You are on page 1of 11

Application of Large Three-Dimensional Finite-Element

Analyses to Practical Problems


Fook-Hou Lee1; Sze-Han Hong2; Qian Gu3; and Pengjun Zhao4

Abstract: This paper deals with the usefulness and feasibility of three-dimensional (3D) finite-element analysis in geotechnical engineering
practice. The usefulness of 3D analysis is illustrated through two examples. The first example deals with a collapse problem and highlights the
Downloaded from ascelibrary.org by New York University on 05/17/15. Copyright ASCE. For personal use only; all rights reserved.

presence of significant three-dimensional effects even in what may appear to be a two-dimensional problem. The second example deals with
the effect of soil flow on piles preinstalled within an excavation area. This problem is increasingly encountered in urban high-rise construction
wherein piles are often preinstalled before basement excavation. The lessons learned from these two examples are then discussed, and the
viability of conducting three-dimensional analysis with relatively modest computing means is explored. DOI: 10.1061/(ASCE)GM.1943-
5622.0000049. © 2011 American Society of Civil Engineers.
CE Database subject headings: Finite element method; Three-dimensional analysis; Excavation.
Author keywords: Finite element; Three-dimensional; Excavation; Iterative solver.

Introduction excavations. This paper illustrates some other situations involving


excavation but not corners and soldier piles, wherein 3D effects
Finite-element (FE) analysis has been assuming an increasingly are significant, by using two case histories. These case histories
important role in the geotechnical consultant’s office. In Singapore, demonstrate the variety and sublety of 3D effects and the usefulness
at present, construction projects involving temporary earth retain- of 3D analysis. The final part of the paper reviews the possibility of
ing structures or tunneling usually have to be supported by appro- conducting realistic 3D analysis with relatively modest computa-
priate numerical analyses. However, to date, most of the analyses tional means.
conducted are still restricted largely to two-dimensional (2D) plane
strain or axisymmetric analyses. Many practical geotechnical prob-
lems are inherently three-dimensional (3D) and modeling them in Long Trench Excavation: Nicoll Highway
2D often entails a significant degree of simplification. Some of
these are quite obvious; for instance, simplification of building Background
basement excavations into 2D plane strain problems is still widely
practiced. A number of studies involving 3D analysis have been On the April 20, 2004, at 3:30 p.m., a segment of the deep trench
reported on basement excavations in recent years (e.g., Finno and excavation for the Circle Line of the Singapore Mass Rapid Transit
Harahap, 1991; Wong and Patron, 1993; Ou and Chiou, 1993; Lee collapsed, resulting in the collapse of the adjacent Nicoll Highway
et al. 1995; Lee et al. 1998; Blackburn and Finno, 2007). Apart (Fig. 1). As part of the postcollapse investigation, the Land
Transport Authority requested researchers from the National Uni-
from the previously mentioned, 3D effects have also been reported
versity of Singapore to conduct a 3D FE study to investigate the
in other situations. For example, soldier-piled excavations are often
significance of three-dimensional effects on the collapse and the
analyzed by using 2D FE analyses with properties that are averaged
events leading up to it.
over a certain span of the wall. Hong et al. (2003) noted that such an
approach can also give rise to modeling errors. Soil Profile and Properties
However, three-dimensionality can occur in other situations
in addition to corners and soldier piles in excavations. Many of The collapse occurred on reclaimed land. The ground surface was
these other situations are less well-understood than basement covered by a layer of sand fill underlain by soils of the Kallang
Formation, which was characterized by the presence of soft marine
1
Professor, Dept. of Civil Engineering, National Univ. of Singapore, clay. Figs. 2 and 3, which are drawn with respect to the same
Kent Ridge, Singapore (corresponding author). E-mail: cveleefh@nus Origin O, show the boreholes, trench alignment, and the depths
.edu.sg of the various soil layers below ground surface. Two marine clay
2
Director, GeoSoft Pte. Ltd., 2 Kaki Bukit Place #07-00 Tritech layers named upper marine clay (UMC) and lower marine clay
Building, Singapore 416180. (LMC), which are heavily interbedded with fluvial soil types F1
3
Principal Geotechnical Engineer, AECOM Australia Pty Ltd., 3 Forrest and F2 and estuarine deposits, can be seen. The total depth of
Place, Perth, WA 6000, Australia. the base of the lower marine clay layer varies across the site from
4
Senior Geotechnical Engineer, Butler Partners Pty Ltd., 79 Doggett approximately 22 m at the western end to approximately 36 m in
Street, Newstead, QLD4006, Australia.
the eastern end. Fluvial soil type F1 is a predominantly granular soil
Note. This manuscript was submitted on November 3, 2009; approved
on August 23, 2010; published online on August 27, 2010. Discussion per- with small amounts of silt and clay, often described as a silty sand
iod open until May 1, 2012; separate discussions must be submitted for or clayey sand. Fluvial soil type F2 is a predominantly cohesive
individual papers. This paper is part of the International Journal of Geo- soil; typically silt with small amounts of clay or sand. The estuarine
mechanics, Vol. 11, No. 6, December 1, 2011. ©ASCE, ISSN 1532-3641/ deposit is an organic soft soil layer, with properties that are similar
2011/6-529–539/$25.00. to the upper marine clay. For this reason, the estuarine deposit is

INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / NOVEMBER/DECEMBER 2011 / 529

Int. J. Geomech. 2011.11:529-539.


a. Ground domain and jet-grouted zones were modeled by using
eight-noded brick elements.
b. Struts were modeled by using nonlinear, elastoplastic beam
elements. Strut failure under primarily compressive forces
was considered important and was modeled in several ways.
First, by prescribing a peak strength followed by strain soften-
ing, buckling of struts under pure axial loading conditions can
be approximated. Secondly, in strut members that are subjected
to large axial forces, the additional bending moment generated
by the axial forces (P) interacting with the lateral deflection
(Δ) of the strut may be significant. In the analysis, this P-Δ
interaction is captured by performing the analysis in large
deformation mode that allows the geometry of the strut to be
updated, thereby reflecting the effect of lateral deflection.
Downloaded from ascelibrary.org by New York University on 05/17/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 1. Overall view of the Nicoll Highway collapse zone (image by Finally, by prescribing an elastoplastic moment-rotation rela-
Qian Gu) tion, plastic hinge formation can be reflected.
c. Diaphragm wall panels were modeled by using four-noded plate
elements. The section modulus of the wall was measured to be
considered as part of the upper marine clay in this analysis. The the sum of the cracked section modulus (EI) of the concrete and
lower marine clay is underlain by relatively stiff soil of the Old that of the steel reinforcement. In lateral compression, the entire
Alluvium Formation. Table 1 shows some of the inferred engineer- gross section was considered to be effective. Finally, yielding of
ing properties measured at this site. The SPT blow count of the Old the wall in flexural mode was also modeled to capture the pos-
Alluvium soils ranges typically from 20 blows=300 mm at the top sible formation of a plastic hinge in the wall below formation
of the stratum to more than 100 blows=300 mm within approxi- level just before collapse. The wall was modeled by using shell
mately 5 m of depth. The groundwater table is located very close elements with an equivalent thickness of 0.3 m and an equivalent
to the ground surface. In many boreholes, the depth of the ground- Young’s modulus, which starts off at a high value and reduces
water table was measured at less than 1 m from the ground surface. with strain. This models a wall with a cracked section modulus
This is reasonable considering the proximity of this site to the of approximately 40% of the gross section modulus of the
water edge. actual wall and an ultimate bending moment of approximately
2500 kN-m=m; the latter is the design ultimate moment.
Overview of Analysis
d. Diaphragm wall joints: In the actual construction, no global
The finite-element analysis was conducted by using ABAQUS/ waler was used to spread strut forces horizontally across
Standard version 6.3. The meshed model used for the analyses diaphragm wall panels. Instead, individual wall panels were
is shown in Figs. 4(a) and 4(b). The element types used in the directly connected to one or two struts by short lengths of
analysis are as follows: I-sections or by local walers, which spanned only between

Fig. 2. Bore holes, diaphragm wall, and depth of the bottom of the Lower Marine Clay

530 / INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / NOVEMBER/DECEMBER 2011

Int. J. Geomech. 2011.11:529-539.


compression, the two nodes on each connector are constrained
to undergo virtually the same displacement by assigning a very
high axial stiffness of 2 × 105 MN=m to the connector. When a
limiting tensile extension was exceeded, the shear and axial
stiffnesses of the connectors dropped to a very low value, which
simulated the break-up of the joints in tension. The tensile
extension limit was set at 0.1 mm. Given the uncertainties
on the effective length of the wall join, the prescribed limit
is highly approximate. Nonetheless, preliminary studies showed
that reasonable variation of extension limit about this value did
not significantly alter the results of the analysis.
e. Jet-grouted soil: Two jet-grouted layers were created within the
lower marine clay layer before excavation. The properties
adopted for the jet-grouted soil is shown in Table 1. As Fig. 5
Downloaded from ascelibrary.org by New York University on 05/17/15. Copyright ASCE. For personal use only; all rights reserved.

shows, in each layer of jet-grouted soil is an approximately


diagonal gap running between the north and south walls. This
gap corresponds to the alignment of a 66-kV high-voltage cable
that was not diverted during construction. Owing to obstruction
from the cable, the soil underlying it might not have been ade-
quately grouted. As such, the soil within this gap was modeled
as a “weak” jet grout with reduced strength of 100 kPa, as shown
in Table 1.

Material Properties
The analyses were conducted by using total, rather than effective,
stress formulation for several reasons. First, much of the excavation
was undertaken in soft, highly impermeable marine clay, which was
likely to remain undrained during the 4-month excavation period.
Secondly, because the analysis involved a collapsed excavation,
it was considered important to ensure that the undrained shear
strength of the soft clay was accurately reflected. Finally, the pro-
cess of collapse was better simulated dynamically, wherein inertial
effects could be invoked to equilibrate unbalanced forces that were
generated during the collapse. In ABAQUS version 6.3, dynamic
analysis could not be conducted in effective stress.
All soil types were modeled by using Mohr-Coulomb models.
For low-permeability soils, angle of friction was set to 0° and only
the undrained shear strength was used. More sophisticated soil
models were not used because this is a total stress analysis. In ad-
dition, the limited site investigation information available precludes
meaningful evaluation of parameters for more sophisticated mod-
els. Table 1 shows the soil properties used in the analysis. The soil
profile shown in Fig. 3 was fed into the analysis by means of a
program that was written to assign each element its respective soil
type on the basis of the northing, easting, and depth of the element
centroid. To reflect the variation of strength within the same layer of
soil, the upper marine clay was further subdivided into two layers
of equal thickness, with undrained shear strengths as shown in
Table 1. The lower marine clay was also similarly subdivided. As
shown in Table 1, the adopted parameters were on the conservative
Fig. 3. (a) 3D perspective of soil layers starting from ground surface: side of the peak strength from the field vane tests. On the other
fill, fluvial layer (F1), estuarine deposit (E), Upper Marine Clay hand, the relatively low remoulded undrained shear strength indi-
(UMC), and second estuarine layer (E); meshed surfaces represent cated by the field vane measurements suggests a significant amount
inferred interfaces between soil layers; (b) Soil profiles beneath second of softening in the transition from peak strength state to the re-
estuarine (E) layer [continued from Fig. 3(a)]: fluvial type 1 (F1), moulded state. As such, the adopted values of undrained shear
fluvial type 2 (F2), Lower Marine Clay (LMC), fluvial type 2 (F2), strength were considered to be representative values that took into
and Old Alluvium (OA) account both the peak and critical states of the marine clay.
The moduli of the marine clay were estimated by using an Eu =su
ratio of 67 recommended by Dames and Moore (1983), in which Eu
one to three wall panels. Because the diaphragm wall panels is the undrained Young’s modulus and su the undrained shear
were connected to each other only by shear keys, negligible ten- strength. The total stress analyses cannot directly account for
sile and flexural capacity along the horizontal direction existed. consolidation and swelling of the marine clay, which could have
To reflect this, the wall joints were modeled by two-noded occurred during construction; this was accounted for partially
connector elements connecting adjacent wall panels. Under by using a Poisson’s ratio that was less than 0.5. Wroth (1975)

INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / NOVEMBER/DECEMBER 2011 / 531

Int. J. Geomech. 2011.11:529-539.


Downloaded from ascelibrary.org by New York University on 05/17/15. Copyright ASCE. For personal use only; all rights reserved.

Table 1. Soil Properties Used in Nicoll Highway Analysis


Measured properties Properties adopted in analysis
CPT cone CPT shaft SPT blow Young’s Friction
Density ρ resistance qu friction f s Field vane shear count N modulus E Poisson’s Cohesion s or angle ϕ
Soil type (kg=m3 ) (kN=m2 ) (kN=m2 ) strength suv (kN=m2 ) (blows=300 mm) (MN=m2 ) ratio ν su (kN=m2 ) (degrees)
Fill 1,600 5 10 0.3 0 32
Upper marine clay, first half layer (UMC1) 1,600 300–500 5–8 Peak: 10–20; remoulded: 5–10 2 0.48 10 0
Upper marine clay, second half layer (UMC2) 1,600 500–700 8–10 Peak: 20–30; remoulded: 5–10 4 0.48 20 0
Fluvial soil (F2) 1,900 1,600 Approximately 70 10–20 14.5 0.48 70 0
Lower marine clay, first half layer (LMC1) 1,700 800–1,000 13–15 Peak: 22—; remoulded: 8–15 4 0.48 20 0
Lower marine clay, second half layer (LMC2) 1,700 1,000–1,200 15–17 Peak: 38–44; remoulded: 8–15 7 0.48 35 0
Old Alluvium (N < 70) 1,900 Approximately 2,000 Approximately 100 20–70
Old Alluvium (N > 70) 2,000 > 70 100 0.48 350 0
Jet-grouted soil 2,000 90 0.48 450 kN=m2 0
(peak strength) reached
at plastic strain of 1%;
100 kN=m2 (critical
state strength) reached
at 10% plastic strain
Weak jet-grouted soil (beneath 66-kV crossover) 2,000 20 0.48 100 0

Int. J. Geomech. 2011.11:529-539.


excavation—level 10

between 0.4 and 0.5.

532 / INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / NOVEMBER/DECEMBER 2011


66 kV cable alignment is omitted (represented by gaps)

Wroth’s (1975) chart with a plasticity index of 50 for Singapore

permeability, the representative Poisson’s ratio is likely to lie


marine clay leads to an effective Poisson’s ratio of approximately
between 0.254 and 0.371, increasing with plasticity index. Using
Fig. 4. Finite-element mesh: (a) before excavation; (b) after

0.36. In view of this and the fact that the marine clay has a very low-
reported effective Poisson’s ratios of lightly overconsolidated soils
Fig. 5. Upper and lower jet-grout layers; weak jet-grout beneath
The design shear strength of the jet-grouted soil is 450 kPa (this suggests an undrained shear strength of 50–100 kPa. The cone re-
is equivalent to an unconfined compressive strength of 900 kPa) sistance of approximately 1.6 MPa suggests an undrained shear
and the Young’s modulus was measured to be 300 times the shear strength of 70–140 kPa. The value used in the analysis is 70 kPa.
strength. However, preliminary studies showed that the use of this However, parametric analyses showed that both the F1 and F2
design values with an elastic perfectly plastic model results in pre- layers did not feature strongly in the overall behavior. This is also
dicted wall deflection, which is much smaller than those observed consistent with the Committee of Inquiry report (COI 2005), which
before failure. One possible explanation is that the jet-grouted did not ascribe any role to the F1 and F2 in the collapse.
marine clay might have undergone significant postpeak softening. The capacities of the struts adopted were not on the basis of the
Chin’s (2006) study shows that, under undrained triaxial compres- design capacities. Tests conducted by researchers at the Nanyang
sion, cement-treated soil specimens often strain soften after yield- Technological University shows failure of the joints between the
ing, with buildup of positive excess pore pressure. In this analysis, strut and wall at approximately 30% of the design capacities of
postpeak softening of the jet-grouted soil was modeled by speci- the struts, and these measured capacities were used. Most of the
fying a peak shear strength of 450 kPa, which decreased to an struts were preloaded to only 30% of the design preload because
ultimate strength of 100 kPa value at a plastic strain of 10%. The strut load measurements suggest significant loss of preload. The
Downloaded from ascelibrary.org by New York University on 05/17/15. Copyright ASCE. For personal use only; all rights reserved.

Young’s modulus remained unchanged 300 times the shear strength exceptions were those struts that had load cells installed; these
of the jet grout. struts were assumed to be loaded to 100% of their design preload.
Other factors may also influence the results. For instance, the However, as will be discussed, the 3D analysis would suggest that
strength of the jet-grout layer may be highly nonuniform. Chew loss of strut preload is almost inevitable with a sequential preload-
et al. (1997) reported jet grout unconfined compressive strengths ing process.
from the Singapore River contracts one to three ranging from
Construction Sequence
approximately 500 to almost 4,000 kPa, with a mean strength of
approximately 1,300 kPa. This implies a mean undrained shear The construction sequence consisted of 10 layers of excavations
strength of approximately 650 kPa, which is much more than alternating with nine layers of strutting. Each of the first nine layers
the value adopted. However, the response of the weaker locations of excavation was modeled as a uniform lift. What actually took
may dominate overall jet-grout behavior. Jet-grouted soil mass is place during the early phases of excavation was uncertain because
known to have a much lesser mass modulus than the jet-grout detailed record of the as-built construction sequences was unavail-
material. For instance, Nakagawa et al. (1996) reported that the able. For the 10th level of excavation, information from the Land
back-analyzed mass modulus of a jet-grout blanket used for exca- Transport Authority shows that the excavation proceeded from east
vation support was only approximately 13% of its designed value. to west and that the collapse occurred just after the soil at the
Similarly, O’Rourke et al. (1998) reported that the mass modulus eastern end of the excavation was removed. This was modeled
of deep mixing soil-cement columns was only 50% of the value in the analysis.
obtained from element tests.
As Fig. 3 shows, two F1 layers are in the site: one below the fill Comparison with Field Measurements
and another above the F2 layer. Both are relatively thin and their Fig. 6 shows the computed and measured wall deflection at the last
influence on the overall behavior of the ground was considered to stage of excavation shortly before the collapse. The measured wall
be insignificant. For this reason, the upper layer of F1 was treated deflection profiles were obtained from two inclinometers installed
as part of the sand fill, whereas the lower layer was treated as part of opposite one another, just behind the south and north walls within
the F2. For the F2, the SPT blow count of 10–20 blows=300 mm the collapsed zone. The computed deflection was taken from

Wall deflection (m) Wall deflection (m)


-0.2 0 0.2 0.4 0.6 -0.6 -0.4 -0.2 0 0.2
0 0
S337
5 5
S336
S335
10 10
S334
15 S333 15
Measurement
20 20
Depth (m)
Depth (m)

25 25

30 30

35 35

40 40

45 45

50 50

Fig. 6. Measured and computed wall deflection within the collapse zone shortly before collapse: (a) south wall; (b) north wall

INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / NOVEMBER/DECEMBER 2011 / 533

Int. J. Geomech. 2011.11:529-539.


five strut sections around the locations of the two inclinometers. deflects much more than the north wall just before collapse
The computed and measured deflection profiles were in reasonable (Fig. 6). At later stages of the collapse, the north wall joints also
agreement, considering the uncertainties and complexities involved open up because of the effects of local curvature and differential
in the analysis. For the south wall, the computed depth of maximum deflection among panels because of “snap-through” of the north
deflection was more than the measured depth by approximately 5 m. wall. As Fig. 1 shows, wall break-up along wall-to-wall joints was
This may be because of the nonuniformity in the jet-grout layer. indeed evident postcollapse.
Strut buckling also initiates from the zone of largest wall de-
Collapse Behavior flection and spreads laterally eastward, westward, and vertically
Fig. 7 shows the development of wall deflection during the exca- upward. However, eastward propagation was limited as the collapse
vation and collapse, viewed from the eastern end of the excavation. initiates close to the eastern end. Toward the end of the analysis, the
As is shown large wall deflection initiated after the 10th level of failure zone has propagated vertically to the top level of struts and
excavation from the bottom of the excavation and approximately laterally westward over approximately 15 strut sections. At this
four to five strut sections away from the eastern end of the exca- point, the computation could no longer continue.
vation; this is also consistent with eye-witness accounts of the ini- As Fig. 7(a) shows, the wall deflection at this eastern end is al-
Downloaded from ascelibrary.org by New York University on 05/17/15. Copyright ASCE. For personal use only; all rights reserved.

tial location of strut buckling (COI 2005). ready more than that at other zones within the modeled domain,
As the collapse progresses, the diaphragm wall panels start to even during the early stages of excavation. This can be attributed
break-up at the wall-to-wall joints, initiating with the south wall to the thicker layer of soft marine clay at this end of the excavation
joints. This can be attributed to the development of circumferential and the possible presence of a weak jet-grouted zone beneath the
stresses in the walls as the latter deflects. Both the north and south 66 kV cable. This illustrates the possible importance of hetero-
wall alignments are curvilinear. As the south wall deflects into the geneity and local variations to overall behavior. One may indeed
excavation, the radius of curvature of the wall alignment also in- surmise that, had the last level of excavation not started from
creases, thereby giving rise to tensile strains along the wall lines. the eastern end, in which the soft soil layer is the thickest, the col-
Such tensile strains are not uniformly distributed; because the joints lapse might have been avoidable.
are weaker and more flexible than the wall panels, tensile strains The break-up of the wall joints prevented significant redistrib-
localize at the wall joints, leading eventually to the break-up of the ution of the strut loads before buckling of the struts, which might
joints. Examination of the computed results indicates that many of have occurred if a competent global waler system had been in
the connectors representing the wall joints broke up even before place. Arching in soft soil conditions is not an effective means
collapse becomes evident. of redistributing strut loads because it requires large ground defor-
On the other hand, as the north wall deflects into the excavation, mation to take place. One may also surmise that, in a curvilinear
the radius of curvature decreases, thereby giving rise to compres- trench excavation such as this, the absence of competent global
sive strains along the wall lines, which decrease the likelihood walers may lead to numerous zones of instability because it is
of wall break-up. The difference between the circumferential ac- almost impossible to ensure that each and every strut is perfectly
tions of the north and south walls also explains why the south wall aligned with the earth pressure resultant behind each wall panel.

Fig. 7. Wall deflection profiles at various stages of analysis (viewed from eastern end); darker contours represent larger deflection; (a) after strut level
2 (from the top); (b) just after initiation of collapse; (c) during collapse; (d) at the end of analysis

534 / INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / NOVEMBER/DECEMBER 2011

Int. J. Geomech. 2011.11:529-539.


Excavation-Pile Interaction: Common Services reclamation sand fill. This was underlain by a layer of marine clay
Tunnel with a thickness varying from approximately 13 to 24 m. Around
the excavation area was a 5–6 m thick F2 fluvial soil layer under-
Background lying the marine clay. Below this laid soils of the Old Alluvium
formation. All four boreholes showed a thin transition layer of
The Common Services Tunnel was built to accommodate services Old Alluvium in which the SPT blow count increased rapidly from
in the Marina Bay area in Singapore. It was constructed by a cut- approximately 20 to 80 blows=300 mm. This was underlain by a
and-cover approach, which involves excavating a deep trench into stiff Old Alluvium layer with blow count ranging from approxi-
the ground. The depth of the excavation varies from approximately mately 80 to well over 100 blows=30 mm. The depth of the
10 to 28 m below ground surface. At the section analyzed, the for- groundwater was approximately 2.3 m A layer of jet-grouted soil
mation level was 19 m below ground surface. The temporary earth was also installed within the excavation area just below formation
retaining wall was a composite structure composed of continuous
level to serve as an underground strut.
sheet piles on the retained side installed back-to-back with a row of
soldier piles on the excavation side. The sheet piles and soldier piles Finite-Element Model
Downloaded from ascelibrary.org by New York University on 05/17/15. Copyright ASCE. For personal use only; all rights reserved.

were fabricated from grade S275 steel, which has a nominal yield
stress of 275 MN=m2 . The sheet piles used were FSP IV sections, A commercial software, GeoFEA (GeoSoft 2006), was used for this
driven to a depth of approximately 24.8 m below ground surface. study because trials indicated that its iterative solver enabled much
The second moment of area and unit mass per meterrun of the faster runtimes than ABAQUS with a desktop personal computer;
sheet pile wall were 1:04 × 103 m4 and 380 kg=m-length of pile, the speed-up ratio was of the order of three to 10 times. As shown in
respectively. The soldier piles used were 610 × 305 × 149 kg=m Fig. 9, the ground domain was represented by 20-noded brick el-
universal beam sections, spaced 800 mm centertocenter and driven ements with pore pressure degrees of freedom, which allowed fully
to a depth of approximately 37.5 m below ground surface. The coupled analyses to be conducted. The ground surface was assumed
second moment of area and mass per unit length of each soldier to be flat; this is reasonable because the site was on reclaimed land.
pile were 1:259 × 103 m4 and 149.2 kg. The retaining walls were The finite-element mesh is shown in Fig. 9. The sheet pile wall
supported by cross-struts and walers. Before excavation, 1.5 m was modeled by using reduced integration brick elements to en-
diameter bored piles were also predriven into the excavation area hance its ability to model bending (Hong et al. 2003). Pore pressure
to support the tunnel. Three-dimensional analyses were conducted degrees of freedom were not incorporated into the sheet pile wall.
to study the additional bending moments and uplift forces exerted This effectively models the sheet pile wall as an impermeable wall.
on these bored piles by soil movement in the excavation. Above the toe of the sheet piles, the bending stiffness of the soldier
piles is smeared and added onto that of the sheet piles. This is
Site Conditions unlikely to deviate significantly from reality because the soldier
The Common Services Tunnel is located in reclaimed land over- piles are closely spaced and the walers help to distribute loading
lying soft marine clay. Owing to the relatively small number of laterally. Beam elements were used to model the soldier pile seg-
boreholes available at this location, the spatial variation in stratifi- ments below the toes of the sheet piles, walers, and struts.
cation could not be deduced and therefore could not be fully con- Six bored piles were modeled within the trench excavation fol-
sidered in this study. The few boreholes sunk reflect the typical lowing the layout in the field; these were modeled by using reduced
profile of reclaimed land overlying soft marine clay. As Fig. 8 integration brick elements without pore pressure degrees of free-
shows, the top 10–12 m of soil was a loose to medium dense dom with the same cross-sectional area as the actual bored piles.

Fig. 8. Soil profile deduced from four boreholes at the Common Services Tunnel site

INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / NOVEMBER/DECEMBER 2011 / 535

Int. J. Geomech. 2011.11:529-539.


Downloaded from ascelibrary.org by New York University on 05/17/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 9. Finite-element mesh: (a) third-angle view after excavation; (b) plan view with piles annotated

The Young’s modulus was also adjusted so that the flexural rigidity at the far end of the retained soil domain was constrained to remain
EI is also the same as that of the actual piles. constant (that is, at their initial hydrostatic values) with respect to
The four vertical sides of the mesh were fixed against movement time. This allows the far end to act as a recharge boundary should
normal to the respective sides, whereas the base of the mesh was the excavation cause any drawdown.
fixed against all movements; this is reasonable because the already
stiff soil at the base is not expected to slide along the even stiffer Material Properties
soil or rock beneath the base. The initial groundwater table was Tables 2 and 3 summarize the engineering properties of the respec-
assumed to be located 1 m below ground surface. Pore pressure tive soil layers adopted in this study. As Tables 2 and 3 show, the

Table 2. Properties of Fill, Transition Layer, and Old Alluvium (Modelled as Mohr-Coulomb Materials)
Bulk unit Effective Young’s Effective Effective Effective angle Angle of Coefficient of Earth pressure
weight γ modulus E0 Poisson’s cohesion c0 of friction ϕ0 dilation ψ permeability k coefficient
Soil type (kN=m3 ) (kN=m2 ) ratio ν 0 (kN=m2 ) (degrees) (degrees) (m=s) K0
Fill 19 10,000 0.3 2 30 0 1 × 106 0.5
Fluvial clay 19 8,000 0.3 5 25 0
Transition 20 33,000 0.3 5 35 10 1 × 107 0.7
Old Alluvium 20 130,000 0.3 20 37 10 1 × 107 1.0
Jet grout 16 150,000 0.2 400 0 0 1 × 109 0.7

536 / INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / NOVEMBER/DECEMBER 2011

Int. J. Geomech. 2011.11:529-539.


Table 3. Properties of Marine and Fluvial Clays (Modelled as Modified Cam Clays)
Bulk unit Isotropic Isotropic Reference void Critical state Effective Coefficient of Earth Isotropic over-
weight γ swelling compression ratio (taken at 1 kPa friction Poisson’s Permeability k pressure consolidation
Soil type (kN=m3 ) index κ index λ confining stress) ecs coefficient M ratio ν 0 (m=s) coefficient K 0 ratio
Marine clay 16 0.093 0.27 2.62 0.87 0.25 1 × 109 0.7 1.5

marine clay was modeled by using the modified Cam Clay model. angle of friction corresponding to peak strength ϕ0max , and the criti-
Lee (2008) reported that the modified Cam Clay gives a reasonable cal state angle of friction ϕ0c , which has the form
description of the behavior of remoulded, reconsolidated Singapore
marine clay. However, the in situ clay is a structured clay, and ϕ0max  ϕ0c ¼ 0:8ψmax ¼ 5I R ðdegreesÞ ð1Þ
the use of the modified Cam Clay model to represent in situ marine in which ψmax = maximum angle of dilation; and I R = relative dilat-
clay would be conservative if one matches the critical state ancy index. Eq. (1) only holds for plane strain conditions and for
Downloaded from ascelibrary.org by New York University on 05/17/15. Copyright ASCE. For personal use only; all rights reserved.

strength of the model to that of the actual clay. The use of the sand. Bolton (1986) also suggested that for triaxial conditions
Mohr-Coulomb model with effective stress parameter was not
recommended for Singapore marine clay because it was considered ϕ0max  ϕ0c ¼ 3I R ðdegreesÞ ð2Þ
to be unconservative (COI 2005). In fact, it was one of the main
factors leading to the underdesign of the Nicoll Highway excava- Taken together, Eqs. (1) and (2) imply that
tion discussed earlier. In principle, GeoFEA allows users to
ϕ0max  ϕ0c ¼ 0:48ψmax ð3Þ
prescribe their own soil models. However, the limited site investi-
gation data preclude the use of a more sophisticated soil model. In which would allow the maximum angle of dilation to be estimated.
any case, no structured soil model has so far been validated for As shown in Table 4, the use of Eq. (3) with triaxial test data from
Singapore marine clay. three samples leads to maximum angles of dilation ranging from
The other soils in this site are much stiffer, and their behavior approximately 14 to 16° on the basis of an assumed critical state
is similar to that of stiff overconsolidated soils, with dilation angle of friction of 30°. For the Mohr-Coulomb model, a strain-
and strain softening rather than strain hardening. For such soils, averaged angle of dilation is probably more appropriate than the
the use of strain hardening models is inappropriate. Because the maximum value. In this study, parametric studies were conducted
strain-softening portion of the stress strain is difficult to model ac- with different angles of dilation for the Old Alluvium, ranging from
curately, a conservative approach that is widely used in practical 0° to the maximum. As will be discussed later, reducing the angle of
geo engineering and also used in this paper is the Mohr-Coulomb dilation did not always lead to more conservative results.
model with the strength set close to the critical state strength rather Table 5 shows the properties of reduced integration brick ele-
than the peak strength. Admittedly, in such soils, it may be impor- ments used to model the sheet pile–soldier pile combination and
tant to model the preyield nonlinear stress-strain behavior of the the bored piles. The sheet pile–soldier pile combination was mod-
soil. However, this is often difficult to achieve in practical geo con- eled by using anisotropic material properties to reflect the fact that
struction projects such as this, owing to lack of sufficiently reliable its bending rigidity in the horizontal direction is much smaller than
and accurate data that would justify the use of such models. For that in the vertical direction. In Table 5, the value of the Young’s
this reason, small strain parameters are usually unavailable. Hence, modulus in the horizontal direction Ehh is set at 1=10 of that in the
linear elastic behavior was assumed before yielding. vertical direction. The actual ratio is unknown, and different ratios
The angle of dilation was not available from site investigation between 1=100 and 1 were tried. It was found that, when the ratio
data. The fill and F2 layers were not considered to be strongly dilat- drops below 1=10, further decrease in this ratio did not change the
ive; hence their angle of dilation was measured to be 0°. However, results significantly, presumably because the bending rigidity of
triaxial test data showed that the transitional and stiff Old Alluvium the lateral waler was sufficient to redistribute the loading in the
soils were dilative. Bolton (1986) proposed an empirical relation- horizontal direction. In the analysis, the ratio of 1=10 is adopted.
ship for estimating the maximum angle of dilation ψmax , the secant Table 6 shows the beam type used for the struts, rakers, and walers.

Table 4. Angles of Dilation Estimated from Angles of Friction


Depth of Angle of friction—peak Critical state angle of Maximum angle of
Borehole consideration (m) strength ϕ0max (degrees) friction ϕ0c (degrees) dilation ψmax a (degrees)
A4 MZ2 44.0–44.9 37 30 14
A3 MZ1 49.0–50.0 38 30 16
A12 MZ2 54.5–55.5 37 30 14
a
See Eq. (3).

Table 5. Properties of Sheet Pile–Soldier Pile Combination and Bored Pile Modelled by Using Reduced Integration Brick Elements
Equivalent Young’s modulus in Equivalent Young’s modulus in Poisson’s Poisson’s
Structure horizontal direction Eh (kN=m2 ) vertical direction E v (kN=m2 ) ratio ν hh ratio ν hv
Sheet pile–soldier pile combination 1,723,420.7 1,723,4207 0.2 0.2
(above toe of sheet pile)
Bored pile 23:753 × 106 23:753 × 106 0.2 0.2

INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / NOVEMBER/DECEMBER 2011 / 537

Int. J. Geomech. 2011.11:529-539.


Axial Force (kN)
Table 6. Properties of Walers, Struts, and Rakers
-12000 -8000 -4000 0
Structure type (beam) Section type (kg=m) 18 -2160
-2290 -1873
Waler 1 2 × UB400 × 400 × 168
Waler 2 2 × UB400 × 400 × 232 23
Strut 1 2 × UB610 × 305 × 149
Strut 2 2 × UB838 × 292 × 226 28
Raker 1 2 × UB610 × 305 × 149
Raker 2 2 × UB838 × 292 × 226

Depth (m)
33 -9950 -8220

-9474
GeoFEA has a built-in library of standard beam elements that 38
allows the properties to be automatically set once the beam size
Downloaded from ascelibrary.org by New York University on 05/17/15. Copyright ASCE. For personal use only; all rights reserved.

is selected.
43
The properties of the bored pile do not reflect the real properties Pile 6
of concrete. Instead, they were adjusted to match the effective Pile 4
flexural rigidity of the actual pile, which was computed by ignoring 48 Pile 2 -3219 -2729
the rigidity of the tensile section of the concrete. -3067

Bored Pile Displacement, Bending Moment, and Fig. 11. Tensile force profiles for bored piles 2, 4, and 6
Tension Profiles
Fig. 10 shows the deflection and bending moment profiles of three uniform. The bond strength between jet grout and bored pile is
bored piles in the excavation area after excavation to final forma- likely to have significant effect on the tension in the pile. Among
tion. As is shown, pile deflection is largest at the top and decreases the piles, pile 2 which was nearest to the wall, has slightly less
monotonically with depth, reaching 0 at the stiff Old Alluvium stra- tension. This is not surprising; in a narrow trench excavation, heave
tum. The bending moment profiles reflect the deflection profiles may increase with distance away from the face and a larger heave
with positive bending moment at the top because of the coupling will lead to a larger tensile force.
of the heaving jet-grout layer and the deflected pile. Below that, the Parametric studies showed that the main factor affecting the
bending moment is generally negative because of the fixed end con- bending moment and tensile force is the bond strength between
dition imposed by the stiff stratum. This suggests that the piles jet grout and pile. Increasing the bond strength increases the tensile
behave much as fixed-ended beams, with the stiff stratum and the force on the piles but suppresses the heave and limits the lateral
jet-grout acting as fixed ends. Comparison of bending moment in movement and thereby the bending moment. Reducing the bond
the piles shows that the piles nearest to the wall were subjected to strength has the opposite effect.
the largest bending moment, whereas the furthest piles were sub- The effect of reducing the angle of dilation of the Old Alluvium
jected to the smallest bending moment. was also studied. Reducing the angle of dilation leads to a slight
As shown in Fig. 11, in contrast to bending moment, which is decrease in the bending moment and the tensile force. This is be-
limited to the upper segment of the pile, the tensile force is more cause a smaller angle of dilation reduces the “locking-up” effect of

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


-30 -20 -10 0 -30 -20 -10 0 -15 -10 -5 0
18 -25.1 18 -16.9 18 -9.3
-25.5
1302 -19.8 1540 1274
23 23 23 -10.8

28 28 28
Depth (m)
Depth (m)

Depth (m)

33 33 33
-1189 -1075 -568

38 38 38

43 43 43

48 48 48

-2000 -1000 0 1000 2000 -2000 -1000 0 1000 2000 -1000 0 1000 2000
(a) Bending Moment (kNm) (b) Bending Moment (kNm) (c) Bending Moment (kNm)

Legend
Deflection
Bending moment

Fig. 10. Deflection and bending moment profiles of three bored piles in the excavation: (a) pile 2; (b) pile 4; (c) pile 6

538 / INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / NOVEMBER/DECEMBER 2011

Int. J. Geomech. 2011.11:529-539.


the soil in a confined situation, thereby allowing the soil to move large-scale 3D analyses are now within reach of the design office.
past the pile more readily and generating less lateral drag and uplift The rapidity of the advances is shown by comparing the scales and
on the piles. Wall deflection is, however, slightly increased. This computational times of the two previous problems. The Nicoll
comparison illustrates the complexity of real scenarios and the Highway analysis, conducted in 2004, was a total stress analysis
fallacy of the commonly held belief that using a smaller angle involving approximately 150,000 degrees of freedom. It took ap-
of dilation will necessarily give a more conservative result. proximately 5 days to analyze from the start of the excavation to the
At the time of writing, excavation work was drawing to a close 10th lift of excavation, which is just before collapse. The Common
but the writers were unable to get access to field data because they Services Tunnel analysis, conducted in 2007, was an effective
were still not officially released. In any case, the purpose of this stress analysis by using higher order elements and involving
discussion is to demonstrate the usefulness of 3D analysis and 400,000 degrees of freedom. It took approximately less than 2 days
to show that such an analysis can be conducted in a practical en- to complete one run of the analysis.
gineering setting, rather than to validate parameters. One may study
the pile-soil interaction problem with a 2D plane strain analysis but
it will almost involve significant simplification and “smearing” in References
Downloaded from ascelibrary.org by New York University on 05/17/15. Copyright ASCE. For personal use only; all rights reserved.

the third dimension.


Blackburn, J., and Finno, R. (2007). “Three-dimensional responses
observed in an internally braced excavation in soft clay.” J. Geotech.
Geoenviron. Eng., 133(11), 1364–1373.
Conclusions: Lessons Learned from the Previous Bolton, M. D. (1986). “The strength and dilatancy of sands.” Geotechnique,
Examples 36(1), 65–78.
Chew, S. H., Lee, F. H., Lee, Y., and Yogarajah, I. (1997). “Jet grouting in
In many real scenarios, 3D effects are not only significant, but they singapore marine clay.” Proc., 3rd Young Geotechnical Conf., T. S. Tan,
may also be subtle, and the interaction among different components S. H. Chew, K. K. Phoon, and T. G. Ng, eds., National University of
may be complex. Two-dimensional idealizations may oversimplify Singapore, Singapore, 231–238.
the problem and lead to unreliable or incorrect results. For instance, Chin, K. G. (2006). “Constitutive behavior of cement treated marine clay.”
in the case of the Nicoll Highway, only 2D plane strain finite- Ph.D. thesis, National University of Singapore, Singapore.
element analyses were conducted before the collapse. Circumferen- Committee of Inquiry (COI). (2005). “Report of the Committee of Inquiry
tial compression-tension forces on the wall line were only highlighted into the incident at the MRT circle line worksite that led to the collapse
by 3D analysis postcollapse. One may thus surmise whether the large- of Nicoll Highway on 20 April 2004.” Vol. 1, Part 1, Ministry of
Manpower, Singapore.
scale global nature of the collapse could have been averted if 3D
Dames, and Moore (1983). “Singapore mass rapid transit system—detailed
analyses had been conducted and a set of competent global walers geotechnical study interpretative report.” Provisional Mass Rapid
installed to enable better load redistribution. In addition, the 3D Transit Authority, Singapore.
analyses would have drawn attention to the criticality of the eastern Finno, R. J., and Harahap, I. S. (1991). “Finite element analysis of HDR-4
end of the excavation, which might have prompted a rethought on the excavation.” J. Geotech. Engrg., 117(10), 1590–1609.
wisdom of starting the excavation at the most critical location. GeoSoft (2006). GeoFEA users’ manual 8.0, Singapore..
Another lesson was also learned regarding preloading of the Hong, S. H., Lee, F. H., and Yong, K. Y. (2003). “Three-dimensional pile-
struts. In the early stages of the study, attempts to make the preloading soil interaction in soldier-piled excavations.” Comput. Geotech., 30(1),
of the struts as uniform as possible turned out to be generally unsuc- 81–107.
cessful. This was because struts within the same layer were preloaded Lee, F. H. (2008). “How useful is numerical analysis in geotechnical
engineering?” Proc., Int. Conf. on Deep Excavation (ICDE) 2008, Land
sequentially in actual construction and in the analysis, not simulta-
Transport Authority, Singapore.
neously. In such situations, preloading another strut next to an Lee, F. H., Yong, K. Y., and Liu, K. X. (1995). “Three-dimensional analy-
already preloaded strut will cause the retaining wall to move back ses of excavation in soft clay.” Proc., 11th African Regional Conf.—
slightly, resulting in a loss of preload in the previously preloaded Cairo ’95, Egyptian Geotechnical Society, Cairo, Egypt.
strut. This cannot be observed in a 2D analysis, which can only sim- Lee, F. H., Yong, K. Y., Quan, K. C. N., and Chee, K. T. (1998). “Effects
ulate, what is in effect, simultaneous preloading of all struts in a layer. of corners in strutted excavations—field monitoring and case histories.”
Hence, the notion that all struts within the layer will have the same J. Geotech. Geoenviron. Eng., 124(4), 339–349.
preload, as was designed, may be a myth; it can only be achieved by Nakagawa, S., Kamegaya, I., Kureha, K., and Yoshida, T. (1996). “Case
simultaneous or iterative preloading, which is hardly ever done. history and behavioural analyses of braced large scale open excavation
The second example highlighted the usefulness of 3D analysis in very soft reclaimed land in coastal area.” Proc., Int. Symp. on Geo-
technical Aspects of Underground Construction in Soft Ground, A. A.
in what is becoming an increasingly common problem in urban
Balkema, London, 179–184.
underground construction works. The three-dimensionality of such O’Rourke, T. D., McGinn, A. J., Dewsnap, J., and Stewart, H. E. (1998).
problems has long been recognized; however, software and hard- “Case history of an excavation stabilized by deep mixing methods.”
ware limitations have, in the past, prevented 3D analyses. Instead, Design and construction of earth retaining systems, ASCE Geotechni-
piles are often smeared into pile walls in 2D plane strain analyses. cal Special Publication, No. 83, R. J. Finno, Y. Hashash, C. L. Ho, and
Two-dimensional analyses had been conducted during initial de- B. P. Sweeney, eds., ASCE, Reston, VA, 41–61.
sign, but these were found to be unable to provide reliable values Ou, C. Y., and Chiou, D. C. (1993). “Three-dimensional finite element
of pile displacement and loadings. The use of 3D analyses will analysis of deep excavation.” Proc., 11th Southeast Asian Geotechnical
allow such problems to be analyzed as what they really are. This Conf., The Institution of Engineers, Malaysia, Kuala Lumpur, Malaysia,
example also highlights the complexity of the interaction among 769–774.
Wong, L. W., and Patron, B. C. (1993). “Settlements induced by deep
the retaining wall, soil, and the piles in the excavation area, and
excavations in Taipei.” Proc., 11th Southeast Asian Geotechnical Conf.,
the fallacy of the possibly commonly held belief that using a lesser The Institution of Engineers, Malaysia, Kuala Lumpur, Malaysia,
angle of dilation is necessarily more conservative. 787–791.
In the past, 3D analyses have been hampered by software and Wroth, C. P. (1975). “In-situ measurement of initial stresses and deforma-
hardware limitations. However, hardware and software advances tion characteristics.” Proc., Specialty Conf. in In-Situ Measurement of
have now reduced computational times to the point that even fairly Soil Properties, ASCE, Reston, VA, 181–230.

INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / NOVEMBER/DECEMBER 2011 / 539

Int. J. Geomech. 2011.11:529-539.

You might also like