Marine Georesources and Geotechnology, 23:117136, 2005

Copyright # Taylor & Francis Inc.

ISSN: 1064-119X print/1521-0618 online
DOI: 10.1080/10641190590953728

Instrumentation and Numerical Analysis of

Cylindrical Diaphragm Wall Movement During
Deep Excavation at Coastal Area
DONG SOO KIM
Department of Civil & Environmental Engineering,
Korea Advanced Institute of Science and Technology,
Taejon, Korea

BYOUNG CHUL LEE

Department of Civil Engineering,
Shingu University, Sungnam, Korea
The lateral deflection of a cylindrical diaphragm wall and the associated ground
movement induced by deep excavation are analyzed by performing site instrumentations and numerical analyses in the coastal area of Korea. Wall lateral deflection,
rebar stress, and pore water pressure were measured and analyzed in eight directions. Variations of soil properties with the decrease of confining pressure are compared by performing various in situ tests before ad after excavation. To calculate the
wall lateral deflection accurately, the effects of small strain nonlinearity, confining
pressure, and the hysteresis loading=unloading loop developed during excavation are
considered in the proposed numerical analysis. By comparing numerical results with
measured ones, the importances of considering small strain nonlinearity and confining pressure reduction in the nonlinear (FEM) are emphasized. Also, the effects of
wall stiffness on the performance of cylindrical diaphragm walls are studied for
future similar excavation in the coastal area.
Keywords cylindrical diaphragm wall, lateral deflection, instrumentation, FEM,
confining pressure, small-strain nonlinearity

Although the reliable evaluation of ground movement has become essential for the
economy and safety of major offshore construction projects, the current techniques
for evaluation of marine soil properties and design procedure show a considerable
lack of accuracy when compared with the instrumented results. This is usually
attributed to a misunderstanding of marine soil behavior under working load
conditions.
The strain level experienced in the soil medium under working load condition
usually ranges below about 0.5%, and soil behaves nonlinearly from the small strain
ranges of about 10
3% (Jardine et al. 1986; Burland 1989). The stiffness of granular
soils and weathered rock is considerably affected by the in situ confinement, and the
Address correspondence to Dong Soo Kim, Department of Civil & Environmental
Engineering, Korea Advanced Institute of Science and Technology, Taejon, Korea.
E-mail: dskim@kaist.ac.kr

117

118

D. S. Kim and B. C. Lee

modulus values based on (SPT) obtained by site investigations before major excavation would be quite different from those values after excavation. In current design
practice, subgrade reaction method and=or a linear finite element method are often
employed for the analysis of deformational behavior of geotechnical offshore structures using soil properties determined by SPT N-values and=or a conventional triaxial tests which cannot properly consider the effects of confinement, small-strain
nonlinearity and hysteresis loading=unloading loop. Therefore, it is necessary to
develop the refined site investigation and numerical analysis methods which can consider the aforementioned effects for the reliable ground movement analysis during
deep excavation at coastal area.
In this article, the case study of the deep excavation for a 56 m-depth cylindrical
inground LNG storage tank at coastal area was investigated. The 1.7 m thick, 80 m
diameter, and 75 m deep cylindrical diaphragm wall was utilized as a retaining structure for the excavation. Detailed site investigations were performed both before and
after excavations to determine the variation in deformational characteristics due to
the decrease of confinement. The small-strain nonlinearity of the site was evaluated
by effectively combining the maximum shear moduli determined by a downhole test
with the modulus reduction curves determined by a resonant column test. A number
of instruments were installed at the diaphragm wall and the adjacent ground, and the
distributions of wall deflection, pore water pressure, and rebar stress were monitored
at various locations during the excavation (Samsung Corporation 1998). The result
of series of numerical analyses was compared with the monitored wall deflection,
and the importance of considering the effects of small-strain nonlinearity and confining pressure was assessed. Finally, the effects of wall stiffness on the performance of
cylindrical diaphragm walls were evaluated for the future similar construction in the
coastal area.

Outline of Diaphragm Wall and Excavation Work

Inground LNG storage tanks were constructed at Inchon in the west coast of Korea.
The capacity of each storage tank is 200,000 kl; its diameter is 80 m; and its inner
excavation depth is 56 m. The 1.7 m thick and 75 m deep cylindrical diaphragm wall
was constructed as a retaining structure to withstand the earth and water pressures
during excavation without any prop system. The horizontal support is achieved by
virtue of hoop-compression stresses on the circular wall. The wall was installed into
the soft rock by 3 m and cast by 26 panels. The earth retaining wall was constructed
by the slurry wall construction method. The cross section of the LNG storage tank is
shown in Figure 1.
The excavation was carried out at eight stages to prevent the occurrence of
nonuniform lateral pressures. The ground water table inside the retaining wall was
lowered by deep wells and maintained its level about 1 m below the bottom of excavation at each stage. The excavation works proceeded smoothly up to the depth of
about 50 m when a thin impermeable clay layer was encountered. Once the clay layer
was excavated, a significant amount of seepage occurred into the excavated area and
further excavation could not be continued. To stabilize the inward seepage, injection
grouting was performed into the weathered and soft rocks through the holes installed
along the diaphragm wall. Detailed site investigation was performed again and the
deformational characteristics of the inside soil can be evaluated before and after
excavation.

Analysis of Wall Movement During Deep Excavation

119

Figure 1. Cross section of inground LNG storage tank.

Site Profile and Soil Properties

Geological Profile
The site consisted of a layer of reclaimed soil of about 10 m, a loose alluvial silty sand
of about 37 m, an alluvial clay of about 3 m, a sandy gravel of 5 m, a weathered rock
of about 17 m and soft rock (Figure 1). The degree of weathering of rock was severe
at the upper part, and its intensity was mitigated at the lower part. The alluvial clay
layer deposited at a depth of about 50 m was of interest in this study. The water table
was located at a depth of about 7 m. Both in situ and laboratory tests including SPT,
downhole, pressuremeter, and resonant column tests were performed to obtain the
deformational characteristics of the site (Table. 1).

Comparison of Soil Properties before and after Excavation

In most excavation work, site investigation is usually performed before the excavation. If the stiffness of soil is affected by confinement, the deformational characteristics of the soil determined after the excavation would be quite different from
those determined before excavation. In this study, the stiffness of the weathered
rock, which played the major role in resisting the movement of the wall, can be
evaluated before and after excavations.
In SPT, the safety type hammer of energy ratio of 60% was used. SPT N values
determined before and after excavations were compared in Figure 2(a), and it is
observed that N values of the weathered rock were significantly reduced as a result
of seepage and decrease of confinement due to the excavation. Menard-type

120

D. S. Kim and B. C. Lee

Table 1. Comparison of soil properties before and after excavation

Depth EL,m
7.4 
2.6

2.6 
16.3

16.3 
30.0

30.0 
39.9

39.9 
42.3

42.3 
47.3

47.3 
63.8

63.8 
72.1

72.1 
100.0

Soil
classification
Reclaimed full
Silty sand
Silty sand
Silty sand
Clay
Sandy gravel
Weathered rock
Soft rock
Hard rock

N-Value
before after
17
24
42
70
73
127
173

85
510

E 28 N (kPa)
before

44667

65883

115295

192158

204420

348630

474906 233335

1400011

Depth (EL, m)

Soil
classification

Unit weight
c (kN=m3)
(before)

7.4 
2.6

2.6 
16.3

17.3 
39.9

30.0 
39.9

39.9 
42.3

42.3 
47.3

47.3 
63.8

63.8 
72.1

72.1 
100

Reclaimed fill
Silty sand
Silty sand
Silty sand
Clay
Sandy gravel
Weathered rock
Soft rock
Hard rock

18.14
18.14
18.63
20.10
18.63
20.59
23.04
23.04
25.00

after

Vs (m=s)
before after
160
200
260
310
310
310
450
750
1500

246
400

Max. shear
modulus
Gmax (Pa)
(before)

Elastic
modulus
Emax (Pa)
(before)

4.736E7
7.400E7
1.284E8
1.970E8
1.826E8
1.922E8
4.759E8
1.434E9
5.738E9

1.260E8
1.968E8
3.417E8
5.240E8
4.857E8
5.113E8
1.266E9
3.815E9
1.526E10

Gmax and Emax are obtained from shear wave velocity Vs.

pressuremeter tests were also performed before and after excavations and elastic
modulus of weathered rock obtained after excavation was found to be considerably
lower than the modulus before excavation as shown in Figure 2(b).
Downhole test was performed before excavation to obtain the shear wave velocity (Vs) profile of the site as shown in Figure 3. At depths of 24 m, 34.5 m, and 40 m
during excavation, Spectral Analysis of Surface Wave (SASW) tests were performed
to evaluate the variation of vs with confinement. Because of the limited testing space,
the receiver spacing in the SASW tests was restricted. It was interesting to note that
the shear wave velocities measured after excavation were lower than the velocities
obtained before excavation, and the importance of considering the confinement
effect in the evaluation of deformational characteristics can be explained. Unfortunately, it was not possible to perform an SASW test at the bottom of the excavation
(at a depth of 55 m) because of construction problems. The shear wave velocity after
excavation was estimated by correcting the downhole test result based on confining
pressure influence factors of weathered and soft rocks determined by resonant
column tests as shown in Figure 3. The elastic modulus values determined by
SPT, PMT, and shear wave velocity measurements before and after excavation are

Analysis of Wall Movement During Deep Excavation

121

Figure 2. Comparison of modulus obtained by SPT and by PMT between before and after
excavation.

compared in Figure 4. In the Korean Specification of Highway Bridges (Korean

Highway Corp. 1996), the elastic modulus can be estimated as 28N based on SPT
N-values. The maximum elastic modulus at small strains was calculated using the
shear wave velocity. Modulus values determined by wave velocities were larger than
those determined by 28N, and this difference can be explained by the difference in

122

D. S. Kim and B. C. Lee

Figure 3. Comparison of shear wave velocity profiles with excavation steps.

strain amplitude as discussed later. The effects of confining pressure reduction on

modulus values before and after excavation are embodied in the proposed numerical
analyses by using the Janbu (1963) formula.

Figure 4. Comparison of elastic moduli between before and after excavation according to
various testing methods.

Analysis of Wall Movement During Deep Excavation

123

Figure 5. Normalized modulus reduced curve by RC tests at average effective confining pressure of each layer.

In the numerical analysis, the layer of weathered rock plays a major role in the
wall lateral deflection. The modulus values of weathered rock were significantly
reduced due to the disturbance and the decrease of confinement during excavation.

Figure 6. Comparison between elastic moduli determined by N-value, downhole test, and RC test.

124

D. S. Kim and B. C. Lee

If one used the modulus values determined before excavation, the stiffness of the
layer would be overestimated in the analysis.

Variation in Elastic Modulus with Strain

Because the soil exhibits the nonlinear behavior even at small strains and the strain
level experienced in the soil media is small (usually less than about 0.5%) under
working stress conditions, the evaluation of small strain nonlinearity is important
for the reliable analysis of wall deformation (Burland 1989). Samples were obtained
at eight depths for resonant column tests and the test results are shown in Figure 5.
Test results were fitted using Ramberg-Osgood model and fitting parameters were
employed in the proposed model discussed later. Then, the nonlinear stress-strain
relations were determined at each layer by effectively combining the maximum
modulus at small strain determined by the downhole test with the modulus reduction
curves determined by resonant column tests (Kim et al. 1997). The variations in elastic modulus values with strain levels are presented in Figure 6. The modulus values
determined by 7N and 28N based on Korean Specification of Highway Bridges
(Korea Highway Corp. 1996) were also plotted for comparison purposes. Modulus

Figure 7. Monitoring points and devices.

Analysis of Wall Movement During Deep Excavation

125

values at each layer were decreased as strain level increased, and it was interesting to
note that modulus value estimated by 7N was considerably low, but values estimated
by 28N were close to the value at a strain level of about 0.1%. Considering the fact
that the strain level in the working stress condition is usually less than 0.11.0%, it
may be appropriate to use 28N as a linear elastic modulus.

Instrumentation
An extensive instrumentation program was implemented, which included (1) inclinometers in eight directions to measure the wall lateral deflection, (2) pore water
pressure transducers in four directions to measure the variation in pore pressures
during excavation at four different depths, (3) rebar stress meters in the radial and
vertical directions, and so on (Figure 7). The measuring section was divided into
eight directions (31, 73, 128, 170, 211, 253, 308, 350). In the directions of 73
and 253 (MAIN LINE 1) all instruments were performed. In the directions of
170 and 350 (MAIN LINE 2) which were almost at a right angle to MAIN LINE
1, all instruments excluding pore water pressure meter in the surrounding ground
were installed. In other directions, minimum required instruments were installed.

Figure 8. Lateral wall deflections and average value in final excavation step.

126

D. S. Kim and B. C. Lee

The variations of wall deflections with depth were monitored at eight directions
by inclinometers installed in diaphragm wall. Circumferential rebar stress meters
were installed at five depths in eight directions. Pore water pressure transducers were
installed to measure the applied water pressure to the diaphragm wall and to confirm
security against permeation through the bottom. Pore pressure transducers were
installed at three depths in four directions outside of the diaphragm wall and at
one depth in four directions inside of the diaphragm wall.

Observed Performance during Excavation

Lateral Deflection of Diaphragm Wall with Excavation Steps
Figure 8 shows the measured wall deflections at eight different inclinometer locations
after final excavation to the depth of 56 m. The inclinometers were installed inside
the wall and the monitored deflections were corrected based on the absolute deflections monitored at the top of the wall accurately using a transit. The difference in
wall deflection at various locations can be explained by the occurrence of eccentric
force derived from surrounding ground.

Figure 9. Comparison of lateral deflection by using rebar stress and inclinometer value at 4th,
6th and final excavation steps.

Analysis of Wall Movement During Deep Excavation

127

The average value of measured deflections is plotted as a solid line in Figure 8.

The excavation was carried out at eight steps during two months, and the diaphragm
wall was supported only by the hoop stress of cylindrical diaphragm. The average
values of lateral wall deflections with excavation steps of 4th, 6th, and final are shown
in Figure 9. Since the shape of the diaphragm wall in cylindrical, the wall lateral
deflection can also be computed indirectly by measuring the rebar stress installed
in the wall. The rebar stresses are also plotted in Figure 9 for the comparison purposes. The calculated wall lateral deflection shape based on the rebar stresses with
excavation steps matched well the average deflection shape monitored by inclinometers, indicating the reliability of using the average deflection shape in the analysis.
Variation in Pore Water Pressure
Pore water pressures were monitored in four directions at depths of 18.4 m (EL-11 m)
and 52.7 m (EL-45.3 m) and outside of wall and at a depth of 70.4 m (EL-63 m) both
inside outside of the wall. The variations in measured pore pressures with time are
shown in Figure 10. On the 7th of October, pore pressures measured at EL-63 m began
to decrease sharply and then stabilized. At that moment, the impermeable clay layer that
existed at a depth of 47 m was excavated and the abrupt seepae started to occur due to
the pressure head difference between the outside and inside of the wall. In the design
stage, the clay layer was not considered as impermeable and the pore water pressure
was considered as hydrostatic, because it was expected that the wall embedded into soft
rock which would constrain the seepage. Judging from the implementation results, the
precise expectation of the permeability of the soft rock and the weathered rock were
important for estimating the real behavior of the surrounding ground and the wall.

Numerical Analysis
Proposed Nonlinear Elastic Model
In this study, the improved numerical modeling method was proposed to estimate
the behavior of the cylindrical diaphragm wall and surrounding ground movement

Figure 10. Results of pore water pressure measurement.

128

D. S. Kim and B. C. Lee

during deep excavation in coastal area. During excavation, soil strains in the working stress ranges usually experienced were below about 0.5% (Burland, 1989). At this
small strain range, stress-strain relationships of soil show nonlinearity and hysteresis
loop. The backbone curve is very important for the evaluation of loading=unloading
behavior because it defines the initial stiffness of the soil at small strains and constitutes the basis for characterizing the stress-strain behavior of soils for nonlinear
analysis. Ramberg-Osgood stress-strain relationship generally fits the nonlinearity
of the experimental data at strain range below 0.1% very well (Anderson and
Richart 1975). In this article, hysteresis loop can be represented by the backbone
curve, which is described by Ramberg-Osgood Parameters, and by assuming
Masing rule.
One form of the Ramberg-Osgood stress-strain equation for the inintial backbone curve can be written as (Eq. 1):

e1

r1
r3
Ei



r1
r3 R
;
C
Ei

where e1 is principal strain, r1
r3 is deviatoric stress, Ei Emax is initial modulus, C is dimensionless coefficient, and R is dimensioless exponenet.
Since the tangential modulus is defined as the instantaneous tangential slope of
the stress-strain curve, the tangential modulus can be calculated from (Eq. 1) by differentiating with respect to the major principal strain as shown in (Eq. 2).
Et

@r1
r3

@e1

1 CR

Ei


r1
r3
Ei

R
1 :

According to the excavation steps, the effective confinement decreases due to the
removal of surcharge load and seepage flow, and soil stiffness inside the wall after
excavation is reduced compared to that before excavation. Based on the site investigations, soil stiffness after excavation is estimated about 30% of the value before
excavation. Therefore, for the precise estimation of deformation behavior, it is necessary to consider a variation of soil stiffness due to confining pressure reduction by
excavation. The initial tangential modulus of stress-strain curve has been observed
to be dependent on confining pressure. Janbu (1963) proposed the following equation (Eq. 3) to consider the dependency of confining pressure for initial tangential
modulus Ei .
 n
r3
Ei Emax KPa
;
3
Pa
where, Pa is atmospheric pressure, initial tangential modulus factor K and stress
influence exponent n are nondimensional material parameters.
If (Eq. 3) is substituted into (Eq. 2), the tangential modulus of backbone curve
considering the effect of confining pressure at working load condition is represented
by (Eq. 4).
 n
KPa Pr3a
Et
4

R
1 :
1
r3
1 CR KParr
n
3 =Pa

Analysis of Wall Movement During Deep Excavation

129

Table 2. Values of parameter used in the proposed numerical analysis

Parameters
Fill
Alluvial soil
Weathered rock

230
420
1000

0.68
0.60
0.62

31622
79433
3981

2.57
2.97
2.21

Then, the tangential modulus for the unloading condition is represented by

considering Masing behavior.
 n
KPa Pr3a
Eur
5

R
1 ;
r1
r3
1 2CR
n
R
1
KPa r3 =Pa
where C and R are Ramberg-Osgood fitting curve parameters, and r1 and r3 are
major and minor principal stresses, respectively. These parameters are determined
from the in situ and laboratory tests and listed in Table 2. The advantages of using
the proposed model is that the soil behavior during excavation can be calculated
automatically and accurately by considering (1) small strain nonlinearity, (2) effect
of confining pressure reduction, and (3) hysteresis loop for unloading and reloading.
The proposed model was implemented in the computer program of ABAQUS
(1995). The validity and accuracy of the proposed model was verified by comparing
the results with hyperbolic model developed by Duncan et al. (1980). Because of the
limited space, the details cannot be included in this paper.

Finite Element Analysis

A series of 2D finite element analyses was performed on a deep excavation site using
the program ABAQUS (1995), where the proposed model is embodied by using
UMAT Subroutine. The parameters of the proposed model (K, n, C, R) were
defined at each layer and listed in Table 2.
For simplicity, an idealized axisymmetric plane strain excavation geometry with
an excavation depth H and a width of 40 m was considered as shown in Figure 11
The site was assumed to be composed of layers of fill, alluvial soil deposit and weathered rock overlying a soft rock stratum. A refined mesh, which consists of approximately 5828 nodes and 5673 elements as shown in Figure 12, was adopted to
minimize the effect of mesh dependency on the finite element modeling. Because
of the symmetry about the excavation centerline, only one-half of the excavation
was considered in the finite element model. The finite element mesh extends to a
depth of 1.0 H below the final excavation platform and laterally to a distance of
3.0 H from the excavation centerline. The soil and wall was discretized using 8-noded
isoparametric axisymmetric plane strain element. In addition, the interface between
the wall and retained soil was modeled using contact pair elements. The nonlinear
behaviors of the fill, alluvial soil deposit, and weathered rock were modeled by using
the proposed model. Soft rock and wall were treated as a linear elastic material. In
this study, excavation was simulated numerically by removing the elements rowwise
from the front of the wall. At each time of such excavation, the excavated surface

130

D. S. Kim and B. C. Lee

Figure 11. Excavation geometry.

was made stress free by calculating the equivalent nodal forces from the removed elements and applying them on the excavated boundary (Ishihara 1970). The implied
load due to excavation was found too large to be applied in a single increment for
the nonlinear soil. Hence the load was spread over a number of increments until

Figure 12. Finite element mesh.

Analysis of Wall Movement During Deep Excavation

131

numerical stability was established and convergence was achieved. An excavation

step is repeated until the excavation reaches the final depth EL-49 m.

Analysis Results
Comparison Between the Measured and Analyzed Results
Measured lateral wall deflections are compared to nonlinear FEM analysis results
with excavation steps in Figure 13. Soil parameters listed in Table 2 reflect the effect
of confining pressure reduction derived from ground excavation and small-strain
nonlinearity of soil modulus. Because the stiffness of the cylindrical diaphragm wall
with thickness of 1.7 m is high, lateral wall deflection is relatively small. FEM analysis results are well consistent with measured values at each excavation steps. On the
other hand, wall deflection obtained from using parameters before excavation shows
much smaller deflections at depths between 40 m and 55 m, compared to the measured data at final excavation step as shown in Figure 14. It is explained that parameters before excavation cannot consider the effect of confining pressure reduction
during excavation.

Figure 13. Comparison between measured and numerical analysis results with excavation
steps.

132

D. S. Kim and B. C. Lee

Figure 14. Comparison of results with soil properties before and after excavation.

Comparison Between Linear and Nonlinear Finite Element Analysis

Though nonlinear analysis predicts the actual soil behavior well, it can be somewhat
difficult to use routinely. As an alternative to nonlinear analysis, linear analysis can
be chosen with a constant value of modulus that corresponds to the strain range of
soil under working load. The deviatoric strain contour is shown in Figure 15 and the
maximum strain level is about 0.1%, which is consistent with the fact demonstrated
by Burland (1989). Wall deflections obtained by linear analysis using soil modulus
values at strains of 0.01% and 0.1% are compared with those by nonlinear analysis
in Figure 16. It has been found that the magnitude of soil modulus at strains of
0.01% and 0.1% are about 80% and 40% of maximum soil modulus, respectively.
Deflection curve obtained by nonlinear analysis is close to the curve obtained by
linear analysis with modulus value at strain of 0.01%, and both results match
reasonably well with measured curve. But the result obtained by linear analysis with
modulus value at strain of 0.1% overestimates the deflection because the modulus
values were underestimated. As shown Figure 15, the average strain level around
the diaphragm wall is about 3.5  10
2% with the maximum strain less than
0.12%, and the difference in deflection curves at each analysis can be explained by
the strain distribution. Even though the linear analysis with modulus values at strain
of 0.01% provides the better result compared to the nonlinear analysis, it is difficult

Analysis of Wall Movement During Deep Excavation

133

Figure 15. Contour of maximum deviatoric strain.

Figure 16. Comparison of results between linear and nonlinear analysis with strain levels.

134

D. S. Kim and B. C. Lee

to choose the proper strain level in the linear analysis, because it depends on various
factors such as soil and wall stiffnesses.
In the proposed case, the wall stiffness of the cylindrical diaphragm was so large
that lateral wall deflection appeared very small, but the effect of the nonlinearity of
soils would influence significantly lateral wall deflection in case of small wall stiffness. Using the numerical analyses performed with altering wall stiffness, the difference between linear and nonlinear analysis can be noticed clearly. For similar deep
excavation in the coastal area, the effects of wall stiffness on the lateral wall deflection have been investigated by varying the wall stiffness of EI 13510, 4752, 2750,
1408, and 594 MNm2, which corresponds to 1.7, 1.2, 1.0, 0.8, and 0.6 m thick cylindrical diaphragm walls, respectively. As shown in Figure 17, the lateral wall deflection decreases with increasing wall stiffness. The shape of lateral wall deflection with
wall stiffnesses shows similar pattern. The maximum lateral wall deflection at the
0.6 m, 0.8 m, 1.0 m and 1.2 m cylindrical diaphragm walls is increased by 93%,
64%, 43% and 27%, respectively, as compared to the maximum lateral wall deflection of the 1.7 m thick wall.
As shown in Figure 5, nonlinearity of soil varies soil modulus with strains. As
wall stiffness decreases, wall deflection increases as shown in Figure 17. The increase
of wall deflection comes from both the decrease of wall stiffness and the decrease of

Figure 17. Curves of lateral wall deflections with various wall stiffnesses.

Analysis of Wall Movement During Deep Excavation

135

soil modulus due to increase of soil strain. By comparing wall deflections obtained
by linear analysis with constant modulus values and nonlinear analysis for all cases
of wall stiffness, respectively, the effect of soil nonlinearity can be explained. In this
analysis, the soil modulus used in linear analysis is 80% of maximum value, which
corresponds the values at strains of about 0.01%, and the effect of soil nonlinearity
can be defined as the difference of maximum wall lateral difference between linear
and nonlinear analysis. It is found that the effect of soil nonlinearity increases with
decreasing wall stiffness. This means that wall lateral deflection is governed by not
only wall stiffness but also soil nonlinearity with decrease of wall stiffness. Therefore,
to evaluate exactly wall deflection and surrounding ground deformation, numerical
analyses must be performed considering the small-strain nonlinearity of soil as well
as the effect of confining pressure reduction during excavation.

Conclusions
The lateral deflections of a cylindrical diaphragm wall induced by deep excavation
are analyzed by performing site instrumentations and numerical analyses in the costal area of Korea. Comparing the elastic modulus values before and after excavation,
it was found that the elastic modulus is reduced significantly after the excavation due
to the effect of confining pressure reduction and the permeation. Therefore, the
effect of confining pressure reduction should be considered appropriately to determine the elastic modulus for an accurate evaluation of the lateral deflection of
diaphragm wall.
The stress-strain model, which enables consideration of the effects of soil nonlinearity and the confinement reduction, was proposed. The propsed model is implemented in ABAQUS, and its applicability to analysis of deep excavation works has
been found to be satisfactory through the verification with in situ measurement
results. To evaluate wall lateral deflection properly, numerical analyses should be
performed considering the effect of small-strain nonlinearity of soil and confining
pressure reduction during excavation.

References
ABAQUS. Users and theory manuals, V. 5.5. Hibbit, Karlson & Sorensen Inc.
Anderson, D. G. and F. E. Jr. Richart. 1975. Effects of shearing of shear modulus of clays.
Journal of the Geotechnical Eng. Div., ASCE 102(9): 975987.
Burland, J. B. 1989. Ninth Lauritis Bjerrum Memorial Lecture: Small is beautiful-The
stiffness of soils at small strains. Canadian Geotechnical Journal 26: 5265.
Duncan, J. M., P. Bryne, K. S. Wong, and P. Marbry. 1980. Strength, stress-strain and bulk
modulus parameters for finite element analyses of stresses and movements in soil masses.
Geotechnical Engineering Research Report No UCB/GT/80-01, University of
California, Berkeley, California.
Ishihara, K. 1970. Relations between process of cutting and uniqueness of solution. Soil and
Foundations 10(3): 5065.
Janbu, N. 1963. Soil compressibility as determined by oedometer and triaxial tests. Proc
Europe Conference on soil Mechanics and Foundation Engineering 1: 1925.
Jardine, R. J., D. M. Potts, A. B. Fourie, and J. B. Burland. 1986. Studies of the influence of
non-linear stress-strain characteristics in soil-structure interaction. Geotechnique 36(3):
377396.

136

D. S. Kim and B. C. Lee

Kim, D. S., K. C. Kweon, S. Y. Jeong, and J. Y. Park. 1997. Evaluation of nonlinear deformation characteristic of soil using laboratory tests and site tests. Journal of the Korean
Geotechnical Society 13(5): 89100 (in Korean).
Korea Highway Corporation. 1996. Korea specification of highway and bridges. (in Korean).
Samsung Corporation. 1998. Subsoil investigation of incheon LNG storage tank, Rpth to
Korea Gas Corporation. (in Korean).