology 63 (2017) 118–132

Fig. 4. Finite element mesh of the TNEC case used for analysis and the location of selected stress points.

Fig. 5. Comparison of measured and computed wall deflections with various degradation of soil modulus parameters in the Hardening Soil Model.

face settlements become greater when Eref toric stresses that meets q ~ ¼ r01 þ ðd  1Þr02  dr03 with
ur decreased to 80% of its
initial value. This indicates that Eref
is a critical input parameter d ¼ ð3 þ sin u0 Þ=ð3  sin u0 Þ. This ellipse has a length of isotropic
ur
pre-consolidation stress ðpc Þ on the p0 -axis and a  pc on the q ~-
that controls the computed wall deflections and the ground surface
settlements. axis, where a is a cap parameter that correlates to K 0NC . In addi-
In order to study the behavior of soil nearby the excavation tion, the mean effective stress ðp0 Þ and the deviatoric stress ðqÞ
area, the soil effective stress path of several points were investi- are defined as p0 ¼ ðr01 þ r02 þ r03 Þ=3 and q ¼ jr01  r03 j, respec-
gated. The effective stress paths of stress points at the retained tively; where r01 is the major principal effective stress, r02 is the
side, namely points N, O, and P, and inside the excavation zone, intermediate principal effective stress, and r03 is the minor princi-
namely points Q and R, were plotted in p0  q and p0  q ~ space. pal effective stress. In some occasions, the out-of-plane horizontal
According to Schanz et al. (1999), the shape of the cap yield surface effective stress that perpendicular to the plane (r0z ) might be smal-
~ space. q
is an ellipse in p0  q ~ is a special stress measure for devia- ler than the in-plane horizontal effective stress (r0x ) and the in-
 A. Lim, C.-Y. Ou / Tunnelling and Underground Space Technology 63 (2017) 118–132 123

Fig. 6. Comparison of measured and computed ground surface settlements with various degradation of soil modulus parameters in the Hardening Soil Model.

Fig. 7. Effective stress paths at the retained and excavated sides of excavation.

plane vertical effective stress (r0y ). Hence, it should be noted that were located in the shear hardening zone. Although they were
r01 and r03 may not be all the in-plane stresses, even under the located in the shear hardening zone, but the characteristics of soil
plane strain condition. It is worth mentioning that the cap yield was still dominated by the elastic behavior because the effective
surface and the initial shear yield surface are governed by the prin- stress path was near vertical. It was proved that the change in
cipal effective stresses not the Cartesian effective stresses. mean effective stress (Dp0 ) was almost zero, due to very small
As illustrated in Fig 7, most of the effective stress paths (points (around 105) generation of plastic volumetric strain (epv ) when
N, O, and P) at the retained side move upward (near vertically) the soil yielded. Under such a condition the elastic shear strain
toward the Mohr-Coulomb failure line, exceeding the initial shear should take a major proportion in the generation of total shear
yield surface and inside the corresponding cap yield surface (at strain because the plastic shear strain is a linear function of the
excavation stage 1–5). For points N and O, they moved downward plastic volumetric strain in the HS model (Schanz et al., 1999).
following the previous loading path (at excavation stages 6 and 7), Furthermore, a point N is selected, which represents the points
meanwhile the effective stress path for point P remained at the retained side, for explaining those paths direction by refer-
unchanged at excavation stages 6 and 7. All of those stress paths ring to Figs. 8 and 9(a). Fig. 8 shows the evolution of the principal
124 A. Lim, C.-Y. Ou / Tunnelling and Underground Space Technology 63 (2017) 118–132

Fig. 8. Evolution of the principal stress rotation at different excavation stages.

(a) (b)

Fig. 9. Principal and Cartesian effective stress paths of (a) Point N and (b) Point R.

effective stress rotation at different excavation stages, while Fig. 9 ment of the soil shear stress may cause the decrease of the effective
illustrates the principal and Cartesian effective stress paths of minor principal stress associated with an increase of major princi-
Points N and R. pal effective stress, referred as the principal effective stress path in
Generally, when soil was excavated (subject to the vertical or Fig. 9a. This is the possible reason of the effective stress paths
lateral unloading force), the pore water pressure nearby excavation direction of soil at the retained side that moves upward as shown
area was decreased owing to the development of the negative in Fig. 7. This observation is consistent with a general effective
excess pore water pressure. In addition, the soil elements moved stress path of idealized excavation case using the finite element
together with the diaphragm wall towards the excavation side, analysis which was conducted by Gebreselassie (2003). But, after
and the shear stress was developed in conjunction with the rota- stage 5, the general trend of the principal effective stress path of
tion of the principal effective stress, as depicted in Fig. 8. As a con- stages 6 and 7 was reversed to paths of stage 1–5, and it would
sequence to the soil in the retained side, when the shear stress not affect much on the effective stress path at p0  q and p0  q ~
induced generation of excess pore water pressure, the drop of hor- space since the stresses difference between the stages 6 and 7 to
izontal total stress was not necessarily equal to the generation of stage 5 only 2–3 kPa.
the negative excess pore water pressure. The development of the As observed at a point R, which was located at the center of
negative excess pore water pressure could be larger than the lat- excavation bottom, the effective stress path in the p0  q and
eral stress release and resulting in the increase of the horizontal ~ space shifts downward at the excavation stages 1–3, then
p0  q
effective stress, referred as the Cartesian effective stress path in experienced a reversal in the direction of stress path in an amount
Fig. 9a. of 180° at the excavation stages 4 until 7, as plotted in Fig. 7. This
For a particular stress point, such as a point N, the amount of the stress path was located below the initial shear yield surface and
principal effective stress rotation was around 10–30°. The develop- inside the corresponding cap yield surface (inside the elastic zone).
 A. Lim, C.-Y. Ou / Tunnelling and Underground Space Technology 63 (2017) 118–132 125

As consequence, the behavior of soil was purely elastic with no P stress path, but with greater shear stress (also inside the shear
plastic deformation were expected to occur. This stress path (Point hardening zone). The Point Q experienced vertical unloading force,
R) also could be explained by referring to Figs. 8 and 9(b). As dis- causing the generation of negative excess pore water pressure rel-
played in Fig. 9b, at the stages 1 and 2, the principal effective stress atively high compared to the point P, at the retained side. In addi-
path moved downward due to the decrease of the major principal tion, the shear stress was developed in conjunction with the
effective stress and unchanged of the minor principal effective rotation of the principal effective stress. In such a condition, it dif-
stress. Then, it reversed by 180° towards the stage 3 and moved fered from the point R, where no shear stress was generated at the
upward until the stage 7. This can be well understood by the fact center of excavation site.
that, the soil at the excavation center was subjected to the vertical Nevertheless, the most significant finding is, for those stress
unloading that caused the effective vertical stress ðr0y Þ gradually to paths which were located inside the shear hardening zone, the
decrease, meanwhile the in-plane effective horizontal stress ðr0x Þ characteristics of soil was still dominated by elastic behavior in
was gradually to increase due to the lateral compression and the which the elastic deformation took a major proportion in control-
out-of-plane effective horizontal stress ðr0z Þ remains unchanged. ling overall deformation. Obviously, the stress path which was
It should be noted that at initial condition, r0x and r0z were identi- inside the elastic zone, the soil characteristic was purely elastic
cal. The detail of the stress changes can be explained as follows. behavior. In such a condition, together with strong findings as
At the earlier stage of excavation (stages 1 and 2), the effective depicted in Figs. 5 and 6, the unloading/reloading modulus should
vertical stress ðr0y Þ became the major principal effective stress ðr01 Þ, be recognized as a critical parameter in controlling deformation
meanwhile the out-of-plane effective horizontal stress ðr0z Þ, behavior of an excavation.
remaining unchanged, was the minor principal effective stress It is worth to mention that those performed analyses only valid
ðr03 Þ. This is the reason of the principal effective stress path move for soil with strain hardening response, like normally and/or lightly
downward vertically. When the soil reached intermediate stage over-consolidated clays. For the strain softening and/or structured
(stages 3 and 4), the in-plane effective horizontal stress ðr0x Þ, kept soils, the more advanced soil models should be adopted, like the
MIT E-3 or Hypoplasticity Clay models. In addition, the excavation
increasing, became the major principal effective stress ðr01 Þ, and
condition and geometry should be an ordinary one, without slop-
the out-of-plane effective horizontal stress ðr0z Þ yet became the
ing ground and/or heavy surcharge. Otherwise, the stress paths
minor principal effective stress ðr03 Þ. The switch of the major prin-
might be different from findings.
cipal stress direction, originally vertical to horizontal (in-plane)
The Mohr-Coulomb model with two approaches, namely, the
direction, caused the rotation of principal effective stress by an
MC Undrained A and the MC Undrained B, were performed with
amount of 90° (see Fig. 8) and the principal effective stress path
the input of Young’s modulus and Poisson’s ratio exactly the same
reversed 180° and moved upward vertically (see Figs. 7 and 9
as unloading/reloading parameters, that are Eur and tur . With the
(b)). When the soil was further excavated to the final stage (stages
MC Undrained A approach, the effective strength parameters were
5–7), the in-plane effective horizontal stress ðr0x Þ yet became the
those used in the HS model. For the MC Undrained B approach, the
major principal effective stress ðr01 Þ and the vertical effective hor-
input of strength parameter was the undrained shear strength,
izontal stress ðr0y Þ, kept decreasing, became the minor principal
which was obtained from a series of the field vane shear test data
effective stress ðr03 Þ. In such circumstances, it is clearly seen that and the laboratory CK0UC and CK0UE tests (Hsieh and Ou, 2012; Ou
not only the effective vertical stress ðr0y Þ and the in-plane effective et al., 1998). Tables 4 and 5 list the input parameters of undrained
horizontal stress ðr0x Þ influenced the effective stress path direction, and drained materials used for the Mohr-Coulomb model analyses,
but also the out-of-plane effective horizontal stress ðr0z Þ has a sig- respectively.
nificant contribution to the direction of effective stress path in As featured in Fig. 10, the computed wall deflections obtained
p0  q and p0  q ~ space. from the MC Undrained A and Undrained B diverse each other.
The point Q, at the excavated site, was located in front of the At the final stage of excavation, the analysis results showed that
diaphragm wall. The general trend of path is similar with the Point the MC Undrained B approach agree well with the computed

Table 4
The Mohr-Coulomb model input parameters of undrained material for the TNEC case.

Soil layer Depth (m) ct (kN/m3) OCR K0 Eur (kPa) tur Su =r0v Su (kPa)

1-a 0–2 18.25 4 1 3851 0.2 0.25 4.6

1-b 2–4 18.25 4 1 9203 0.2 0.25 11.4
1-c 4–5.6 18.25 4 1 11,963 0.2 0.25 15.3
3-a 8–12 18.15 1.4 0.59 15,947 0.2 0.27 28.5
3-b 12–16 18.15 1.25 0.56 22,176 0.2 0.27 37.5
3-c 16–20 18.15 1.25 0.56 27,777 0.2 0.27 46.5
3-d 20–24 18.15 1.25 0.56 32,022 0.2 0.27 55.5
3-e 24–27 18.15 1.23 0.55 35,336 0.2 0.27 63.4
3-f 27–30 18.15 1.23 0.55 39,571 0.2 0.27 70.1
3-h 30–33 18.15 1.23 0.55 42,328 0.2 0.37 105.4
5 35–37.5 19.13 1 0.5 49,361 0.2 0.47 154.4

Table 5
Input parameters of drained material of the Mohr-Coulomb model in the TNEC case.

Soil layer Depth (m) SPT-N ct (kN/m3) K0 u0 (deg) E0 (kPa) t0

2 5.6–8 7 18.93 0.5 30 68,351 0.3
4 33–35 23 19.62 0.45 33 265,473 0.3
6 37.5–45 26 19.62 0.43 35 300,247 0.3
126 A. Lim, C.-Y. Ou / Tunnelling and Underground Space Technology 63 (2017) 118–132

Fig. 10. Comparison of measured wall deflections with computations from several soil constitutive models of the TNEC case.

results from the HS model, even closer to the field measurements. struts were installed as the lateral supports. Fig. 12 depicts the
The MC Undrained A resulted smaller wall deflections as compared excavation depths, locations of the temporary struts and the soil
to those from the MC Undrained B. because the MC Undrained A layer profiles. Compared to the TNEC soil profiles, the Sungshan
over-estimates the undrained shear strength of the clay soil. In a III, silty sand, disappears at this site, and Sungshan II and IV merge
normally consolidated clay, the MC Undrained A approach obvi- into a single layer of silty clay that is located above GL 25 m,
ously over-estimates the undrained shear strength and under- between GL 25 and 31 m is Sungshan I, which is characterized
estimates the excess pore pressure. In order to overcome this issue, as the silty sand. Below the silty sand, GL 31 m, is the mudstone.
the MC Undrained B approach can be used as an alternative solu- The ground water level was located 3 m below the ground surface
tion. The MC Undrained B approach forces soil to fail at a specified and the pore water pressure is assumed hydrostatic.
undrained shear strength, which can be easily determined from The finite element mesh for the analysis following that dis-
various tests such as field vane shear tests or triaxial tests. Hence, played in Fig. 4. The left vertical boundary was set at the middle
in addition to the unloading/reloading parameters for the Mohr- of the excavation site, which was 17.5 m from the diaphragm wall,
Coulomb model, the undrained shear strength is also a critical the right boundary was also 100 m from the diaphragm wall
parameter in controlling the deflection of diaphragm wall. (around 5 times of excavation depth), and the bottom boundary
The computed ground surface settlements of soil models, as was at the mudstone, i.e., 31 m below the ground surface (around
shown in Fig. 11, did not agree well with the field measurements, 2 times of excavation depth).
especially at initial and middle stages of excavation. For the final Similar with the TNEC case, the structural members were
stage of excavation, they were generally close to field measure- assumed to behave as linear-elastic. The input parameters for the
ments. The results of deformation analysis were generally consis- structural elements are listed in Tables 2 and 6. In the analyses,
tent with Orazalin et al. (2015). the clay layers were modelled with the HS model, the HSS model,
and the MC Undrained B models, while the silty sand layer was
3.2. The Formosa excavation project simulated with the MC drained analysis. All of the input parame-
ters for those models were adopted from the TNEC case because
3.2.1. The Formosa project and analysis description these both cases were in the same geological subzone and were
In this section, the Formosa excavation case (Ou et al., 1993; close to each other. Tables 7 and 8 display the input parameters
Hsieh and Ou, 1998) was used to validate the consistency of the of the Formosa case for the HS, HSS, and the MC models,
procedure of parameter determination that established in the pre- respectively.
ceding section. The Formosa case has a distance of around 2 km
from the TNEC case and infest a similar geological subzone. The 3.2.2. Result and discussion
Formosa case was a 30-story reinforced concrete building with As shown in Fig. 13, the computed wall deflections from the HS
four basement levels. The excavation site occupied 2311 m2 in and the MC Undrained B models agreed well with the field mea-
rectangular shape; about 75 m in length and 35 m in width. The surements at the middle and final stages of the excavation, though
final excavation level was 18.45 m, and a diaphragm wall with the MC Undrained B yielded slightly greater wall deflections com-
31 m in depth and 0.8 m in thickness, was used to retain the soil pared to the others. Fig. 14 shows the comparison of the computed
behind the diaphragm wall. The excavation was completed using ground surface settlements and the results of field measurements.
the bottom up construction method, in which the temporary steel The computed ground surface settlements trend were generally
 A. Lim, C.-Y. Ou / Tunnelling and Underground Space Technology 63 (2017) 118–132 127

Fig. 11. Comparison of measured ground surface settlements with computations from several soil constitutive models of the TNEC case.

close to the field measurements. Moreover, the investigated effec-

tive stress paths of the HS model showed that most of the stress
paths have similar trend with the stress path of the TNEC case as
shown in Fig. 7. In general, the deformation characteristics
between the TNEC and the Formosa cases were consistent each
other.

3.3. The ‘‘Deep Clay” excavation: hypothetical case

Besides the HS and the MC model, the HSS model was also
applied to analyze the deformation characteristics of the TNEC
and the Formosa case histories, in which the characteristics of
small strain in the soil has been considered in the HSS model.
The model parameters of the TNEC and the Formosa cases are
listed on Tables 1 and 7, respectively. In general, the performance
of the HSS model on wall deflection computation was sound. As
depicted in Figs. 10 and 13, the computed wall deflections of the
HSS model for the TNEC and Formosa excavation cases were close
to field measurements at the middle and final stages of excava-
tions. However, The HSS model only yielded slight improvement
on the computed ground surface settlements of the TNEC and For-
mosa excavation cases, compared to the others, though the small
strain characteristic of soil is considered, as depicted in Figs. 11
Fig. 12. Profile of the Formosa case excavation sequence and subsurface soil layers. and 14. The similar results were also observed by Mu and Huang

Table 6
Dimension of lateral support input parameters for the analysis of the Formosa case.

Stage Type A (m2) s (m) E (GPa) t

1 Steel 0.012 3.3 210 0.15
2 Steel 0.0174 3.3 210 0.15
3 Steel 0.0219 3.3 210 0.15
4 Steel 0.0348 3.3 210 0.15
5 Steel 0.0438 3.3 210 0.15
6 Steel 0.0438 3.3 210 0.15

Note: A = cross section area; s = spacing; E = Young’s modulus and t = Poisson’s ratio.

Table 7
The Hardening Soil and Hardening Soil Small model input parameters of the Formosa case.

Soil layer Depth (m) ct (kN/m3) u0 (deg) OCR K0 Eref Eref Eref
ur (kPa) Gref
50 (kPa) oed
(kPa) 0 (kPa)

1-a 0–2 17.6 30 4 1 7934 5554 23,802 70,152

2-a 2–6 17.6 30 1.25 0.56 10,170 7119 30,510 67,865
2-b 6–10 17.6 30 1.25 0.56 10,319 7224 30,958 59,899
2-c 10–14 17.6 30 1.25 0.56 10,469 7328 31,407 54,858
2-d 14–18 17.6 30 1.25 0.56 10,369 7259 31,108 51,257
2-e 18–22 17.6 30 1.23 0.55 10,209 7146 30,626 48,746
2-f 22–25 17.6 30 1.23 0.55 9758 6831 29,275 46,776

Note: m = 1; Rf = 0.9; tur = 0.2; c0:7 = 104.

128 A. Lim, C.-Y. Ou / Tunnelling and Underground Space Technology 63 (2017) 118–132

Table 8
The Mohr-Coulomb model input parameters of the Formosa case.

Soil layer Depth (m) ct (kN/m3) OCR K0 Eur (kPa) tur Su =r0v Su (kPa) u0 (deg)
1-a 0–2 17.6 4 1 4203 0.2 0.25 4.4 –
2-a 2–6 17.6 1.25 0.56 10,375 0.2 0.27 16.4 –
2-b 6–10 17.6 1.25 0.56 15,962 0.2 0.27 24.9 –
2-c 10–14 17.6 1.25 0.56 21,706 0.2 0.27 33.4 –
2-d 14–18 17.6 1.25 0.56 26,959 0.2 0.27 41.9 –
2-e 18–22 17.6 1.23 0.55 31,662 0.2 0.27 50.3 –
2-f 22–25 17.6 1.23 0.55 34,725 0.2 0.27 57.8 –
Drained material
3 25–31 19.72 – 0.45 192,367 0.3 – – 33

Fig. 13. Comparison of measured wall deflections with computations from several soil constitutive models of the Formosa case.

Fig. 14. Comparison of measured ground settlements with computations from several soil constitutive models of the Formosa case.
 A. Lim, C.-Y. Ou / Tunnelling and Underground Space Technology 63 (2017) 118–132 129

Fig. 15. Geometry of the hypothetical Deep Clay case.

Fig. 16. Comparison of wall deflections curves of the Deep Clay case.

(2016), in which the inverse analysis has been adopted to improve and minimum width of geometry size were three times and two
the calculation results of the HSS model. In addition, Johansson and times (a part from the excavation width) of the excavation depth,
Sandeman (2014) also concluded that it is difficult to see any respectively. In addition, if non-small strain soil model was applied
advantages of using the HSS model rather than the other models. (i.e., the MC or the HS models), they recommended a layer of small
Hence, the Deep Clay case was assumed to investigate the com- strain stiffness zone, which has an equal thickness with the exca-
puted ground surface settlements of the soil models. vation depth, need to be introduced at the bottom of model geom-
etry for detain the effect of small strain on ground settlement
3.3.1. Description of the ‘‘Deep Clay” hypothetical case computation.
Vermeer and Wehnert (2005) proposed a minimum size of a As shown in the computed ground settlement of case studies
finite element analysis model geometry in performing the finite (Figs. 11 and 14), none of the selected soil models can satisfied
element deformation analysis. They suggested the minimum depth the result of field measurements well, including the HSS model
130 A. Lim, C.-Y. Ou / Tunnelling and Underground Space Technology 63 (2017) 118–132

Fig. 17. Comparison of ground surface settlements of the Deep Clay case from the hardening soil model and the Hardening Soil Small model.

where the small strain characteristic of soil is considered. Both of stiffness layer was considered, only the unloading/reloading mod-
case studies satisfied the minimum width of geometry size, but ulus of that layer was increased by factor of 5 from the initial value
they did not satisfy the minimum depth of geometry size because (see Fig. 15). The construction sequences and the parameters of the
the hard stratum was existed prior to the three times of excavation structural members of this hypothetical case mimic the simulation
depth. Thus, the geometry size constraint of performed analyses of the TNEC excavation case.
(i.e., the TNEC and the Formosa cases) might be the possible reason
for the accuracy of computed ground settlement.
3.3.2. Result and discussion
The objective of conducting the analysis of the hypothetical
In the Deep Clay case, the computed wall deflections and
case, namely Deep Clay, is to investigate the performance of those
ground surface settlements from the HSS model are set as the ref-
three soil models in computing the ground surface settlements if
erence values due to its capability of modelling the small strain
the minimum size of model geometry was fulfilled and a small
characteristic of soil. The small strain zone geometry (see Fig. 15)
strain zone was introduced in the model geometry. In these simu-
was not be applied to the HSS model with which only the original
lations, two types of model geometry were performed, namely the
geometry was used. As shown in Fig. 16, with considering small
original geometry and the zone of small strain stiffness geometry.
strain zone, the computed wall deflections from the MC Undrained
When the zone of small strain stiffness geometry was performed, a
B and the HS models are identical with those from the HSS model.
layer of small strain stiffness zone, which has an equal thickness
This denotes that the small strain zone has no effect in the com-
with the excavation depth was introduced at the bottom of model
puted wall deflections. On the contrary, with the consideration of
geometry. Fig. 15 depicts the typical geometry of the Deep Clay
the small strain zone, it could improve the computed ground sur-
case. The model geometry was 80 m in depth and 160 m in width.
face settlements for both the HS model and the MC Undrained B
The homogenous clay soil was assumed to have the same soil prop-
models, as plotted in Figs. 17 and 18, respectively. The computed
erties as Sungshan IV layer in the TNEC case histories. Likewise
ground surface settlements from the MC Undrained B with consid-
previous analyses, the HS model, the HSS model, and the MC
ering the small strain zone geometry was close to those from the
Undrained B model were used to model the clay soil with the iden-
HSS model at the final stage of excavation, but not at the initial
tical input soil properties as Sungshan IV layer. If the small strain
and middle stages. Even though the improvement only appeared

Fig. 18. Comparison of ground surface settlements of the Deep Clay case from the Mohr Coulomb Undrained B model and the Hardening Soil Small model.
 A. Lim, C.-Y. Ou / Tunnelling and Underground Space Technology 63 (2017) 118–132 131

at the final stage of excavation, it still becomes a great benefit, Acknowledgements

especially for the MC model users. Hence, the MC Undrained B
model could be respected as an ideal soil model for practical appli- The authors would like to thank the Ministry of Science and
cation, especially when performing a deformation analysis of deep Technology – Republic of China for providing financial support
excavation in the normally or lightly over-consolidated clays, with through contract No. NSC 101-2221-E-011-110-MY3.
or without considering the effect of the small strain, where the
input unloading/reloading modulus, unloading/reloading Poisson’s References
ratio and the undrained shear strength parameters are well
determined. Benz, T., Vermeer, P.A., Schwab, R., 2009. A small-strain overlay model. Int. J. Numer.
Anal. Meth. Geomech. 33, 25–44.
Brinkgreve, R. et al. (Eds.), 2010. Plaxis 2D 2010 – User’s Manual. Plaxis BV,
Netherlands.
4. Conclusions
Calvello, M., Finno, R., 2004. Selecting parameters to optimize in model calibration
by inverse analysis. Comp. Geotech. 31, 410–424.
Based on the finite element analyses, field monitoring data of Chen, R., Meng, F., Li, Z., Ye, Y., Ye, J., 2016. Investigation of response of metro
tunnels due to adjacent large excavation and protective measures in soft soils.
case histories, and parameters deduced from laboratory and in-
Tunnel. Undergr. Space Technol. 58, 224–235.
situ tests, the conclusions can be drawn as following: Corral, G., Whittle, A.J., 2010. Re-analysis of deep excavation collapse using a
generalized effective stress soil model. In: Proceedings of the 2010 Earth
1. For the Hardening Soil model, most of the soils inside and out- Retention Conference. Geotechnical Special Publications, Bellevue, WA, USA.
GSP 208.
side excavation zone were in the elastic behavior. Even though Gebreselassie, B., 2003. Experimental, Analytical and Numerical Investigations of
the effective stress path of soils adjacent to the diaphragm wall Excavations in Normally Consolidated Soft Soils (Dissertation). University of
have undergone yield, but the characteristics of those soils were Kassel.
Ho, I.H., Hsieh, C.C., 2013. Numerical modeling for undrained shear strength of clays
still dominated by the elastic behavior. Hence, the unloading/ subjected to different plasticity indexes. J. GeoEng. 8, 91–100.
reloading parameters are predominant in a deformation analy- Hong, Y., Ng, C.W.W., Liu, G.B., Liu, T., 2015. Three-dimensional deformation
sis of an excavation case. For elastic perfectly plastic soil model, behaviour of a multi-propped excavation at a ‘‘greenfield’’ site at Shanghai soft
clay. Tunnel. Undergr. Space Technol. 45, 249–259.
like the Mohr-Coulomb model, when the effective stress Hsieh, P.G., Ou, C.Y., 1998. Shape of ground surface settlement profiles caused by
undrained analysis was performed, the strength parameters excavation. Can. Geotech. J. 35, 1004–1017.
should be expressed in terms of the undrained shear strength Hsieh, P.G., Ou, C.Y., 2012. Analysis of deep excavations in clay under the undrained
and plane strain condition with small strain characteristics. J. Chin. Inst. Eng. 35,
ðc ¼ Su ; u ¼ 0Þ and the unloading/reloading parameters (Eur
601–616.
and tur ) should be adopted in order to obtain the reasonable Hsieh, P.G., Ou, C.Y., Lin, Y.K., Lu, F.C., 2015. Lesson learned in design of an
computed wall deflections; as accurate as those from the Har- excavation with the installation of buttress walls. J. GeoEng. 10 (2), 63–73.
Hwang, R.N., Lee, T.Y., Chou, C.R., Su, T.C., 2012. Evaluation of performance of
dening Soil model.
diaphragm walls by wall deflection paths. J. GeoEng. 7, 1–12.
2. The effect of small strain stiffness in the computed ground sur- Johansson, E., Sandeman, E., 2014. Modelling of a Deep Excavation: A Comparison of
face settlements can be distinguishable if the soft soil was deep Different Calculation Methods to In-Situ Measurements (Thesis). Chalmers
enough (at least 3 times of excavation depth). The computed University of Technology.
Kung, T.C., Ou, C.Y., Juang, C.H., 2009. Modeling small-strain behavior of Taipei clays
ground surface settlements from the Mohr-Coulomb model for finite element analysis of braced excavations. Comput. Geotech. 36, 304–319.
and the Hardening Soil model were close to the computed Lim, A., Ou, C.Y., Hsieh, P.G., 2010. Evaluation of clay constitutive models for
ground surface settlements from the Hardening Soil Small analysis of deep excavation under undrained conditions. J. GeoEng. 5, 9–20.
Mašín, D., 2005. A hypoplastic constitutive model for clays. Int. J. Numer. Anal.
model if the small strain layer (as thick as the excavation depth) Meth. Geomech. 29, 311–316.
was realized at lower part of the model geometry. Mašín, D., Boháč, J., Tůma, P., 2011. Modelling of a deep excavation in a silty clay. In:
3. The effective vertical stress ðr0y Þ or the in-plane effective hori- Anagnostopoulos, Pachakis, Tsatsanifos (Eds.), International Proceedings of 15th
European Conference on Soil Mechanics and Geotechnical Engineering, vol. 3.
zontal stress ðr0x Þ was not necessary the major or minor princi- Athens, Greece, pp. 1509–1514.
pal effective stress. In several circumstances, the out-of-plane Mašín, D., 2014. Clay hypoplasticity model including stiffness anisotropy.
effective horizontal stress ðr0z Þ could be the major or minor Geotechnique 4, 232–238.
Mu, L., Huang, M., 2016. Small strain based method for predicting three-
principal effective stress, as well. Hence, r0z also has a significant dimensional soil displacements induced by braced excavation. Tunnel.
contribution to the direction of effective stress path in p0  q Undergr. Space Technol. 52, 12–22.
and p0  q ~ space. Ng, C.W.W., Lings, M.L., 1995. Effects on modeling soil nonlinearity and wall
installation on back-analysis of deep excavation in stiff clay. J. Geotech. Eng.
4. The soil at the retained side moved together with the dia- Geoenviron. Eng. 121 (10), 687–698.
phragm wall towards excavation side, and the rotation of prin- Orazalin, Z.Y., Whittle, A.J., Olson, M.B., 2015. Three-dimensional analyses of
cipal effective stress was in conjunction with the development excavation support system for the stata center basement on the MIT campus. J.
Geotech. Eng. Geoenviron. Eng. 141 (7), 05015001.
of the shear stress. As a consequence, the induced negative Ou, C.Y., Hsieh, P.G., Chiou, D.C., 1993. Characteristics of ground surface settlement
excess pore water pressure was not necessary equal to the drop during excavation. Can. Geotech. J. 30, 758–767.
of the horizontal total stress. In such a condition, the develop- Ou, C.Y., Lai, C.H., 1994. Finite element analysis of deep excavation in layered sandy
and clayey soil deposits. Can. Geotech. J. 31, 204–214.
ment of the negative excess pore water pressure could be larger Ou, C.Y., Liao, J.T., Lin, H.D., 1998. Performance of diaphragm wall constructed using
than the lateral stress release and resulting in the increase of top-down method. J. Geotech. Eng. Geoenviron. Eng. 124, 798–808.
the horizontal effective stress. Ou, C.Y., Liao, J.T., Cheng, W.L., 2000. Building response and ground movements
induced by a deep excavation. Geotechnique 50, 209–220.
5. The soil at the excavation center was subjected to vertical Ou, C.Y., 2006. Deep Excavation-Theory and Practice. Taylor and Francis.
unloading that caused the effective vertical stress ðr0y Þ to Powrie, W., Pantelidou, H., Stallebrass, S.E., 1998. Soil stiffness in stress paths
decrease, meanwhile the in-plane effective horizontal stress relevant to diaphragm walls in clay. Geotechnique 48, 483–484.
Schanz, T., Vermeer, P.A., Bonnier, P.G., 1999. The hardening soil model: formulation
ðr0x Þ was gradually to increase due to the lateral compression and verification. In: Beyond 2000 in Computational Geotechnics – 10 Years
and the out-of-plane effective horizontal stress ðr0z Þ remains Plaxis. Rotterdam, pp. 281–296.
unchanged. Whilst the in-plane effective horizontal stress was Schmidt, B., 1966. Earth pressure at rest related to stress history, Discussion. Can.
Geotech. J. 3 (4), 239–242.
greater than the vertical effective stress and the out-of-plane Shi, J., Liu, G., Huang, P., Ng, C.W.W., 2015. Interaction between a large-scale
effective horizontal stress and without development of shear triangular excavation and adjacent structures in Shanghai soft clay. Tunnel.
stress, the rotation angle of principal effective stress was 90°. Undergr. Space Technol. 50, 282–295.
Surarak, C., Likitlersuang, S., Wanatowski, D., Balasubramaniam, A., Oh, E., Guan, H.,
These phenomena will cause the reversal of effective stress path 2012. Stiffness and strength parameters for hardening soil model of soft and
direction in an amount of 180°. stiff Bangkok Clays. Soils Found. 52 (4), 682–697.
132 A. Lim, C.-Y. Ou / Tunnelling and Underground Space Technology 63 (2017) 118–132

Teng, F.C., 2010. Prediction of Ground Movement Induced by Excavation Using the Whittle, A.J., Kavvadas, M.J., 1994. Formulation of the MIT-E3 constitutive model for
Numerical Method with Consideration of Inherent Stiffness Anisotropy overconsolidated clays. J. Geotech. Eng. 120, 173–199.
(Dissertation). National Taiwan University of Science and Technology. Whittle, A.J., Davies, R.V., 2006. Nicoll highway collapse: evaluation of geotechnical
Tseng, T.Y., 2012. Investigation of Stiffness Paramaters of Effective Stress Undrained factors affecting design of excavation support system. In: Proceeding of
Analysis in Deep Excavation (Thesis). National Taiwan University of Science and International Conference on Deep Excavations, Singapore.
Technology. Zhang, E., Goh, A.T.C., Xuan, F., 2015. A simple prediction model for wall deflection
Vermeer, P.A., Wehnert, M., 2005. Beispiele von FE-Anwendungen – Man lernt nie caused by braced excavation in clays. Comp. Geotech. 63, 67–72.
aus. In: FEM in der Geotechnik – Qualität, Prüfung, Fallbeispiele, Hamburg, pp. Zhao, J., Künzli, O., 2016. An introduction to connectivity concept and an example of
101–119. physical connectivity evaluation for underground space. Tunnel. Undergr. Space
Wang, L., Juang, C.H., Atamturktur, S., Gong, W., Khoshnevisan, S., Hsieh, H.S., 2014. Technol. 55, 205–213.
Optimization of design of supported excavations in multi-layer strata. J.
GeoEng. 9, 1–12.