Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Ng, C. W. W. & Yan, R. W. M. (1999). Geotechnique 49, No.

6, 825834

Three-dimensional modelling of a diaphragm wall construction


sequence
C . W. W. N G  a n d R . W. M . YA N 

typical dimensions of diaphragm wall panels constructed are 8 m long (L), 06 m wide (W ) and
16 m deep (D). The numerical simulation of the
diaphragm wall construction has been conducted
using a nite difference programme FLAC3D
(Itasca, 1996). The three-dimensional nite difference mesh and the notations for the stress state
adopted are given in Figs 1(a) and 1(b) respectively. The construction sequence of three diaphragm wall panels simulated is shown in Fig. 2.
One plane of symmetry, x 0, is used in generating a nite difference mesh for the analysis. The
planes, x 0 and x 50, are allowed to move
freely in the y and z directions but not in the x
direction. Similarly, the planes, y 50 and
y 50, are free to move in the x and z directions
but not in the y direction. At the boundary plane,
z 50, all movements are restrained.

KEYWORDS: deformation; diaphragm wall; Gault;


stress transfer; three-dimensional.

INTRODUCTION

It is well-known that construction of diaphragm


walls and bored piles will cause considerable stress
changes in heavily overconsolidated soils and can
induce substantial ground deformations (Burland &
Hancock, 1977; Symons & Carder, 1993). The
mechanisms of stress transfer and their inuence
on ground movements have been investigated by
many researchers in the eld (Tedd et al., 1984),
using various numerical methods and techniques
(Higgins et al., 1991; Ng, 1989; Gunn et al., 1993;
Ng et al., 1995) and a geotechnical centrifuge
(Powrie & Kantartzi, 1996).
Among the numerical methods and techniques
adopted and reported, Ng et al. (1995) carried out a
pseudo three-dimensional analysis of diaphragm
wall installation effects in an attempt to explain the
observed substantial stress reduction but small
ground deformation during the construction of a
diaphragm wall in Gault Clay at Lion Yard Cambridge. In this technical paper, results of a this threedimensional back-analysis of the diaphragm wall
installation sequence at Lion Yard are reported. The
computed results are compared with eld observations as well as some numerical calculations from
the previous pseudo three-dimensional analyses. Key
stress transfer mechanisms and ground deformations
from this truly three-dimensional analysis are identied, explained and conrmed.

Soil model and parameters


The ground succession at Lion Yard is about
3 m gravel/made ground overlying 38 m of Gault
Clay, which in turn overlies the Lower Greensand.
The initial ground water table is located at about
3 m below ground. The Gault Clay has been
modelled as a linear elasticperfectly plastic isotropic material with a MohrCoulomb yield surface. For consistency and ease of comparisons, the
soil model and the soil parameters adopted in this
paper are identical to those adopted by Ng et al.
(1995). Model parameters used are summarized in
Table 1. Justications of the adopted parameters
are discussed by Ng (1992) and Ng et al. (1995).
For comparisons between the current analysis
and the pseudo three-dimensional analysis (Ng et
al., 1995), identical modelling techniques have
been applied to account for the variation of stiffness with strain within a linear analysis. The stiffness for soils adjacent to any one of the panels
(see Fig. 3) which are being constructed, has been
assigned to a relatively low stiffness ratio,
Eu =cu 1000, to allow for the reduction of soil
stiffness at modest shear strains as a result of the
effects of the excavation process. This low stiffness
ratio has since remained unchanged. For all other
unaffected soils, a higher stiffness ratio, Eu =cu
4000, has been adopted. The adoption of the
simple soil models and the articial division of

NUMERICAL MODELLING TECHNIQUES AND


MODEL PARAMETERS

Finite difference mesh and boundary conditions


In order to compare with and verify the results
from the pseudo three-dimensional analysis of the
diaphragm wall construction at Lion Yard Cambridge (Ng et al., 1995), the same construction
sequence has been selected and analysed. The
Manuscript received 17 Nov 1998; revised manuscript
accepted 20 Aug 1999.
Discussion on this paper close 30 June 2000; for further
details see p. ii.
 Hong Kong University of Science & Technology.

825

826

NG AND YAN

x
100 m

50 m

y
Panel 2

Panel 1

Panel 3

50 m

z
(a)

x
zz

zy

zx

xz

yx
yy

xy
xx

yz

z
(b)

Fig. 1. Diagrams showing: (a) nite difference mesh; (b) three-dimensional stress state

827

DIAPHRAGM WALL CONSTRUCTION SEQUENCE MODELLING

x
Panel 2 Panel 1

0.3 half-panel
width

Axis of
symmetry

Panel 3

16 m

8m

z
Panel 1
x

Panel 2

Panel 3

y
Section Section
AA
BB

B2 B B1

Construction sequence
1B

1C

2B

2C

3B

3C

1st character: Panel No.


2nd character: Bentonite (B)
Concrete (C)

0.25 m
2

B B B

Fig. 2. Construction sequence of diaphragm walling


Table 1. Soil parameters adopted in the analysis
Gravel
Description
Unit weight (s )
Coefcient of earth pressure (K 0 )
Drained angle of friction (9)
Effective cohesion (c9)
Drained Poisson's ratio
Drained Young's modulus (E9)

Gault Clay
Data
s 18 kN=m3
035
408
0 kPa
02
50 000 kN=m2

two different zones of stiffness are clearly not very


desirable. This is because the continual variation of
soil stiffness with strain and the path dependency
of soil stiffness response are not included. Clearly,
further renement of the analysis is needed once
the fundamental three-dimensional stress transfer
mechanisms due to diaphragm wall installation
have been understood.
Modelling procedure
Figure 2 shows the construction sequence modelled in the nite difference analyses. The modelling procedures are summarized as follows.

Description
Unit weight
Coefcient of earth pressure
Undrained angle of friction
Undrained shear strength
Undrained Poisson's ratio
Undrained Young's modulus

Data
s 20 kN=m3
K 0 1:5 (assumed)
08
cu 60 10z9 kPa
0495
Eu =cu 4000 or 1000

(a) Excavate the rst trench (panel No. 1) in one


step and apply hydrostatic bentonite pressure
(unit weight 10:1 kN=m3 ) on the trench
faces simultaneously (i.e. the 1B stage).
(b) Cast the concrete panel by increasing the
lateral pressure inside the panel from the
bottom using a theoretical bilinear wet concrete pressure envelope proposed by Ng (1992)
and Lings et al. (1994).
(c) Cure the concrete panel No. 1 by replacing the
trench with elastic concrete elements and
removing the applied bilinear concrete pressures on the trench faces simultaneously (i.e.
stage 1C).

828

NG AND YAN

Wall
3

Gravel

Panel No. 1

4
16
2

Eu/cu 5 4000
M

Plan view, section MM

Elevation view

(b)

Soil with Eu/cu 5 1000

(a)

x
Eu/cu 5 4000

Panel No. 2

Panel No. 2

y
8

Eu/cu 5 4000
Panel No. 1

Panel No. 3

Panel No. 1
4

y
Eu/cu 5 4000

y
2

Plan view, section MM

Plan view, section MM

(c)

(d)

Fig. 3. Locations of soil elements with low stiffness: (a) overview; (b) during the construction of panel No. 1; (c)
during the construction of panel No. 2; (d) during the construction of panel No. 3 (all dimensions in metres)

(d ) Excavate panel No. 2 and apply bentonite


pressures inside the trench (stage 2B).
(e) Cast the concrete panel by increasing the
bentonite pressure to the bilinear concrete
pressure and cure the panel by replacing the
trench with elastic concrete elements and
removing the applied bilinear concrete pressures on the trench faces simultaneously (i.e.
stage 2C).
( f ) Apply the same modelling procedures to
construct panel No. 3.
Details of the modelling procedures are given by
Yan (1998).
STRESS DISTRIBUTIONS AND GROUND
DEFORMATIONS

As dened by Ng et al. (1995), the horizontal


arching mechanism refers to the redistribution of
total horizontal stress laterally via the shear stress
component ( yx ), as shown in Fig. 1(b). The downward load transfer mechanism denotes the redistribution of horizontal load vertically downward by
shearing ( zx ).

Stress distributions behind the diaphragm wall


Construction of diaphragm walls inevitably
causes stress changes in the ground. The computed
distributions of total normal horizontal stress (xx )
with depth, 035 m behind the wall, at sections A
A, B B and B B are shown in Figs 4(a),
4(b) and 4(c) respectively. During the construction
of panel No. 1, the normal horizontal stress at
section AA is governed by the bentonite and wet
concrete pressures inside the trench. Shear stresses,
yx and zx, are effectively zero at this section.
This results in a signicant reduction of total
normal horizontal stress above the toe of the wall
at the centre of panel No. 1. On the contrary,
below the toe of the wall there is an increase of
total normal horizontal stress as a result of stress
transfer via the downward load transfer mechanism
by the shear stress ( zx ) during the excavation
process from the top to the toe of the panel.
However, this increase becomes insignicant at
about half of the panel depth, below the toe of the
wall.
It can be seen from Fig. 4(a) that the stress
reductions above the toe do not equal the stress
increases below the toe of the wall. This implies

829

DIAPHRAGM WALL CONSTRUCTION SEQUENCE MODELLING

Stress: kPa
0

200

400

600

1B
1C
2B
2C
3B
3C

1B
1C
2B
2C
3B
3C

10

lK
tia
0

15

15

l
tia
Ini

rete
conc
Wet
onite
Bent

20

rete
conc
Wet
onite
Bent

Depth: m

Ini

Depth: m

K0

20

600

10

Stress: kPa
200
400

25
(b)

25
(a)
Stress: kPa
0

200

400

600

0
1B
1C
2B
2C
3B
3C

10

15

lK

tia

Ini

20

rete
conc
Wet
onite
Bent

Depth: m

25
(c)

Fig. 4. Total horizontal stress, xx , behind the wall at: (a) section AA; (b) section B2 B2 ; (c)
section B B

that some of the initial stresses are redistributed


laterally through the horizontal arching mechanism
via the shear stress component ( yx ), discussed in
detail by Ng & Yan (1998) who simulated the
installation effects of a single diaphragm wall
panel. Equal stress reduction above the toe and
stress increase below the toe of the wall has been
reported by Ng et al. (1995) in their pseudo threedimensional analyses. The difference in the computed behaviour between the true and pseudo
three-dimensional analyses is that the downward
load transfer and the horizontal arching mechan-

isms were uncoupled in the pseudo three-dimensional analyses. In their two-dimensional plane
strain analysis, which is one of the two-stage
analyses of the pseudo three-dimensional approach,
any reduction of total normal horizontal stress
above the toe will cause an equal amount of stress
increase below the toe of the wall. Thus, the
pseudo three-dimensional analysis overestimated
the stress reduction attributed to the downward
load transfer mechanism. On the other hand, the
horizontal arching mechanism transfers lateral
stress (via the shear stress, yx ) from the centre to

830

NG AND YAN

the edges of the panel, resulting in an increase in


normal stress at sections B B and B B (see
Figs 4(b) and 4(c)).
Subsequent construction of the two adjacent
panels, No. 2 and 3, has only a minor inuence on
the stress distribution with depth at section AA
(see Fig. 4(a)) because of the large panel length
(8 m) which limits the extent of the horizontal
arching action. Nevertheless, a small stress increase
can be seen above and below the toe of the wall
due to the construction of the two adjacent panels.
No signicant changes of stress distribution behind
panel No. 1 were computed by modelling two additional panels outside panels No. 2 and 3 and
the lateral stress distribution behind panels No. 2
and 3 moved towards that behind panel No. 1 (Yan,
1998). Due to the limited length of this paper, for
detailed computed results, refer to Yan (1998).
Although the construction of panel No. 1 does
not cause any signicant reduction of total normal
horizontal stress at section B B above the toe

of the wall (see Fig. 4(b)), the construction of


panel No. 3 leads to a large increase in total
normal horizontal stress at section B B along
the depth as a result of the horizontal arching
mechanism via the shear stress component, yx .
The increased normal stress is well above its initial
K 0 value. This is because after the construction of
panel No. 1, the stiff concrete panel provides a
strong support for the horizontal arching mechanism to take place, and hence results in a higher
horizontal stress than those values at section B
B (see Fig. 4(c)), where the normal horizontal
stress at this section approximately returns to its
initial K0 conditions in the ground. High local
horizontal stress at the edges of a panel might
affect the performance of the stop ends and waterproong system.
Figure 5 shows the computed three-dimensional
distributions of total normal horizontal stress (xx )
with depth, 035 m behind the wall, along the
length of the wall at the end of installation of the
Stress: kPa
0

:
l, y

he
gt

n
alo
0
ce
n
a
st 24
Di
28
212
216
220
0

wa

200 400 600

16

12

Depth:

10

15

20

Fig. 5. Total horizontal stress, xx , behind the wall at 3C stage

831

DIAPHRAGM WALL CONSTRUCTION SEQUENCE MODELLING

Surface settlements
Figures 7(a) and 7(b) show the vertical surface
settlement varying with perpendicular distance
away from the wall at section AA and at section
BB (an average of computed values at sections
B B and B B ) respectively. Due to the
larger stress reduction at the bentonite than at the
concreting stage, settlements at section AA that
occur at the former stage are approximately twice
those computed at the latter stage during the construction of the rst panel. After concreting, a
gentle settlement trough appears behind the rst
panel, with the maximum settlement occurring
approximately 3 m away. Settlements at section A
A continue to increase slightly during the subsequent construction of the two adjacent panels. The
computed settlement at a distance of about one
trench depth (1D) away is approximately 30% of
the maximum settlement which occurred behind
the wall at the end of construction (at 3C stage).
No signicant settlement can be seen at a normal
distance of approximately one-and-a-half times the
trench depth (1:5D) away from the wall. This is
consistent with the centrifuge results reported by
Powrie & Kantartzi (1996). Similar magnitude and
settlement proles can also be seen at section BB

Stress: kPa
0

200

400

600

800

1000

Measured

Pseudo 3D analysis
3D analysis (3C stage)

10
Depth: m

diaphragm wall panels. Due to the horizontal arching and the downward load transfer mechanisms, a
non-uniform horizontal stress distribution can be
clearly seen behind the wall. Stress reduction at the
centre portion of a panel leads to an increase in
stress at the edges and beyond the side panels (i.e.
panel No. 2 and No. 3). A similar non-uniform
stress distribution was obtained in the pseudo
three-dimensional analyses (Ng et al., 1995). The
degree of non-uniformity might be overestimated
due to numerical peculiarity resulting from the
substantial difference in stiffness between the soil
and concrete elements at the interfaces.
Comparisons between computed and measured
total horizontal stresses are shown in Fig. 6. The
measured values were obtained from seven total
earth pressure cells installed near the centre line of
the panel. For the horizontal stresses above the toe
of the wall, computed values by the truly threedimensional analysis are more consistent with the
eld measurements than those calculated by the
pseudo three-dimensional analysis. Below the toe,
the truly three-dimensional analysis computed a
substantially smaller stress increase than the pseudo three-dimensional analysis. The area of stress
reduction above the toe is clearly not equal to the
area of stress increase beneath the toe. This implies
that stress reduction in the ground is attributed to
both downward load transfer and horizontal arching
mechanisms, which can only be modelled simultaneously by the truly three-dimensional analysis.

K0 5 1.5

15
Level
of toe

20

25

Fig. 6. Predicted and measured total horizontal stresses near the centre line of panel No. 1 after
diaphragm wall construction

(see Fig. 7(b)) at the end of installation. When


comparing these with the eld measurements of
ground settlement of approximately 1 mm occurring between 8 m and 20 m from the diaphragm
wall at Lion Yard (Ng, 1992), there is reasonable
agreement between the computed and measured
values.
Figure 8 shows a three-dimensional settlement
surface behind the wall at the end of construction.
A settlement bowl occurs behind each panel. The
magnitude of the maximum settlement behind the
two subsequent secondary panels (i.e. No. 2 and 3)
is approximately twice that behind the rst (primary) panel. This is due to the accumulation of
settlements from the construction of the rst panel
and the subsequent panel. If the soil stiffness were
permitted to vary with strains, the difference in
settlement behind the primary and the secondary
panel would be ever larger. It should be noted in
the gure that there are some unusual abrupt
changes of deformed shapes at the two edges of
the rst panel. Local numerical peculiarity may
lead to these abnormal results since there was a
substantial difference in stiffness between the rst
constructed concrete panel and the surrounding
soils when the two adjacent panels were being

832

NG AND YAN

Distance away from the wall, x : m


0

10

15

20

25

30

Vertical settlement, z: mm

1B

1C
2B
2C
2

3B
3C

3
(a)
Distance away from the wall, x : m
0

10

15

20

25

30

Vertical settlement, z: mm

1B
1C

2B
2C
3B
2

3C

3
(b)

Fig. 7. Variations of vertical settlement at ground surface with distance normal to the wall at: (a) section AA; (b)
section BB

excavated. However, by taking an average of the


computed values from two sides of the junction (at
section BB) between two panels, a smooth curve
can be obtained (see Fig. 7(b)).
CONCLUSIONS

A truly three-dimensional back-analysis of the


construction sequence of three diaphragm wall
panels installed in stiff Gault Clay has been conducted using a nite difference programme. The
current analysis conrms the roles of two stress
transfer mechanisms: the horizontal arching and
downward load transfer mechanisms during diaphragm wall installation, which were initially identied by Ng et al. (1995) from their pseudo threedimensional analysis. The horizontal arching
mechanism plays a key role in redistributing total
normal horizontal stresses laterally from the centre
to the two edges of, and beyond, a panel. The

redistribution of horizontal load is achieved via the


shear stress component ( yx ). This results in nonuniform horizontal stress distribution behind a diaphragm wall panel, the stress being the smallest at
the centre but increasing in magnitude towards the
edges. Partial recovery of total normal horizontal
stresses occurs during the construction of adjacent
panels. On the other hand, some of the initial total
horizontal normal stresses are shed vertically
downwards beneath the toe of the wall by the
downward load transfer mechanism via the shear
stress, zx . This leads to an increase in horizontal
stress below the toe of the wall. These two mechanisms act simultaneously and result in an average reduction of horizontal stress directly behind
the wall above the toe but an increase of horizontal
stress in neighbouring soil beyond the wall in the
longitudinal direction and below the toe of the
wall.
The calculated horizontal stress increase beneath

833

DIAPHRAGM WALL CONSTRUCTION SEQUENCE MODELLING

Settlement: mm

2
Panel No. 2
Panel No. 1
3

Panel No. 3
216

212

30
25

28
24
0
Distanc
e along
th

4
e wall,

8
y: m

20
m
15
ll, x :
e wa
h
t
m
o
r
f
y

10
12

16

5
20 0

wa

nce a

Dista

Fig. 8. Vertical settlement, z , behind the wall at 3C stage

the toe by the present analysis is noticeably smaller


than the value computed by the pseudo threedimensional analysis. The difference is attributed
to the fact that the pseudo three-dimensional analysis overestimated the amount of downward load
transfer as a result of its two uncoupled plane
analyses.
A settlement bowl appears behind each panel,
while the maximum settlement occurs at a distance
of approximately 20% of the trench depth (i.e.
about 3 m) away from the face of the wall. No
signicant settlement can be seen at a normal
distance of approximately one-and-a-half times the
trench depth away from the wall. The maximum
computed settlements behind the two subsequent
secondary panels (i.e. No. 2 and 3) are approximately twice that behind the rst (primary) panel.

NOTATION
c9 effective cohesion
cu undrained shear strength of Gault Clay
D depth of diaphragm wall panel
E9 effective Young's modulus of gravel
Eu undrained Young's modulus of Gault Clay
K 0 in situ earth pressure coefcient
L length of diaphragm wall panel
W width of diaphragm wall panel
z9 depth below the top of Gault Clay
s unit weight of soil
z ground surface settlement
normal stress
shear stress
9 angle of friction

REFERENCES
Burland, J. B. & Hancock, R. J. R. (1977). Underground
car park at the House of Commons, London: geotechnical aspects. Structural Engineer 55, No. 2, 87100.
Gunn, M. J., Satkunananthan, A. & Clayton, C. R. I.
(1993). Finite element modelling of installation effects. In Retaining structures pp. 4655. London:
Thomas Telford.
Higgins, K. G., Potts, D. M. & Symons, I. F. (1989).
Comparison of predicted and measured performance
of the retaining walls of the Bell Common tunnel.
TRRL contractor report 124.
Itasca (1996). Fast Lagrangian analysis of continua
(FLAC-3D), version 1.1, user manuals. Itasa Consulting Group, Inc., Minnesota, USA.
Lings, M. L., Ng, C. W. W. & Nash, D. F. T. (1994). The
lateral pressure of wet concrete in diaphragm wall
panels cast under bentonite. Proc. Instn. Civ. Engrs
Geotech. Engng 107, July, 163172.
Ng, C. W. W. (1992). An evaluation of soilstructure
interaction associated with a multi-propped excavation. PhD thesis, University of Bristol, ch. 7.
Ng, C. W. W., Lings, M. L., Simpson, B. & Nash, D. F.
T. (1995). An approximate analysis of the three
dimensional effects of diaphragm wall installation.
Geotechnique, 45, No. 3, 497507.
Ng, C. W. W. & Yan, R. W. M. (1998). Stress transfer
and deformation mechanisms around a diaphragm
wall panel. Journal of Geotechnical and Environmental Engineering, ASCE 124, No. 7, 638648.
Powrie, W. & Kantartzi, C. (1996). Ground response
during diaphragm wall installation in clay: centrifuge model tests. Geotechnique 46, No. 4, 725
739.
Symons, I. F. & Carder, D. R. (1993). Stress changes in
stiff clay caused by the installation of embedded
retaining walls. In Retaining structures, pp. 227236.
London: Thomas Telford.

834

NG AND YAN

Tedd, P., Chard, B. M., Charles, J. A. & Symons, I. F.


(1984). Behaviour of a propped embedded retaining
wall in stiff clay at Bell Common Tunnel. Geotechnique 34, No. 4, 513532.

Yan, R. W. M. (1998). Three-dimensional modelling of


diaphragm wall installation. MPhil thesis, Department
of Civil Engineering, Hong Kong University of
Science and Technology.

You might also like