Journal of GeoEngineering, Vol. 11, No. 1, pp.

1-12, April 2016 Chang-Yu Ou: Finite Element Analysis of Deep Excavation Problems 1
http://dx.doi.org/10.6310/jog.2016.11(1).1

2015 TGS Geotechnical Lecture:

FINITE ELEMENT ANALYSIS OF DEEP EXCAVATION PROBLEMS
Chang-Yu Ou 1

ABSTRACT
The finite element method is a powerful tool for the analysis of deep excavation problems. Since soil is a highly nonlinear,
plastic and porewater pressure dependent material, to realistically predict the deformation of soil, rational formulation of
constitutive models and flow rules are necessary. This paper presents the commonly used constitutive soil models suitable for the
analysis of deep excavation problems where the input parameters can be derived from the basic soil properties in the effective
stress drained and undrained analyses. Moreover, the total stress undrained analysis was an alternative for the undrained clay
where the stiffness of the undrained clay has to be obtained from empirical formulae. The finite element method can also be used
for the analysis of stability problems. With the inclusion of elastic-plastic properties of structure members in the analysis and
simulation of center posts in excavations, the finite element analysis can result in good estimate the factor of safety of excavations.
Finally the three dimensional finite element analysis of excavations with buttress walls was presented to demonstrate its
applicability. It is found that the wall deflection and ground settlement for excavations with buttress walls can be well predicted.
The mechanism of buttress walls in reducing the lateral wall deflection can also be understood. Design of buttress walls in
excavations would be more reasonable.

Key words: Effective stress undrained analysis, total stress undrained analysis, stress paths, deformation analysis, stability analy-
sis, three dimensional analysis.

1. INTRODUCTION forces is influenced by many factors: the conditions of soil layers,

the level and pressures of groundwater, the excavation depth, the
Design of an excavation necessarily considers its stability
excavation width and so on. The finite element method is capable
and serviceability. The former concerns the safety of the excava-
of simulating these factors and therefore the results derived from
tion system and the latter aims to avoid the damage of adjacent
the method would be more accurate than those derived from sim-
properties such as buildings and public utilities. In the past sev-
ple stability formulae and empirical methods. Therefore, the fi-
eral simple formulae for stability analysis (e.g., Terzaghi 1943;
nite element method has been applied to the analysis of deep
Bjerrum and Eide 1956; JSA 1988; Tschebotarioff 1951; Hsieh et
excavation problems quite extensively.
al. 2008) have been developed based on theory of bearing capac-
It is generally understood that the accuracy of the finite ele-
ity or limiting equilibrium, with little consideration of the entire
ment method depends on the formulation of constitutive models
structural system. Moreover, many empirical methods (e.g.,
and selection of soil parameters. Since the stress strain behavior
Bowles 1986; Clough and O’Rourke 1990; Hsieh and Ou 1998;
of soil is highly non-linear, anisotropic, porewater pressure de-
Osman and Bolton 2006; Ou and Hsieh 2011; Hsieh and Ou 2016;
pendent, and plastic, the formulation of constitutive models is
Peck 1969) have also been derived to estimate the amount of
usually very complicated. Moreover, the selection of soil param-
deflection of retaining walls and ground settlement. Those meth-
eters (e.g., strength and stiffness parameters) is a crucial step in
ods were basically derived from the monitoring results of exca-
the finite element analysis. A common problem in the analysis of
vation case histories, representing the effects of every relevant
deep excavations is that the soil tests data are often limited or of
element on deformation. Normally it can lead to effective predic-
low-quality due to the difficulty in taking undisturbed in situ
tions, without much complexity, for similar excavation projects,
samples.
in terms of soil conditions, construction methods, and engineer-
The objective of this paper is to present the commonly used
ing design. Nevertheless, those predictions are only applicable to
analysis methodology of the finite element method including
normal excavations. For some excavations with special designs
total stress analysis and effective stress analysis for clay, consti-
such as ground improvement, installation of cross walls and but-
tutive model, deformation analysis and stability analysis. Finally,
tress walls, or mixed top-down with bottom-up excavations, the
the three dimensional (3D) analysis will be elucidated to address
prediction from empirical methods may lead to inaccurate results.
the importance of the 3D behavior of excavations.
The stability and movements of an excavation are engen-
dered by unbalanced forces acting on the wall due to the removal
of soils within the excavation zone. The magnitude of unbalanced
2. EFFECTIVE AND TOTAL STRESS
Manuscript received March 10, 2016; revised March 14, 2016; ac- ANALYSES
cepted March 15, 2016.
1
Professor (corresponding author), Department of Civil and Con-
struction Engineering, National Taiwan University of Science and In the geotechnical analysis the geomaterial must be classi-
Technology, Taipei, Taiwan (e-mail: [email protected]). fied as either drained or undrained material. Drained material
2 Journal of GeoEngineering, Vol. 11, No. 1, April 2016

certainly needs to be analyzed with effective stress analysis in q

which excess porewater pressure is set equal to 0. There are two Kf compression line
types of undrained analysis, namely, total stress undrained analy-
sis and effective stress undrained analysis. A
In the total stress undrained analysis, the soil and water are A2 B
treated as a single material and the ground water level is set at the
bottom of the finite element mesh to avoid the influence of '
A1 A1
'
ground water on the soil and structure in the analysis. All the A2 '
parameters of soil should be expressed in terms of total stress, for B0
A0
'
A0 B0
example, strength parameters (c,   0), undrained Young’s
modulus (Eu) and undrained Poisson’s ratio (  0.495). The
'
coefficient of the at-rest earth pressure is equal to 1.0. The pa- B2 B1
'
p' , p
B1
rameters of the material used in the analysis are not intrinsic
properties because they are affected by the development of ex- Total Stress Path
cess porewater pressure that does not appear in the analysis. B2 Effective Stress Path
Therefore, the parameters derived or measured from laboratory
or field tests should be with the same stress path as those in the
Kf extension line
field condition to ensure a similar development of excess
porewater pressure. Although the strength parameters can be
measured from the unconsolidated undrained test or vane shear
test, the undrained Young’s modulus can only be estimated from
Fig. 1 Possible stress paths in excavations (Ng 1999)
empirical formulae.
In the effective stress undrained analysis, soil and water are
treated as two phases and the actual location of ground water
level should be set. All the parameters of soil are expressed in undrained material can be conducted with effective stress un-
terms of effective parameters, for example, effective strength drained analysis or total stress undrained analysis. The common-
parameters (c and ), effective Young’s modulus (E) and effec- ly used effective stress model and total stress undrained model
tive Poisson’s ratio (), etc., The coefficient of the at-rest earth will be introduced in the following sections.
pressure should be conformed with the geostatic condition. Most
importantly, the stresses computed from the finite element meth- 3.1 Effective Stress Model
od are affected by the generation of excess porewater pressure Diaz-Rodriguez et al. (1992) studied the yield behavior of
that can be computed from the finite element computation pro- natural clays in different areas, in which the effective friction
cedure. The procedure for the effective stress undrained analysis
angle ranges from 17.5 to 43 (Fig. 2). The shape of the yield
can be summarized as
surface should be strongly influenced by the geological formation
1. Input the effective stress parameters (e.g., c, , E and ), and constitutions of soil because it changes from areas to areas.
compute the constitutive stiffness matrix of soil [D’] As shown in Fig. 2, when values of the friction angle are low, the
2. Input water stiffness, compute the constitutive stiffness ma- yield surfaces are in a flat shape. With the increase of the friction
trix of porewater [Df] angle, the yield surfaces rotate upward and become thick and
3. Compute [D] [D’] [Df], then form the element stiffness chunky. If K0  1 sin is assumed, the yield surface seems to
matrix [KE] be asymmetric with respect to the K0 line because the upper part
4. Assemble the [KE] into global stiffness matrix [KG] of the yield surfaces above the K0 line appears to be fatter than
5. Use the standard finite element procedure [R] [KG][dn] to the lower part. However, for simplification, the elliptical shape of
compute nodal displacements [dn] where [R] is nodal forces the yield surface, symmetric with respect to the K0 line, is often
6. Through {} [B][dn] to compute strain in an element {} assumed for most of constitutive models of soil.
7. Through {} [D’] {} to obtain the effective stress {} Figure 3 compares the yield surface of the typical Taipei
silty clay, as established by Chin et al. (1994), with those pre-
8. Through {f} [Df] {} to yield the porewater pressure {f}
dicted by the Modified Cam-clay (abbreviated as MCC, Roscoe
and Burland 1968), MIT S1 (Pestana and Whittle 1999), RANI
3. CONSTITUTIVE SOIL MODEL AND ITS (Ou et al. 2011), and S-CLAY1 (Wheeler et al. 2003) models, in
which the effective friction angles are assumed to be 30. All of
APPLICATION the yield surfaces are elliptical in shape. The MCC model obvi-
ously predicts the yield stress far from the test data. The yield
In an excavation, the stress alteration associated with verti-
cal and horizontal stress relief is very complex. It depends on stresses predicted by the MIT S1, RANI and S-CLAY1 models
several factors, such as type of retaining wall, excavation depth, are generally in good agreement with the test data above the K0
stress history of soil, etc. Figure 1 illustrates possible stress paths line and the S-CLAY1 model obtains the yield stress closer to the
in an excavation from the result of finite element analysis (Ng right part of the test data.
1999). As we understand, the stress strain response of soil is The MCC model is based on the critical state theory and
stress path dependent and to realistically simulate the defor- originally means to simulate the behavior of normally and near-
mation behavior of soil, good constitutive models are certainly normally consolidated clays under triaxial compression test condi-
required. As mentioned in the proceeding section, the behavior of tions. The yield surface of the MCC model in the p-q stress
 Chang-Yu Ou: Finite Element Analysis of Deep Excavation Problems 3

0.6 ments for intermediate to final stages. This can be explained by the
fact that the MCC yield surface is symmetric to the hydrostatic line
and the real soil yield surface is symmetric to the about K0 line
(Fig. 3). At the early stages, the soil is subjected to small unload-
0.4 ing force, causing the stress state to be the inside of the MCC
 '  43 yield surface, and meanwhile also the inside of the real soil yield
 1   3
2 vm 
surface as predicted by S-CLAY1 model. The deformation be-
35  0
32  ~3 havior predicted by the MCC model thus matches the real be-
28
0.2 
7 havior. This is why the wall displacement prediction curves are
 ~2 close to the field measurements at early stages. When excavation
25
23 17.5

advances deeper, i.e., intermediate to final stages, the unloading
force was large enough to cause the stress state of the soil to be in
0 the plastic state, i.e., path A-B-C-E and relative large defor-
0 0.2 0.4 0.6 0.8 1.0 mation occurs (Fig. 3). However, the stress state predicted from
 1   3 the model may be still inside the MCC yield surface. Hence, the
2 vm computed wall displacements are smaller than the field meas-
urements.
Fig. 2 Yielding surfaces of natural soils (Diaz-Rodriguez et al. To obtain better analysis results, the parameter , directly
1992) related to movements induced by unloading force, should be ad-
justed. Considering that use of the MCC model may cause the
normally consolidated soil in front the wall to be in the plastic
state while the overconsolidated soil may be still in the elastic
state, we therefore raise the -value from the original value
CSL (0.1 0.15) to 0.25 for the soil at the depth of 12 ~ 37.5 m
0.8 Taipei silty
clay yielding (normally consolidated clay) while the -value of the soil at the
region
depth of 0 to 12 m (overconsolidated clay) remains unchanged.
The rest of the input parameters remain to be unchanged. Figure
5 shows the comparison of measured wall displacements and
q /  vc

0.4 Modified Cam - Clay those computed using the MCC model with the adjusted parame-
ters. The computed wall displacements generally agreed with the
field measurements.
MIT S 1
Both Figs. 4 and 5 show that the computed surface settle-
ments were much smaller than the field measurements for the soil
0
RANI
S- CLAY 1 near the wall but larger than the field measurements for the soil
far away from the wall. This is attributed to the fact that the small
strain characteristics are not considered in the MCC model.
The Hardening Soil model (abbreviated as the HS model,
0.4 Schanz et al. 1999), is a true second order model for soils in gen-
0 0.2 0.4 0.6 0.8
eral. The HS model requires 9 parameters, i.e., three reference
p  /  vc stiffness parameters ( E50 ref
for triaxial compression, Eurref for
ref
triaxial unloading/reloading or elastic Young’s modulus, Eoed
Fig. 3 The yield surface of the Taipei silty clay from tests and ref
for oedometer loading) at a reference stress level p , a power, m,
those predicted from various models
for the stress dependent stiffness formulation, the pure elastic
Poisson’s ratio or unloading/reloading Poisson’s ratio, ur, the
space is ellipse in shape and symmetric with respect to the hy- Mohr-Coulomb strength parameters (, c), the K0-value in pri-
drostatic line. The modified Cam-clay constitutive model in- mary one-dimensional compression (K0nc), and failure ratio, Rf.
volves five parameters, i.e., the frictional constant, M, the iso-
Eurref is a key parameter to influence the movements of excava-
tropic logarithmic compression index, , the swelling index, ,
tions. For undrained clay, the Eurref can be determined from the
pure elastic or unloading-reloading Poisson’s ratio, ur, and pure
elastic or unloading-reloading Young’s modulus, Eur. The yield swelling index , exactly the same as that in the MCC model (Lim
surface is mainly determined by the parameter  and M. The Eur, et al. 2010). For sand, the Eurref can be determined from the SPT
a key parameter to control computed movements of excavations, (standard penetration test) correlation (Khoiri et al. 2014).
can be derived from the swelling index  (Lim et al. 2010). Figure 6 shows the comparison of the measured wall dis-
Figure 4 shows the comparison of measured wall displace- placements and those computed by the HS model for the TNEC
ments of the TNEC case and those computed using the MCC mod- case. The computed wall displacements were generally close to
el with the parameters directly from the laboratory tests, that is, the field measurement. Compared with the MCC model, the HS
real soil parameter (Lim et al. 2010). The wall displacements are model gives better computed ground surface settlements. The HS
close to field measurements at early stages (stages 1 and 2) while model may be suitable for the analysis of deep excavation prob-
the computed wall deflections are smaller than the field measure- lems.
4 Journal of GeoEngineering, Vol. 11, No. 1, April 2016

Displacement (cm) Distance from the wall (m)

141210 8 6 4 2 0 0 10 20 30 40 50 60 70 80 90 100
0 0

Settlement (cm)
1 23
5 -2
10 2 4
15
3 5 -4
5 4 6
Depth (m) 20 6 -6
7 7
excavation stage
25 -8
30
35 Field measurements
measurement
40 MCC model
45
50
Fig. 4 The comparison of measured wall displacement and ground settlement profile with computed using the MCC model (real soil
parameters)

Displacement (cm) Distance from the wall (m)

141210 8 6 4 2 0 0 10 20 30 40 50 60 70 80 90 100
0 0

Settlement (cm)
2
5 3 -2
1
10 4
15 2 -4
3 5
Depth (m)

6 4 6 -6
20 7 5 excavation stage
7
25 -8
30
35 measurements
Field measurement
40 MCC model
45
50

Fig. 5 The comparison of measured wall displacement and ground settlement profile with computed using the MCC model (adjusted
parameter, /  0.25)

1
2
3
2 4
5
6 4 3 6 7 excavation stage
5
7
Field measurements
Field measurements
HS model
HS model

Fig. 6 The comparison of measured wall displacement and ground settlement profile with computed using the HS model

The hardening soil small strain model (abbreviated as HS does not have a good effect in improving the analysis accuracy.
small model, Benz et al. 2009) evolves from the hardening soil The wall displacements computed from the HS small model for
model with the consideration of small strain characteristics of all stages were generally close to those from field measurements.
soil. In the HS small model, two additional parameters are re- The HS small model gives slightly better prediction in surface
quired: The reference shear modulus at small strain ( G0ref ) and settlements though the prediction results are still far from the
shear strain (0.7) at which the secant shear modulus equal to 0.7 field measurements.
G0ref . For analysis, the G0ref can be obtained from the bender Lim and Ou (2016) has studied various stress paths of the
element test and the 0.7 can be set equal to 105. Figure 7 shows soil in an excavation with the analysis of the TNEC case using
the analysis results using the model with 0.7 105. Compared the HS model. As shown in Fig. 8, the typical stress paths all
with the results by the HS model (Fig. 6), the HS small model moved vertically, either exceeding the initial shear yield surface
 Chang-Yu Ou: Finite Element Analysis of Deep Excavation Problems 5

1 2
3
2 4
5
4 3
6 7 excavation stage
6 5
7
Field measurements
Field measurements
HS
HS model
Small model

Fig. 7 The comparison of measured wall displacement and ground settlement profile with computed using the HS small model

C
L 9) if the effective strength parameters are applied (the MC un-
N (GL -10m)
drained A approach). On the other hand, if the same effective
250 GL -19.7m
unloading/reloading deformation parameters are used but the
R O (GL -20m)

MC line Q
P (GL -30m)
undrained shear strength was specified, i.e., su and   0 as the
strength parameters (Fig. 9(c)), the analysis is referred to as the
200 MC undrained B approach. As featured in Fig. 10, the computed
wall deflections obtained from the MC Undrained A and Un-
Initial
InitialSYS
SYS
drained B diverse each other. At the final stage of excavation, the
Q 7
150 results from the MC Undrained B approach agree well with those
q (kPa)

P 7 from the HS model, even closer to the field measurements. The

0 MC Undrained A resulted in smaller wall deflections as com-
100 O 6 pared to those from the MC Undrained B because the MC Un-
7
7 0 drained A overestimates the undrained shear strength of clay.
5 R The MC model is certainly applicable to the analysis of ex-
7
50 0 cavations in drained material like sand/gravel. Research has
N 2
CYS (z 20m)
CYS(z=20m) proved that Mohr-Coulomb model is applicable to sand sandy
(z 10m)
CYS(z=10m)
CYS soil as long as appropriate selection of the stiffness parameters
CYS (z
CYS 30m)
(z=30m)
0 (Khoiri et al. 2014).
10 20 30
0
0 0 0
p ' (kPa) 3.2 Total Stress Model
Fig. 8 Effective stress paths in the retained and excavated sides The   0 MC model is a total stress model, which involves
(note: CYS denotes cap yield surface; SYS denote shear four parameters, namely, the undrained strength parameters (c, )
yield surface) and the undrained deformation parameters (Eu, u). With the total
stress analysis, the undrained Poisson’s ratio u should be equal
to 0.5 for the saturated clay under the undrained condition and
(SYS) (points N, O, P and Q) but still inside the cap yield surface the cohesion intercept should be equal to the undrained shear
(CYS), or inside the initial shear yield surface and cap yield sur- strength su and   0. The undrained elastic Young’s modulus, Eu,
face (point R). Results from finite element analysis also found should be determined based on empirical correlations or from
that the change in the volumetric strain along those typical stress back analysis of well-documented case histories. As long as an
paths was almost equal to 0 and the generation of plastic volu- appropriate selection of Young’s modulus, the wall deflection
p
metric strain ( v ) was very small, around 105 when the soil can be predicted with a good accuracy. Figure 11 shows the
yielded, i.e., exceeding the initial shear yield surface. It was comparison of wall deflection from the field measurement with
therefore proved that when the state of stress has not yet reached those from the MC model where the undrained Young’s modulus
to the failure line, the soil was mainly dominated by elastic be- was assumed to be equal to 500 su. Similar to the MCC model
havior. The unloading/reloading modulus should be recognized and HS model, the MC model is unable to predict the surface
settlement well.
as a critical parameter in controlling deformation behavior of an
The undrained soft clay model, abbreviated as the USC
excavation.
model, is a stress path dependent total stress model, which con-
Therefore, the MC model were able to simulate the defor-
siders the following characteristics of soil (Hsieh and Ou 2012):
mation behavior of the soil in excavations under the undrained
condition. Since the MC model is unable to predict the nonlinear 1. Variation of undrained shear strength with principal stress
behavior of soil before failure, it would underestimate the gener- rotation;
ation of the excess porewater pressure under the undrained con- 2. Variation of Young’s modulus with the increase of stress
dition and thus overestimates the undrained shear strength (Fig. level;
6 Journal of GeoEngineering, Vol. 11, No. 1, April 2016

Figure 12 compares the wall displacements from field

measurement with those predicted from the USC model. The
computed wall deflections for all stages agreed well with the
field observations. The development of the wall deflection shape
with the construction sequence and the location of the maximum
wall deflection computed from the model were also very close to
those observed in field observations. The computed surface set-
tlements were also in good agreement with the observed settle-
ments. The location of the maximum surface settlement comput-
ed from the model was almost the same as that from field obser-
vations. The USC model predicted both wall deflection and sur-
face settlement well.
(a) Real soil (b) MC undrained A approach

4. STABILITY ANALYSIS AND FAILURE

MECHANISM OF EXCAVATIONS

The stability of deep excavations is a main concern for prac-

tical engineers. The failure of excavations is often characterized
by the collapse of the support system and the large inward
movement of the surrounding soil. Conventional methods, in-
cluding Terzaghi’s method, Bjerrum and Eide’s method, and the
slip circle method, are commonly used for the evaluation of basal
heave stability of excavations in soft clay. In these methods, a
failure surface of soil is first assumed, and the factor of safety
(FS) against basal heave is then calculated as the ratio of the re-
(c) MC undrained B approach
sistant force to the driving force. Although this calculation pro-
cedure is convenient for practical application, the actual failure
Fig. 9 Effective and total stress paths of the soil subject to
surface may not be the same as that assumed in the conventional
compression
methods. Moreover, the methods are unable to identify if the
failure is due to the soil or structures.
3. High stiffness of soil at small strain; The stability of excavations has been studied by many re-
searchers using the finite element method (FEM) with reduced
4. Rational way to determine the undrained shear strength.
shear strength (e.g., Goh 1990; Faheem et al. 2003; Do et al.
The USC model also employed the concept of the yield sur- 2013). However, all of the previous studies used the elastic sup-
face to differentiate the Young’s modulus between the primary port system and did not model the existence of center posts in
loading and unloading/reloading states. The unloading/reloading excavations that are used to support the horizontal struts, such
Young’s modulus, Eur, also degrade, from the range of small that the FEM might not sufficiently simulate the behaviors of
strain, with the increase of strain or stress level due to the devel- excavations. It is found that the factor of safety are overestimat-
opment of pore water pressure. ed.

Field measurements
Field measurements
MC undrainedAA
MC undrained
MC undrained
MC undrainedBB

Fig. 10 The comparison of measured wall displacement and ground settlement profile with computed using the MC undrained A and B
approached, respectively
 Chang-Yu Ou: Finite Element Analysis of Deep Excavation Problems 7

Displacement (cm) Distance from the wall (m)

141210 8 6 4 2 0 0 10 20 30 40 50 60 70 80 90 100
0 0

Settlement (cm)
2
5 1 -2
2 3
10 4
15
3 -4
4
5
Depth (m)
5 -6
20 7 6 6
25 7 excavation stage -8
30
35 measurements
Field measurement
40 MC model
model
45
50
Fig. 11 The comparison of measured wall displacement and ground settlement profile with computed using the  0 MC
model (E/su  500)

Displacement (cm) Distance from the wall (m)

141210 8 6 4 2 0 0 10 20 30 40 50 60 70 80 90 100
0 0

Settlement (cm)
12
5 1 3
-2
10 2 4
3 5 -4
15 5 4 6
Depth (m)

20 7 6 -6
25 7 excavation stage -8
30
35
40 Field measurement
measurements
USC model
model
45
50
Fig. 12 The comparison of measured wall displacement and ground settlement profile with computed using the USC model

The author and his group furthermore performed the stabil- Wall
Walldeflection
deflection,(mm)
mm 98% M
98% Mstage
stage maxmax
ity analysis of four failure cases in which the retaining wall and -200 0 200 400
99% M
99% Mstage
stage maxmax
lateral supports were simulated with the elastic-plastic behavior 0 Mstage
M 1.00
stage maxmax = 1.00
and the center posts were considered in the finite element model -4 500
(Do et al. 2016). The four cases are the Taipei Rebar Broadway -8
Heave, mm
m
Depth,(m)

400
excavation, the Taipei Shi-Pai case, the Hangzhou subway exca-
-12 300
Depth

vation and the Nicoll highway excavation. It is found that the

-16 200
factors of safety computed from the finite element was close to or
smaller than 1.0 for all failure cases. The yielding of the struts or -20 100
walls first caused a rapid increase in the wall deflection and soil -24 0
heave and then the failure of the excavations. Figure 13 gives the 0 4 8 12 16 20 24 28
wall deflection and soil heave at the excavation bottom at the Distance to wall
Distance to wall,(m)
m
final stage for different values of  Mstage for the Taipei Rebar Fig. 13 Wall deflection and soil heave of Taipei Rebar Broad-
Broadway case where  Mstage is the ratio of the unloading way case as using the elastoplastic support system
applied successfully in calculation to that caused by excavation at
the final stage. The maximum strength reduction ratio was found
incidentally to be 1.00 when excavation reached to the final stage, the plastic bending moment (Mp) and axial force (Np). During ex-
which also corresponds to the  Mstage max value of 1.00. cavation, the internal force curves of the struts at the first and sec-
When excavation approached to the failure, the wall exhibited a ond layers tended to develop toward the horizontal (M) axis.
kick-out deflection mode, and the maximum deflection (up to Their behaviors were significantly affected by the bending moment.
400 mm) occurred at the wall toe. The soil heave also exhibited a On the other hand, the curves of the struts at the third layer grew
toward the vertical (N) axis, indicating a higher impact of the axial
consistent deformation mode, where the maximum heave (up to
force on this layer than the bending moment. The struts at the sec-
500 mm) developed at the center of the excavation.
ond and third layers started to yield at the fourth excavation stage.
Figure 14 shows an interaction diagram of the internal forc-
The struts at all the four layers yielded in the final excavation stage.
es (M, N) of the struts and walls in the finite element analysis In addition, the internal forces of the struts at the fourth (lowest)
where the support system was assumed to be elastic-plastic. The layer appeared at the topmost crest of the area enveloped by the
behaviors of the struts and walls are elastic when their internal boundary lines in this stage. This showed a predominant impact of
force curves develop within the corresponding areas enveloped the axial force on this layer, and therefore, the development of
by the boundary lines that were constructed from the values of plastic hinges along the strut lengths could be inferred.
8 Journal of GeoEngineering, Vol. 11, No. 1, April 2016

N, kN/m ○ ○ ○ 1st strut layer

-14000 △ △ △ 2nd strut
1st strutlayer
layer tion, it is necessary to adopt effective measures to limit thewall
1 2nd
rd strut layer
2 ◇ ◇ ◇ 3 strut layer
3 -10000 3rd strut layer deflection or ground settlement. Ground improvement is a com-
-1000 5 5 4th strut
□ □ □ 4th layer
strut layer
4 mon measure to reduce the excavation-induced ground move-
5 wall
wall
4 ments (e.g., Gaba 1990; Liu et al. 2005; Parashar et al. 2007).
4 Recently, cross walls and buttress walls have been used as the
-500 alternative measures widely. The author and his group have
Excavation 3 5
stage shown that use of cross walls in deep excavations can reduce the
5 2 3 1 23 4 wall deflection to a very small amount (e.g., Ou et al. 2006; Ou
0 4 M kNm/m et al. 2011). Since the mechanisum of cross walls in reducin the
-1000-600 -200 -100 0 100 200 600 1000
5 movements has been understood, a relevent simplified analysis
BL of 1st strut was developed (Hsieh et al. 2012). However, use of cross walls
500 layer in a very wide excavation would be costly. Therefore, buttress
BL of 2nd strut layer walls have been adopted recently as an alternative of cross walls
BL of 3rd and 4th strut for the protection of adjacent buildings during deep excavation.
1000
10000 layers However, the mechanism of buttress walls in limiting the
BL of wall wall defleciton is very complicated. Some engineers employed
14000
Note:
Note: the concept of ground improvement to design buttress walls, that
BLrepresents
BL representsboundary
boundary lines.
lines. is, the average strength and stiffness of the soil in the passive
At the55ththstage,
Atthe stage, internal
internal forces
forces to M
corresponding
corresponding Mstage
to stage max zone increased due to the buttress walls. Some treat buttress
max
are are plotted.
plotted. walls to function as T-beams as those in reinforced concrete
structures, enhancing the capability of moment-resistance of the
Fig. 14 Interaction diagram of internal forces of elastic-plastic
diaphragm wall and some used frictional resistance between two
structural elements of Taipei Rebar Broadway case
sides of the buttress wall and adjacent soil to provide additional
resistance to against the movement of the diaphragm wall. To
In addition, the FEM analysis of other three cases were also have a good design of buttress walls, it is necessary to perform
performed and the factors of safety were also less than 1.0. It is finite element analysis or to understand the mechanism of but-
also found that the yield or failure of the excavations was initiat- tress walls in reducing the wall deflection.
ed from either the struts or walls rather from the soil. The results Figure 15 shows the Park-2001 project in Taipei. The dia-
from the FE analysis were consistent with the field observation. phragm wall was 21 m in depth and 0.6 m in thickness. Figure
15(b) and 15(c) show the dimension and depth of the buttress
5. THREE DIMENSIONAL ANALYSIS walls. Final excavation depth was 8.6 m. Figure 15(c) shows the
construction sequence and subsoil profile and the basic soil
To avoid the damage of adjacent buildings during excava- properties.

(a) Excavation plan (b) Types of buttress wall (c) Profile of the excavation

Fig. 15 The Park-2001 excavation project

 Chang-Yu Ou: Finite Element Analysis of Deep Excavation Problems 9

Figure 16 plots a comparison of the computed and monitored

wall deflections at SI4, near the location of a buttress wall using Deflection (mm) Deflection (mm)
the HS model. The computed wall deflections were in general 0 20 40 60 80 100 0 20 40 60 80 100120
0
close to the monitored values but slightly overestimated at the last
two stages. For the evaluation of effectiveness of buttress walls in 4
reducing the wall deflection, analysis of the excavation with as-
sumption of no buttress walls was performed and the results are 8

Depth (m)
also shown in Fig. 16. The computed maximum wall deflections 12
with buttress wall and without buttress walls were 76.1 mm and
41.5 mm, respectively. Installation of buttress wall in this case 16
certainly had good effects in reducing the wall deflection.
20
Figure 17 shows the computed wall deflections for the Stage 1 Stage 2
excavation where the retangular shape buttress walls, with the 0
same area as the T-shape in the Park-2001 project, were
employed. Results show that with consideration of frictional 4
resistance between the buttress walls and the adjacent soil, the
computed wall deflections were less than those for without 8

Depth (m)
buttress walls. The computed wall deflections for frictionless 12
case were almost the same as those for without buttress walls.
This implies that the main mechanism of the rectangular shape 16
buttress walls in reducing the wall deflection was due to the fric-
tional resistance between buttress walls and the adjacent soil. The 20
Stage 3 Stage 4
combined bending stiffness of the diaphragm wall from the dia-
24
phragm wall and the rectangular shape buttress wall plays insig- Computed result (with BW)
nificant role in the reduction of the wall deflection. This is be- Monitored result Computed result (without BW)
cause when the buttress walls were demolished along with the
removal of soil, the buttress walls below the excavation bottom Fig. 16 Monitored and computed wall deflections at SI4 for the
were mainly “pushed” by the diaphragm wall rather than provid- Park-2001 excavation project
ing the bending resistance against the deformation of the dia-
phragm wall.

Deflection (mm) Deflection (mm) Deflection (mm) Deflection (mm)

0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100 0 20 40 60 80 100 120

5
Depth (m)

10

15

20
Stage 1 Stage 2 Stage 3 Stage 4
With outBW
Without BW
With BWwithout
With BW without frictional
frictional resistance
resistance
With BW
With BWwith frictional
with resistance
frictional resistance

Fig. 17 Wall deflections for the assumed excavations without buttress walls and with buttress walls
10 Journal of GeoEngineering, Vol. 11, No. 1, April 2016

To study the mechanism of buttress wall in reducing the Deflection (mm)

Deflection (mm)
wall deflection, the Park-2001 project was analyzed with the -20 0 20 40 60 80 100
assumption of the retaingular shape of buttress walls with the 0
lenght of 5 m, 10 m and 20 m, respectively. Figure 18 show the
variation of the computed wall deflections for the scenarios with 4
buttress walls and without buttress walls, respectively. The
amount of wall deflection certainly decreased with the increasing 8

(m)
length of buttress walls. To investigate the mobilization of shear

Depth (m)
strength, the relative shear stress ratio, rel, is defined as the ratio 12

Depth
of shear stress to the shear strength of a soil for a given effective
normal stress. The rel 1.0 implies that the soil adjacent to the 16
buttress wall was at the failure state and its shear strength or fric-
tional resistance was fully mobilized. Figure 19 illustrates the 20
variation of the relative shear stress ratio on the surface of the
buttress wall in the shallow excavation in clay and in sand, with 24
the buttress wall length. The rel was smallest near the diaphragm
wall and generally increased with the increasing distance from WithoutBW
Without BW  10m
Lrr=10m
the diaphragm wall, up to a value of 1.0. The rel was mostly  5m
Lrr=5m L  20m
Lr=20m
r
close to 1.0 for the short length of the buttress wall, e.g., Lr  5 m
and majorly much less than 1.0 for the longer length of the but- Fig. 18 Computed wall deflections for different lengths of the
tress wall, e.g., Lr  20 m (Fig. 19). rectangular buttress wall in excavations

Diaphragm wall

Lr  5m Lr  10m Lr  20m

Fig. 19 Distribution of relative shear stress ratio for the cases with different lengths of the rectangular buttress wall

Figure 19 shows a relatively small shear stress ratio within 6. CONCLUSIONS

the first 2.0 m from the diaphragm wall for all lengths of the but-
tress wall. This phenomenon can be explained by the fact that the The following conclusions can be drawn from the above
movement at any location of the buttress wall was all the same, studies:
almost the same as the movement of the diaphragm wall. The soil 1. The effective stress undrained analyses can generally result
in front of the diaphragm wall, say, 2.0 m from the diaphragm in good prediction of wall deflections for excavations in clay
wall, directly pushed by the diaphragm wall, should have almost if the yield surface of the model and flow rule meet the de-
the same amount of movement as the diaphragm wall or buttress formation characteristics of soil. The input parameters can
wall. Therefore, the relative displacement or relative shear stress be derived from the basic properties of soil. However, the
ratio between the buttress wall and the adjacent soil within the ground settlement cannot be predicted well unless the char-
first 2.0 m was very small but it increased gradually with the acteristics of high stiffness at small strain is incorporated in
increasing distance from the diaphragm wall. It was clear that if the constitutive model of soil. The total stress undrained
the buttress wall length was less than 2.0 m, the buttress wall was model can also yield a good prediction of the movements in
unable to restrain the wall deflection although the combined an excavation where the strength parameters of soil should
bending stiffness from the contribution of the diaphragm wall be determined from laboratory or field tests but the Young’s
and buttress wall seems increased. modulus can only be selected from empirical correlations.
 Chang-Yu Ou: Finite Element Analysis of Deep Excavation Problems 11

2. For excavations under the undrained condition and normal Do, T. N., Ou, C. Y., and Lim, A. (2013). “Evaluation of factors of
construction, the soil in excavations is mostly dominated by safety against basal heave for deep excavations in soft clay using
elastic behavior. As long as the undrained shear strength can the finite element method.” Journal of Geotechnical and Ge-
oenvironmental Engineering, 139, 21252135.
be specified, the effective stress MC model with   0 can
result in good prediction of wall deflection. Faheem, H., Cai, F., Ugai, K., and Hagiwara, T. (2003).
“Two-dimensional base stability of excavations in soft soils us-
3. Generally speaking, conventional formulae for stability ing FEM.” Computers and Geotechnics, 30(2), 141163.
analysis were derived from general bearing capability equa-
Gaba, A. R. (1990). “Jet grouting at Newton station.” Proceedings of
tions or simple force equilibrium. Though those have been
the 10th Southeast Asia Geotechnical Conference, Taipei, 7779.
used for decades, the mechanism of failure of excavations
are unable to be found through those methods. Finite ele- Goh, A. T. C. (1990). “Assessment of basal stability for braced ex-
ment method can be employed to estimate the factor of cavation systems using the finite element method.” Computers
and Geotechnics, 10(4), 325338.
safety of excavations or investigate the mechanism of failure
excavations when the wall and lateral supports can be simu- Hsieh, P. G., Ou, C. Y., and Liu, H. T. (2008). “Basal heave analysis
lated with the elastic-plastic behavior. of excavations with consideration of anisotropic undrained
strength of clay.” Canadian Geotechnical Journal, 45, 788799.
4. Ground improvement, cross walls and buttress walls are
often designed to ensure the safety of adjacent buildings Hsieh, P. G. and Ou, C. Y. (1998). “Shape of ground surface settle-
ment profiles caused by excavation.” Canadian Geotechnical
during deep excavation. Plane strain finite element analysis
Journal, 35, 10041017.
may be used to evaluate the performance of those measures
but just to a certain extent. For some situations such as but- Hsieh, P. G. and Ou, C. Y. (2012). “Analysis of deep excavations in
clay under the undrained and plane strain condition with small
tress walls where its interaction between structures and soil
strain characteristics.” Journal of the Chinese Institute of Engi-
is complicated, three dimensional analysis is able to realisti- neer, 35(5), 601616.
cally evaluate the effectiveness of buttress walls in excava-
tions. The mechanism of buttress walls in the reduction of Hsieh, P. G. and Ou, C. Y. (2016). “Simplified approach to estimate
the maximum wall deflection for deep excavations with cross
wall deflection can also be investigated using the three di- walls in clay under the undrained condition.” Acta Geotechnica,
mensional finite element analysis. 11, 177189.
Hsieh, P. G., Ou, C. Y., and Shih, C. (2012). “A simplified plane
ACKNOWLEDGEMENTS strain analysis of the lateral wall deflection for excavations with
cross walls.” Canadian Geotechnical Journal, 49, 11341146.
The author acknowledges the research grants from the Min- JSA (1988). Guidelines of Design and Construction of Deep Excava-
istry of Science and Technology in Taiwan (previously known as tion. Japanese Society of Architecture, Japan.
National Science Council) in the past. The results in this study Khoiri, M., Ou, C. Y., and Teng, F. C. (2014). “A comprehensive
mostly come from his supervised Master and PhD students. Their evaluation of strength and modulus parameters of a gravelly
contributions in this paper are highly acknowledged. cobble deposit for deep excavation analysis.” Engineering Ge-
ology, 174, 6172.
Lim, A., Ou, C. Y., and Hsieh, P. G. (2010). “Evaluation of clay
REFERENCES constitutive models for analysis of deep excavation under un-
drained conditions.” Journal of GeoEngineering, 5(1), 920.
Benz, T., Vermeer, P. A., and Schwab, R. (2009). “A small-strain Lim, A. and Ou, C. Y. (2016). Stress Paths in Deep Excavations
overlay model.” International Journal for Numerical and Ana- Under Undrained Conditions and Its Influence on Deformation
lytical Methods in Geomechanics, 33, 2544. Analysis, Geotechnical Research Report GT1601, National
Bjerrum, L. and Eide, O. (1956). “Stability of strutted excavation in Taiwan University of Science and Technology.
clay.” Geotechnique, 6(1), 3247. Liu, G. B., Ng, C. W. W., and Wang, Z. W. (2005). “Observed per-
Bowles, J. E. (1986). Foundation Analysis and Design, McGraw-Hill, formance of a deep multistrutted excavation in Shanghai soft
New York. clays.” Journal of Geotechnical and Geoenvironmental Engi-
neering, 131(8), 10041013.
Chin, C. T., Crooks, J. H. A., and Moh, Z. C. (1994). “Geotechnical
properties of the cohesive Sungshan Deposits, Taipei,” Ge- Ng, C. W. W. (1999). “Stress paths in relation to deep excavations.”
otechnical Engineering, Southeast Asian Geotechnical Society, Journal of Geotechnical Engineering and Geoenvironmental
25(2), 77103. Engineering, 124(5), 357363.
Clough, G. W. and O’Rourke, T. D. (1990). “Construction-induced Osman, A. S. and Bolton, M. D. (2006). “Ground movement predic-
movements of in situ walls.” Proceedings of the Design and tions for braced excavations in undrained clay.” Journal of Ge-
Performance of Earth Retaining Structures, ASCE Special otechnical and Geoenvironmental Engineering, 132(4), 465477.
Conference, Ithaca, New York, 439470. Ou, C. Y. and Hsieh, P. G. (2011). “A simplified method for pre-
Diaz-Rodriguez, J. A., Leroueil, S., and Aleman, J. D. (1992). dicting ground settlement profiles induced by excavation in soft
“Yielding of Mexico city clay and other natural clays.” Journal clay.” Computers and Geotechnics, 38, 987997.
of Geotechnical Engineering, 118(7), 981995. Ou, C. Y., Lin, Y. L., and Hsieh, P. G. (2006). “A case record of
Do, T. N., Ou, C. Y., and Chen, R. P. (2016). “A study of failure an excavation with cross walls and buttress walls.” Journal of
mechanisms of deep excavations in soft clay using the finite GeoEngineering, 1(2), 57986.
element method.” Computers and Geotechnics, 73, 153163. Ou, C. Y., Liu, C. C., and Chin, C. K. (2011). “Anisotropic visco-
12 Journal of GeoEngineering, Vol. 11, No. 1, April 2016

plastic modeling of rate dependent behavior of clay.” Interna- Roscoe, K. H. and Burland, J. B. (1968). “On the generalized stress-
tional Journal for Numerical and Analytical Methods in Geo- strain behavior of ‘wet’ clay.” Engineering Plasticity, Heyman
mechanics, 35(11), 11891206. and Leckie, Eds., Cambridge University Press, Cambridge,
535609.
Ou, C. Y., Hsieh, P. G., and Lin, Y. L. (2011). “Performance of ex-
cavations with cross walls.” Journal of Geotechnical and Ge- Schanz, T., Vermeer, P. A., and Bonnier, P. G. (1999). “The harden-
oenvironmental Engineering, 137(1), 94104. ing soil model: Formulation and verification.” Proceeding of
Beyond 2000 in Computational Geotechnics10 Years of
Parashar, S., Mitchell, R., Hee, M. W., Sanmugnathan, D., Sloan, E.,
PLAXIS, Balkema, Rotterdam.
and Nicholson, G. (2007). “Performance monitoring of deep
excavation at Changi WRP project.” Proceedings of the 7th In- Terzaghi, K. (1943). Theoretical Soil Mechanics, John Wiley & Sons,
ternational Symposium on Field Measurements in Geomechanics, Inc., New York.
Singapore, 112. Tschebotarioff, G. P. (1951). Soil Mechanics, Foundations, and Earth
Peck, R. B. (1969). “Deep excavation and tunneling in soft ground.” Structures, McGraw-Hill, New York.
Proceedings of the 7th International Conference on Soil Me- Wheeler, S. J., Näätänen, A., Karstunen, M., and Lojander, M. (2003).
chanics and Foundation Engineering, State-of-the-Art Volume, “An anisotropic elasto-plastic model for natural clays.” Cana-
Mexico City, 225290. dian Geotechnical Journal, 40(2), 403418.
Pestana, J. M. and Whittle, A. J. (1999). “Formulation of a unified
constitutive model for clays and sands.” International Journal
for Numerical and Analytical Methods in Geomechanics, 23,
12151243.