Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Geotechnical Engineering Proceedings of the Institution of Civil Engineers

Volume 164 Issue GE1 Geotechnical Engineering 164 February 2011 Issue GE1
Pages 11–25 doi: 10.1680/geng.9.00078
Mass behaviour of embedded improved Paper 900078
soil raft in an excavation Received 26/09/2009 Accepted 17/02/2010
Published online 06/01/2011
Yang, Tan and Leung Keywords: excavation/grouting/mathematical modelling

ICE Publishing: All rights reserved

Mass behaviour of embedded


improved soil raft in an
excavation
j
1 Haibo Yang BEng, PhD j
3 Chun Fai Leung PhD
Mott MacDonald Singapore Pte Ltd, Singapore Professor, Centre for Soft Ground Engineering, Department of Civil
j
2 Thiam Soon Tan PhD Engineering, National University of Singapore
Professor, Centre for Soft Ground Engineering, Department of Civil
Engineering, National University of Singapore

j
1 j
2 j
3

In deep excavation in soft soils, a layer of soil below formation level is often improved to stabilise the excavation.
The improvement, usually by deep cement mixing or jet-grout piling, will result in the construction of a raft of
overlapping short columns (termed an ‘embedded improved soil raft’). Soil investigation carried out will usually
provide information on vertical cores from such columns. But during excavation, the columns are loaded laterally by
the inward-moving retaining wall: thus the mobilised mass properties in the lateral direction are of importance in the
design. This paper examines the mechanisms of how mass properties are mobilised and their relation to the
elemental properties, the variation of properties within a column, and how the columns are arranged. The analysis
and the simulation of a reported field case history show that for soil-cement columns arranged just in contact with
each other, the mobilised mass stiffness is less than 28% of the elemental material stiffness. The analysis also shows
that the properties of the outer layer have a greater impact on the mass behaviour of the embedded improved soil
raft than those of the inner layer, and this ought to be considered in any site investigations.

Notation strutting cannot be used. To control the wall deflection and its
c cohesion yield stress associated ground movements effectively, one common approach
E Young’s modulus is to improve a layer of soft soils beneath the formation level. A
L overlap parameter common approach to improve the soft soil, as shown in Figure
n total number of samples 1(a), is to mix the in situ soft soils with cement, which can be
R column radius introduced into the ground either by the deep cement mixing
V coefficient of variance (DCM) method or the jet-grouted piling (JGP) method (Porbaha
 Poisson’s ratio et al., 1999; Shibazaki, 2003). Both methods produce short soil-
 normal stress cement columns that overlap with each other to form a continu-
f direct shear strength ously improved composite ground, which acts like a strut below
 friction angle the formation level to help restrain the inward movement of the
ł dilation angle retaining wall. In this paper, this composite ground is termed the
embedded improved soil raft to emphasise that the improvement
1. Introduction is below the formation level, and is over a wide area rather than
Deep excavations in soft soils in urban areas often face serious in strips like a strut. Applications of such an embedded improved
challenges in the control of induced ground movement on nearby soil raft system to support deep excavation have been reported by
structures to limit its adverse impact (Tan and Shirlaw, 2000). many researchers (Hsi and Yu, 2005; O’Rourke and McGinn,
Typically, such deep excavations are supported by diaphragm 2006; O’Rourke and O’Donnell, 1997; Wong and Poh, 2000).
walls or sheet-pile walls. The imbalance caused by the removal of
soils will cause wall deflection, often with the maximum deflec- In construction, to monitor the quality of the embedded improved
tion occurring below the formation level, where conventional soil raft once constructed, vertically cored samples from indivi-

11
Geotechnical Engineering Mass behaviour of embedded improved
Volume 164 Issue GE1 soil raft in an excavation
Yang, Tan and Leung

sive strength and the Young’s modulus for design use. However,
Excavation in reality, when the improved soil raft is called upon to restrain
the retaining wall from moving in, the resistance comes from the
composite strength and stiffness in the horizontal direction, which
are affected by the imperfections and heterogeneity in the
horizontal direction, as well as by how the columns are arranged
and overlapped with each other. The mobilised properties of such
an embedded improved soil raft, used to stabilise an excavation,
is not going to be dependent on the elemental properties alone.
Once it is realised that the restraint comes from the composite
strength and stiffness in the horizontal direction, the non-
Soil raft uniformities in the horizontal direction, even within the same
improved soil column, become an important factor in the
estimation of the composite properties. This aspect, which to date
has seldom been discussed, is the focus of the present paper.
Contact critical
2. Variability in improved soil-cement
columns
Figure 2 shows the distribution of direct shear strengths of
samples cored from deep mixing columns at three locations at a
(a) site in Tokyo (Kawasaki et al., 1984). The direct shear strengths
were replotted, and the strength distribution along the column
radial direction is shown in Figure 3. It can be seen that,
generally, the outer layers have lower strength than the inner
layers. When these columns are arranged and subjected to loads
Embankment
that act on them horizontally, these variations will have an
Fill
important impact on the mobilised properties. In another site in
Singapore, a hybrid system called RAS-JET, which has jet open-
ings at the edges of mechanical mixing blades, is used to form
soil-cement columns in the ground. The soil-cement columns
Soil formed by the RAS-JET system have two distinct, concentric
layers in the radial direction: the inner layer formed by mechani-
Potential slip line cal mixing, and the outer one formed by jet grouting. The outer
jet-grouting layer is often stiffer than the inner mechanical mixing
layer. A typical unconfined compressive strength (UCS) profile of
Soil-cement column the RAS-JET soil-cement column is shown in Figure 4.

With different arrangement of columns, and with variations of


properties such as those described above, the mobilised properties
Composite strength: of the improved soil layer when acting like a strut will be
summation of strength of
affected. A back-analysis on an excavation in Singapore (Pickles
column and soil
and Henderson, 2005) showed that the mobilised mass behaviour
was very different from the results obtained from the cored
(b) samples: the back-analysed effective mobilised mass stiffness was
about 70 MPa, whereas the stiffness from tests on cored samples
Figure 1. Mobilised mechanics of soil-cement treated ground:
was well over 200 MPa. In addition to the usual concern about
(a) excavation loading; (b) embankment loading
the effect of sample disturbance, another major contributor to
account for this very significant (more than 300%) difference is
the fact that the actual mobilised mass properties are dependent
dual soil-cement columns are obtained from the field and tested on the way the columns are arranged, and on how the properties
in the laboratory. The elemental properties (i.e. strength and within each column vary.
stiffness) obtained from these tests, representing the quality of
individual soil-cement columns, are then used for design pur- In current design practice, one approach to estimate the mobilised
poses. Therein lies a key problem: samples are cored vertically, mass properties of such an improved soil raft is to use a weighted
and are usually also tested vertically to obtain both the compres- average method, which sums the properties of each constituent,

12
Geotechnical Engineering Mass behaviour of embedded improved
Volume 164 Issue GE1 soil raft in an excavation
Yang, Tan and Leung

25 11·4 10·3 8·9 7·7


4·7
A-4·2 10·2 15·7 12·9 13·4
5·2
11·6
n ⫽ 108 10·2
7·4 14·9 11·3
20 τf ⫽ 998 kN/m2 10·1 12·2 15·7 12·1 6·8 15·7 11·0 11·8
12·9 9·6 12·1
σ ⫽ 270 kN/m2
Number of samples

9·0 12·2 10·2 14·8 12·2 9·9 12·6 14·1


V ⫽ 0·270 12·4
11·7 13·9 14·3
14·9 10·7 11·1 10·0
15 5·4 10·9 9·2
B.H.
B.H. 5·3 B.H. 11·1
B.H. B.H. 8·4 B.H.
8·2
B.H. 8·5 4·3
B.H. B.H.
10 B.H.
B.H.
B.H. 7·8
B.H. 9·9 B.H. 9·3 9·6
6·3 9·0
9·6 9·7 7·6 13·0 12·1 13·3 11·4 9·2
5 10·0 9·3 11·9 7·4
8·2 11·0 8·0 11·2 11·4 12·4 6·74 8·4
9·4 12·3 10·2 8·3
8·6 9·9 9·3 9·7 11·1 11·4 10·3 12·4
0 7·9 8·4 6·7
9·0
13·5 8·2
0 500 1000 1500 2000 8·0
7·8
11·3
9·2 8·3
Direct shear strength, τf: kN/m2 7·5 10·0 8·8
8·3
7·1 4·6
6·2

(a)

9·1 9·9
25 8·7 8·3 10·6 9·9 11·1
A-13·2
10·3 14·9 14·0
n ⫽ 124 7·8 12·0 15·4 10·2 10·7
15·3
19·2
15·7
τf ⫽ 1262 kN/m2 10·3 12·6 9·8 8·3 11·9 14·9 12·9
11·5
20 11·3
σ ⫽ 279 kN/m2 12·6 14·9 11·4 11·5
Number of samples

13·5 11·0 8·1 10·3 11·1 11·8


V ⫽ 0·222 16·9 14·0
12·1 16·0 11·7
14·7 13·6 10·7 13·9 7·1
15 12·4 17·4 24·3 10·7
10·4 11·9 B.H.
12·7 12·1 15·4 9·5 13·0
B.H.
B.H.
10·1 11·0 14·2 11·9
10 13·9 13·2 14·8
B.H. 12·0
14·7 15·0
10·8 11·7 13·8 13·3 11·1
B.H. 14·8
10·0 12·3 12·1 21·6 17·5 11·6 11·4
5 13·8 10·5 9·4 10·4
11·4 11·5 13·5 18·6 23·1 15·4 13·0 10·8
9·4 13·7 12·0 13·0
14·3 10·3 13·6 14·1 16·6 11·8 12·1 12·4
0 13·2 15·6
13·0 10·9
0 500 1000 1500 2000 2500 12·2 15·1
12·1
11·3
13·1
2
8·1 9·3 13·3(c)
Direct shear strength, τf: kN/m 12·8 11·2 13·7 13·9 11·7
13·3 10·5
(b)

10·5 9·6
8·0
13·1 12·4 2·9
A-14·2 11·7 14·4 5·8
11·7 9·1
15·8 12·6
20 n ⫽ 108 11·3 14·7 15·4 16·4 13·0 8·2
τf ⫽ 1267 kN/m2 16·3 12·6 13·6 11·0
10·9 10·3 16·6 14·1 16·0
σ ⫽ 396 kN/m2 14·7 9·1
Number of samples

13·1 17·2 14·8 12·3


15 V ⫽ 0·313 10·3 13·3 16·2
13·6 12·6 15·0 11·7
13·5 17·0 16·6 2·2
10·0 12·0 15·9
B.H. B.H.
10 5·7 5·5
19·4 B.H. 15·9
B.H. 20·3 7·9
15·3 15·6 18·1 17·5
13·3
5 11·4 13·2 13·4 10·8 19·0 19·8 19·7 16·1
9·3 4·5 10·7 11·5
17·0 10·4 9·8 19·3 21·1 8·1 17·3 10·5
15·5 15·4 10·2
0 11·1 14·2 10·1 8·1 11·6 15·0 5·6 8·9
0 500 1000 1500 2000 2500 13·9
10·9 8·3
8·6 10·5
Direct shear strength, τf: kN/m2 13·9 11·4 3·5 13·6
12·1 6·9 14·9 19·3 15·8
7·5 16·3

(c) (τf ⫻102 kN/m2)

Figure 2. Horizontal variability of deep mixing columns (Porbaha,


2002)

13
Geotechnical Engineering Mass behaviour of embedded improved
Volume 164 Issue GE1 soil raft in an excavation
Yang, Tan and Leung

2·0
line cuts through the improved ground, such as from an embank-
ment loading, as shown in Figure 1(b). However, the mobilised
mechanics of an embedded improved soil raft during an excava-
tion (Figure 1(a)) are very different. The improved soil raft is
1·5
compressed horizontally as the excavation proceeds; it is also
Normalised strength

subject to basal heave as the overburden is gradually removed


(Shirlaw, 2003; Tanaka, 1993). The contact between the soil-
cement columns and the properties near the contact area are
1·0
therefore critical for the composite mass to resist the excavation
loading. If one soil-cement column is missing as a result of
complicated site conditions, it will have little impact on the
0·5 overall behaviour in the embankment loading case, whereas the
effect could be much more drastic for the excavation loading
case, since the horizontal load transfer chain could be broken; in
using the improved soil layer like a strut, the weakest unit in the
0 chain can affect the overall performance.
0 0·2 0·4 0·6 0·8 1·0
Sample locations: 0, column centre; 1, column edge To illustrate this point graphically, consider two cases of two
identical columns in point contact and being pressed into each
Figure 3. Strength variation in the radial direction of the soil-
other laterally, as shown in Figure 5. Each column is divided into
cement columns, normalised to the average strength of the centre
two concentric layers that have the same cross-sectional area. In
samples (after Kawasaki et al. (1984))
the first case the outer layer is soft while the inner layer is very
stiff, whereas for the second case the outer layer is very stiff and
the inner layer is soft. When mobilised to resist the lateral
5 loading, the composite behaviour of the two columns together
would be different. In the first case, with the soft outer layers
touching each other, the mobilised mass behaviour will be closer
Unconfined compressive strength: MPa

4 to that of two soft columns being pressed against each other,


whereas in the second case the behaviour is more akin to that of

3
Well-mixed soil

Poorly mixed soil

Mechanical mixing part Jet grouting part

0
0 0·35 0·70 1·05 1·40
(a)
Distance from column centre: m
Outer layer area ⫽ Inner layer area
Figure 4. Typical unconfined compressive strength profile of a
soil-cement column formed by RAS-JET at Nicoll Highway field
trial, Singapore

weighted by its respective volume ratio within the composite


(Porbaha, 2000). Sometimes an additional parameter is introduced
into the method to refine its prediction of mass properties (Ou et (b)
al., 1996). The underlying philosophy of this method is that the
properties for each constituent are mobilised concurrently, and Figure 5. Analogy: (a) poorly mixed soil as outer layer; (b) well-
therefore the weighted average summation gives the best estima- mixed soil as outer layer
tion. This philosophy works for the case where a potential slip

14
Geotechnical Engineering Mass behaviour of embedded improved
Volume 164 Issue GE1 soil raft in an excavation
Yang, Tan and Leung

two very stiff columns being pressed together. On the other hand, Soil-cement column
if a weighted average method based on area is used, there would
be no difference between the two cases. This simple example
demonstrates that if such an improved soil is used like a strut in
an excavation to restrain the inward movement of the retaining
wall, then an understanding of how an array of such columns In situ soil
behaves collectively when loaded laterally is important.

Thus this study is concerned with how to interpret the mobilised


mass properties, given some information about the elemental
properties, the way the properties could vary, and how they are
arranged. Through this, an insight will be gained into how such
an improved soil raft consisting of overlapping columns will (a) (b)
behave when mobilised to restrain a retaining wall from moving
inwards during an excavation. Some guidelines will also be
recommended for engineers to consider in the design of such an
embedded improved soil raft to stabilise an excavation. This
paper will not deal with the issue of how the elemental properties
are obtained, as that is an issue of characterisation, a topic that
has been dealt with in great detail by other researchers.
L
3. Study approach R
L
The behaviour of an embedded improved soil raft in an excavation
is complicated. To make the problem amenable, systems with
simple geometric, material and loading configurations will be
examined first; then complexities will gradually be introduced into
the model in order to build up the necessary understanding of this (c) (d)
complicated problem in a systematic manner. Figure 6 shows the
logic of the analyses to be discussed here. The mass behaviour of Figure 6. Geometric configurations of studies conducted: (a) point
one row of columns in point contact and being compressed contact; (b) point contact grid; (c) overlapping; (d) equilateral
laterally is examined first. This is in fact similar to the Hertz triangle grid
contact problem involving two columns in contact, and to which
an analytical solution exists if the column is uniform and its
material model is isotropic and linearly elastic (Johnson, 1985). chosen as they are the simplest constitutive models, as the main
The Hertz contact problem serves both as a calibration of the focus of this study is on the mechanics of mobilisation and not
numerical models and as a starting step for understanding how the the modelling of material. The first is an isotropic linearly elastic
elemental properties change when an array of columns is being model, which is the simplest in a numerical analysis. The elastic
mobilised as a group. A reported case history (Nakagawa et al., model is able to characterise the initial response of the composite
1996), where the soil-cement columns were arranged in a point ground when loaded. As the load increases, a failure criterion is
contact grid, is then analysed. The simulation of the case history needed to capture the yielding of the individual column. This
will also take into account the contribution of the confining effects necessitates the use of a second material model, which is the
of the in situ soils and the adjacent soil-cement columns on the elastic perfectly plastic model using the Mohr–Coulomb failure
mobilised mass properties. After that, the columns in the row are criterion, which is commonly used in practice to model such
allowed to overlap with each other (Figure 6(c)), to study the improved soil layers (Hsi and Yu, 2005). Figure 7 shows the
effects of overlapping on the mobilised mass behaviour. Finally, stress, strain and stiffness relationships for the two elemental
multiple overlapping columns arranged in an equilateral triangular material models. The magnitude of the stiffness is set to
grid are studied. Taking advantage of symmetry in loading and 300 MPa, a typical value for soil-cement samples from the field.
boundary conditions, only the shaded area in each geometric The unconfined compressive yield stress of the elastic perfectly
configuration, which is sufficient to represent the behaviour of the plastic material model is assumed to be 3000 kPa. This would
whole row of columns, is modelled in the numerical simulations. give a failure strain of 1%, which falls into the reported range
Simulation of the Hertz contact problem and all the subsequent from field samples (Wen, 2005).
problems are completed using the general-purpose finite-element
package Abaqus/Standard (HKS, 2003). 4. Hertz contact problem
The finite-element mesh for the Hertz contact problem is shown
Two elemental material models are considered. These models are in Figure 8. The left boundary was fixed horizontally but allowed

15
Geotechnical Engineering Mass behaviour of embedded improved
Volume 164 Issue GE1 soil raft in an excavation
Yang, Tan and Leung

6000
Linear elastic

Elastic perfectly plastic


5000

4000
Stress: kPa

3000

2000

1000

0 Figure 8. Mesh, boundary and loading conditions for Hertz


0 0·5 1·0 1·5 2·0
Strain: % contact problem

500
Figure 9. The normal stress is normalised by its value at the
centre of the contact area, and the distance from the contact
centre is normalised by the length of the contact path. The
400 normal stress is at a maximum at the centre of the contact area,
and drops to zero at the edge of the contact area. The analytical
solution (Johnson, 1985) to the Hertz contact problem is also
300 shown in this figure, and the near-perfect agreement between the
Stiffness: MPa

analytical and numerical solution indicates the ability of the


numerical model to simulate this contact problem.

200 The mass strain is defined as the displacement at the top


boundary divided by the height of the mesh, and the mass stress
is taken as the vertical force in the rigid body divided by the
width of the mesh (Figure 8). The mobilised mass stiffness is
100
then deduced from the mass stress–strain curves. The mobilised
mass stiffness of the two columns subject to lateral compression
is normalised by the elemental stiffness of the column, and
0 plotted in Figure 10. It is only 22% of the elemental stiffness
0 0·5 1·0 1·5 2·0
1·0
Normalised normal stress along

Strain: %

Figure 7. Elemental material models: linear elastic and elastic 0·8


contact patch

perfectly plastic
0·6

0·4
to move vertically. Compressive loading from the neighbouring Analytical solutions
column was modelled by prescribing uniform displacements at 0·2 Numerical results
the top boundary to press the column against a rigid body (HKS,
2003). As noted before, the material model used is an isotropic 0
0 0·2 0·4 0·6 0·8 1·0
linearly elastic model. At the start of the analysis, the bottom was
Normalised distance from contact centre
in point contact with the rigid base, which was fixed in all
degrees of freedom. Intuitively, we expect that as the load is Figure 9. Solutions to Hertz contact problem: numerical
applied, the area of contact will grow from a point to a patch. compared with analytical
The normal stress distribution over the contact area is shown in

16
Geotechnical Engineering Mass behaviour of embedded improved
Volume 164 Issue GE1 soil raft in an excavation
Yang, Tan and Leung

0·40
when the columns are mobilised as a whole to resist the lateral
0·35 loading, the properties of the outer layers are more important. To
Normalised mass stiffness

make this observation clear, analyses were conducted to simulate


0·30 the scenarios as shown in the analogy in Figure 5. In the analyses,
the soil-cement columns are divided into two concentric layers
0·25 with the same cross-sectional areas. In one run, the outer layer
was assigned a stiffer material while the inner layer was assigned
0·20 a less stiff material. Both materials are assumed to be linearly
elastic perfectly plastic with the Mohr–Coulomb failure criterion
0·15
(Table 1). The material properties for the two layers were
0·10
swapped in the next run. The results of the mass stress–strain
0 0·5 1·0 1·5 2·0 and stiffness relationships are shown in Figure 12. It can be seen
Mass strain: % that when the outer layer was assigned a stiffer material than the
inner, the mass response was stiffer than in the first case. Given
Figure 10. Normalised mass stiffness of two columns in Hertz
that the cross-sectional areas for the two layers are the same, this
contact
set of results clearly shows that the outer layer is more important
than the inner layer in evaluating the mass responses of an
embedded improved soil raft for use in an excavation.
upon initial loading, increasing to about 29% when the mass
strain reaches 1%. This points to the need to differentiate the This example also points to another important observation,
properties of the individual columns from those of the columns namely that the arrangement has a significant impact on the
acting as a composite to withstand lateral compression. mobilised properties. For the case where the outer layer was
assigned to be stiffer, the mass stiffness is about 12% of the
Figure 11 shows the shear stress (von Mises) contours of the material stiffness of the stiffer layer. This ratio is about 6.7% for
column. High stress concentration was observed in the region that the case where the inner layer is stiffer. More quantitative
is close to the contact path. If the cylinders are divided into transformations of the elemental material stiffness to the mass
several concentric layers from centre to edge, the outer layers stiffness for different geometric and layering configurations will
would bear higher stresses than the inner ones. In other words, be shown in later parts.

S, von Mises stress


(ave. crit.: 75%)
⫹1·129 ⫻ 107
⫹1·035 ⫻ 107
⫹9·405 ⫻ 106
⫹8·464 ⫻ 106
⫹7·524 ⫻ 106
⫹6·583 ⫻ 106
⫹5·643 ⫻ 106
⫹4·703 ⫻ 106
⫹3·762 ⫻ 106
⫹2·822 ⫻ 106
⫹1·881 ⫻ 106
⫹9·410 ⫻ 105
⫹5·829 ⫻ 102

Figure 11. Von Mises stress contours for Hertz contact problem

17
Geotechnical Engineering Mass behaviour of embedded improved
Volume 164 Issue GE1 soil raft in an excavation
Yang, Tan and Leung

500
Material parameter Stiffer Less stiff
layer layer

Young’s modulus, E: MPa 300 75 400


Poisson’s ratio,  0.495 0.495 Outer layer stiffer
Cohesion yield stress, c: kPa 1500 375
Friction angle, : degrees 0 0

Mass stress: kPa


Dilation angle, ł: degrees 0 0 300

Table 1. Elemental material properties for the layered columns

200

5. Mobilised mass properties of improved


Inner layer stiffer
soil layer
100
5.1 One row of columns in point contact
Figure 13 shows the mass stress, strain and stiffness relationships
of one row of columns in point contact being compressed
laterally; this is essentially an extension of the Hertz contact 0
0 0·5 1·0 1·5 2·0
problem. In this analysis, it is reiterated that the model used to Mass strain: %
model each element is the isotropic linearly elastic model.
However, this figure shows that when the row of columns is
50
mobilised as a whole upon lateral loading, the mobilised mass
stress–strain behaviour is non-linear. A simple conclusion is that
in practice, even if the test shows that the properties of cored
samples are linear, the properties that are actually mobilised to 40
provide the lateral restraint are likely to be non-linear.
Outer layer stiffer

In Figure 13, when the elemental property is assumed to be linear


Mass stiffness: MPa

elastic, the mobilised mass stress–strain curve is concave up- 30


wards (solid line in the figure), suggesting that the overall
behaviour is strain-hardening: that is, the mass stiffness increases
as the columns are gradually loaded. Quantitatively, the mass
20
stiffness for the isotropic linearly elastic case upon initial loading
is about 22% of the elemental material stiffness, and can increase
up to 34% at a mass strain of 2%. The reason for this stiffening Inner layer stiffer
behaviour is the change in contact area as the load, and thus the 10
strain, is increased. At the start of the analysis, the columns are
only in point contact. However, as the load is increased, the area
of the mobilised zone gradually increases. This increase in the
area of contact leads to a stiffer mass response. Since no failure 0
0 0·5 1·0 1·5 2·0
criterion was prescribed in an isotropic linear elastic model, there Mass strain: %
is no limit to the shear stress increase within each column. The
mass stiffness can therefore continue to increase, albeit at a Figure 12. Mass behaviour of columns as arranged in the analogy
reduced rate.

In a more realistic simulation, a failure criterion ought to be behaviour of the dashed line in Figure 13. So the mass stiffness is
introduced. In this case, the Mohr–Coulomb failure criterion is about 23% of the elemental material stiffness when it starts to
coupled with the elastic model to simulate this failure behaviour. drop as the system is further loaded.
Once the failure criterion is introduced into the elemental
material model, part of each column will undergo yielding once 5.2 Simulation of a reported case history
the shear stresses within the columns reach the elemental material In a deep excavation in Tokyo (Nakagawa et al., 1996), soil-
strength. These yielded parts, if sufficiently large, dominate the cement columns are arranged just in contact with each other to
mass behaviour, and will drastically reduce the mass stiffness form an improved soil layer to stabilise the excavation. The
even though a bigger area has been mobilised, as shown by the authors found that the improved soil layer was much less effective

18
Geotechnical Engineering Mass behaviour of embedded improved
Volume 164 Issue GE1 soil raft in an excavation
Yang, Tan and Leung

0·006 Kh: designed K value


Linear elastic K⬘h: analysed K value at each stage
1·00 2·0
H: thickness of improvement
Elastic perfectly plastic δ: displacement of improved range
0·005 DB: final depth of excavation
D: depth of excavation at each stage
Normalised to the 0·75 1·5
elemental material
Normalised mass stress

0·004 stiffness
H/δ

K⬘h/K h

H/δ
0·50 1·0
0·003
K⬘h/K h

0·002 0·25 0·5

0·001
0 0
0 0·25 0·50 0·75 1·00
D /DB
0
0 0·5 1·0 1·5 2·0
Figure 14. Correlation of depth of excavation with displacement,
Mass strain: %
and K value (reprinted with permission of Taylor & Francis from
0·40 Nakagawa et al., 1996)

0·32 weighted average method but excludes the contribution from in


situ soils. The designed value thus obtained represents the
elemental material properties of the soil-cement columns.
Normalised mass stiffness

0·24 To quantitatively assess the impact of geometry arrangement on


the differences between the mobilised mass properties and the
elemental properties, numerical analysis of soil-cement columns
arranged as in the reported case history, namely just in contact in
0·16
a grid, is carried out using Abaqus. Only compressive loads from
the inward-moving retaining walls are considered. Thus the
symmetries in loads and geometry require that only a quarter of a
0·08 soil-cement column and its surrounding soils need be modelled.
However, such an arrangement, where columns are just in contact
and are surrounded by soft in situ soils, poses numerical
difficulties in two ways. First, it is observed in Abaqus that it is
0
0 0·5 1·0 1·5 2·0 very difficult to generate mesh for the soils in the model with
Mass strain: % elements of reasonably good shape at the sharp corners near the
contact point of two soil-cement columns. Even if a mesh is
Figure 13. Mass behaviour of one row of columns in point obtained, severely distorted elements are inevitable at the sharp
contact corners, where unfortunately the in situ soils undergo severe
deformations and thus require high-quality mesh to capture the
behaviour properly. Second, when soft soils are in between two
than assumed in the design in controlling the deflections of the stiff columns, some nodes in one column will penetrate into its
retaining walls. As shown in the solid line in Figure 14, their neighbouring column upon lateral loading, missing very impor-
back-analysed coefficient of horizontal subgrade reaction from tant interactions between the two adjacent soil-cement columns.
the field (Nakagawa et al., 1996), which essentially reflects the The results thus obtained would not be able to account properly
mobilised mass properties in the lateral direction, was between for the transfer of lateral loads from one column to the other. To
225 and 1000 tf/m (2205 and 9800 kN/m), only 3.0–13.3% of the overcome this, a special treatment (shown in Figure 15) is
design value of 7497 tf/m (73 470 kN/m). The design value was employed by removing a very small amount of the in situ soils at
obtained by ‘multiplying improvement rates by original coeffi- the sharp corners and letting the newly exposed surface of the
cients of the improved soil columns’, which in essence is a soil-cement column contact a rigid base upon loading in the

19
Geotechnical Engineering Mass behaviour of embedded improved
Volume 164 Issue GE1 soil raft in an excavation
Yang, Tan and Leung

S, von Mises stress S, von Mises stress


(ave. crit.: 75%) (ave. crit.: 75%)
3 3
⫹3·4 ⫻ 10 ⫹3·4 ⫻ 10
3 3
⫹3·1 ⫻ 10 ⫹3·1 ⫻ 10
3 3
⫹2·8 ⫻ 10 ⫹2·8 ⫻ 10
3 3
⫹2·5 ⫻ 10 ⫹2·5 ⫻ 10
3 3
⫹2·3 ⫻ 10 ⫹2·3 ⫻ 10
3 3
⫹2·0 ⫻ 10 ⫹2·0 ⫻ 10
3 3
⫹1·7 ⫻ 10 ⫹1·7 ⫻ 10
3 3
⫹1·4 ⫻ 10 ⫹1·4 ⫻ 10
3 3
⫹1·1 ⫻ 10 ⫹1·1 ⫻ 10
2 2
⫹8·5 ⫻ 10 ⫹8·5 ⫻ 10
2 2
⫹5·7 ⫻ 10 ⫹5·7 ⫻ 10
2 2
⫹2·9 ⫻ 10 ⫹2·9 ⫻ 10
1 1
⫹1·1 ⫻ 10 ⫹1·1 ⫻ 10

2 NAKAGAWA 1996: PLANE STRAIN, A QUARTER, SPECIAL TREATMENT


2 NAKAGAWA 1996: PLANE STRAIN, A QUARTER, SPECIAL TREATMENT
ODB: TN_FL66.odb ABAQUS/Standard 6·4-4 Wed May 07 10:15 ODB: TN_FL66.odb ABAQUS/Standard 6·4-4 Wed May 07 10:15
3 1 1
Step: STEP_LOAD, LATERAL COMPRESSION, TWO SIDES CONSTRAINED Step: STEP_LOAD, LATERAL COMPRESSION, TWO SIDES CONSTRAINED
Increment 52: Step Time ⫽ 1·000 Increment 52: Step Time ⫽ 1·000
Primary Var: S, von Mises stress Primary Var: S, von Mises stress

Figure 15. Numerical model to simulate the reported improved


soil layer: left, model overview; right, close-up view of the sharp
corner

Abaqus model. This special treatment allows lateral forces to be situ soft soils and those of the improved soils based on Nakagawa
transmitted to the nearby soil-cement columns, and at the same et al.’s (1996) reported K values.
time accounts for the interaction with the surrounding soils, albeit
with small inaccuracies introduced. Similar treatment is applied The normalised mass stiffness from the simulation is shown in
to the corner at the right-hand side in Figure 15 to simulate the Figure 16. The mobilised mass stiffness is about 28% of the
constraining effects of the neighbouring row of soil-cement elemental properties of the soil-cement columns, slightly higher
columns. The model is loaded up to 0.33% mass strain, which is than the mobilised mass stiffness of one row of soil-cement
comparable to the mobilised level in the field (Nakagawa et al., columns just in contact with each other, because of the confining
1996). Compared with the model in the previous section, the effects of the surrounding soils and the adjacent columns. This
model as shown in Figure 6(b) will take into account the mobilised mass stiffness is still considerably lower than the
confining effects of the surrounding soils and the adjacent soil- elemental material stiffness. At least a 70% reduction is expected
cement columns. The material properties used for the improved due to such a geometric arrangement alone, in which the uniform
soils and the in situ soils are shown in Table 2. This set of deep cement mixing columns are arranged just in contact with
parameters largely reflects the typical material properties of in each other in a grid in soft soils. Though this is a simple three-
dimensional model of a representative unit, its ability to account
for a large part of the reduction of the mobilised lateral stiffness
from the elemental stiffness of the soil-cement columns is clearly
Material parameter Improved soil In situ soil demonstrated. The soil-cement columns arranged in such a
fashion are able to mobilise only very limited zones near the
Young’s modulus, E: MPa 396 60
contact point when they are laterally loaded. A large portion of
Poisson’s ratio,  0.495 0.495
the column experiences low mobilised stresses, as shown in the
Cohesion yield stress, c: kPa 1980 30
shear stress contours in Figure 15.
Friction angle, : degrees 0 0
Dilation angle, ł: degrees 0 0
As discussed earlier in this paper, soil-cement columns that are
Table 2. Material properties for the simulation of Nakagawa et formed in the field are often not uniform, and tend to show
al.’s (1996) case layered properties. A parametric study is then performed to assess
the effects of layering on the mobilised mass stiffness in the

20
Geotechnical Engineering Mass behaviour of embedded improved
Volume 164 Issue GE1 soil raft in an excavation
Yang, Tan and Leung

1·8 Outer layer: ⫻4 half the distance between the centres of two adjacent columns
1·6 Outer layer: ⫻2 (Figure 6(c)), is introduced to facilitate discussion. A smaller L
Normalised mass stiffness

Uniform columns value means a larger amount of overlap. The columns would be
1·4 Outer layer: ⫻1/2
in point contact if L equalled the column radius R. For the
1·2 Outer layer: ⫻1/4
triangular grid shown in Figure 6(d), L is (ˇ3/2)R, which gives L
1·0 ¼ 0.779 m for R ¼ 0.9 m.
0·8
0·6 Figure 17 shows that, compared with columns arranged to just
touch each other, when the columns are arranged to partly overlap
0·4
each other in a row, the mobilised mass stiffness at a very early
0·2
stage increases much more significantly. For example, the mass
0 stiffness is about 55% of the elemental material stiffness when
0 0·05 0·10 0·15 0·20 0·25 0·30 0·35 0·40
Mass strain: %
the overlap parameter L is 0.850 m. This set of results confirms
the importance of ensuring that such improved soil columns must
Figure 16. Mobilised mass stiffness of the improved soil layer be constructed to have adequate overlapping. To highlight this
(after Nakagawa et al., 1996) difference, Figure 18 compares the stress contours for the case of
columns that are just touching each other with the case where the
columns are overlapping. An examination of the von Mises stress
reported case history. The parametric study here is not a back- contours at 0.02% mass strain (Figure 18) shows that the stresses
analysis of the field case but a further assessment of the impact for the point contact case (just touching each other) are
of layering on the mobilised mass stiffness after taking into concentrated mainly near the contact point, whereas for the
account the geometric arrangement of soil-cement columns in an overlap case this is spread to a much wider area.
embedded improved soil raft. In the parametric study, the deep
cement mixing columns were assumed to have two concentric Obviously the next question is how the mobilised stiffness relates
layers that were of the same cross-sectional areas. The strength to the amount of overlap, as increased overlap would imply a
and the stiffness of the outer layers were varied from a quarter to higher cost of construction. Intuitively, the mass property of these
four times the initial values while those of the inner layers were overlapping columns should approach the elemental property as
kept unchanged. Such a range of properties in different layers the overlap becomes increasingly large. Since the mass stiffness
covers the possible variations as reported by Kawasaki et al. varies with the mass strain, its value at 0.8% mass strain was
(1984) and the Nicoll Highway field trial. The mobilised mass extracted and plotted in Figure 19 for comparison, and this figure
stiffness obtained from the parametric study is normalised to the shows that the mass stiffness approaches the elemental material
stiffness of the inner layer as shown in Figure 16. If the outer stiffness as the overlap parameter L decreases or the amount of
layer is stiffer, as in the Nicoll Highway field trial case, the overlap increases. An important point to note is that even with a
mobilised mass stiffness can reach about 55% of the inner layer very small overlap, for example L ¼ 0.897 m, the initial mass
stiffness. On the other hand, if the outer layer is less stiff, as stiffness is nearly twice that for the case where there is no
reported by Kawasaki et al. (1984), the mobilised mass stiffness overlap, just a point contact. Thus a small overlap is shown to
can be as low as 12% of the inner layer stiffness, a further have a large impact on the stress distribution within the column,
reduction from the mobilised mass stiffness of the uniform soil-
cement case. When reasonable variations of material properties in 1·0
different layers are considered in the model, the mobilised mass Overlap parameter, L ⫽ 0·850 m
0·9
Overlap parameter, L ⫽ 0·880 m
stiffness can range from about 55% to about 12% of the
Normalised mass stiffness

0·8 Overlap parameter, L ⫽ 0·897 m


elemental stiffness of the inner layers. This clearly suggests the
0·7 Overlap parameter, L ⫽ 0·900 m (point contact)
importance of knowing the arrangement of the columns as well
0·6
as the variation of properties within each soil-cement column in
0·5
assessing the mobilised mass properties.
0·4
5.3 One row of overlapping columns 0·3
The previous two sections have shown that if the columns are 0·2
arranged to just touch each other, the mobilised mass stiffness is 0·1
significantly lower than that of the elemental behaviour, because 0
0 0·5 1·0 1·5 2·0
the original point contact has room to deform. To overcome this,
Mass strain: %
in practice the columns are usually arranged to overlap each
other. This configuration is simulated next, as shown in Figure Figure 17. Effects of overlapping on the mass behaviour of one
6(c). Again, an isotropic linear elastic model is assumed for the row of overlapping columns
individual element. An overlap parameter L, which is defined as

21
Geotechnical Engineering Mass behaviour of embedded improved
Volume 164 Issue GE1 soil raft in an excavation
Yang, Tan and Leung

S, von Mises stress S, von Mises stress


(ave. crit.: 75%) (ave. crit.: 75%)
2
⫹9·000 ⫻ 10 ⫹9·000 ⫻ 102
⫹8·500 ⫻ 102 ⫹8·500 ⫻ 102
⫹8·000 ⫻ 102 ⫹8·000 ⫻ 102
⫹7·500 ⫻ 102 ⫹7·500 ⫻ 102
⫹7·000 ⫻ 102 ⫹7·000 ⫻ 102
⫹6·500 ⫻ 102 ⫹6·500 ⫻ 102
⫹6·000 ⫻ 102 ⫹6·000 ⫻ 102
⫹5·500 ⫻ 102 ⫹5·500 ⫻ 102
⫹5·000 ⫻ 102 ⫹5·000 ⫻ 102
⫹4·500 ⫻ 102 ⫹4·500 ⫻ 102
⫹4·000 ⫻ 102 ⫹4·000 ⫻ 102
⫹3·500 ⫻ 102 ⫹3·500 ⫻ 102
⫹3·000 ⫻ 102 ⫹3·000 ⫻ 102
⫹2·500 ⫻ 102 ⫹2·500 ⫻ 102
⫹2·000 ⫻ 102 ⫹2·000 ⫻ 102
⫹1·500 ⫻ 102 ⫹1·500 ⫻ 102
⫹1·000 ⫻ 102 ⫹1·000 ⫻ 102
⫹5·000 ⫻ 101 ⫹5·000 ⫻ 101
⫹0·000 ⫻ 100 ⫹0·000 ⫻ 100

Figure 18. Von Mises stress contours at 0.02% mass strain: left,
very small overlap case; right, point contact case

As discussed in the early part of this paper, soil-cement columns


in the field tend not to be homogeneous, often having properties
1·0
Normalised mass stiffness

in the outer layer that are different from those of the inner layer.
It is therefore necessary to examine the effects of such material
0·8
layering on the overlapping soil-cement columns. Figure 20
0·6 shows the results for one row of overlapping columns with two
layers having equal cross-sectional areas but different stiffnesses.
0·4 The overlap parameter L is 0.897 m. Similar to the mass behav-
Elemental stiffness
Normalised mass stiffness iour shown in Figure 12, the mass stiffness is higher when the
0·2 outer layer is assigned the stiffer material. A stiffer outer layer
also helps maintain the initial mass stiffness until a larger mass
0 strain is mobilised, which is a very important contribution to the
0·90 0·85 0·80 0·75 0·70 0·65 0·60
control of wall deflections and associated ground movements in
Overlap parameter, L: m
deep excavation.
Figure 19. Mass stiffness at 0.8% mass strain against the overlap
parameter L (for R ¼ 0.9 m) Parametric studies were carried out to evaluate the effect of the
ratio of stiffness of the outer layer to that of the inner layer on
the initial mass stiffness for different degrees of overlap. For each
overlap parameter the stiffness ratio is varied from 0.125 to 8, a
as a larger area is mobilised right from the start of loading. To range that covers the reported variation from the field (Kawasaki
achieve 80% of the elemental material stiffness, the overlap et al., 1984). The soil-cement columns are assumed to have equal
parameter L should be less than 0.75 m for one row of uniform cross-sectional areas. The results are shown in Figure 21. The
soil-cement columns with radius of 0.9 m. If a line is drawn vertical axis shows the initial mass stiffness normalised by the
between the centres of two connecting soil-cement columns, 30% elemental stiffness of the inner layer. Generally, increasing the
of the column radius must overlap another column along that degree of overlap helps make the composite system stiffer; and
line. this effect is more evident when the stiffness ratio increases. In
other words, even if sufficient overlap in the field is ensured, the
What Figure 19 also shows is that, once the uniform soil-cement benefit from overlapping can still be offset if the outer layer is
columns are overlapped to some extent, the mass stiffness weaker than the inner layer. It is therefore important to improve
becomes less sensitive to the degree of overlap. However, as this the workmanship in the field to ensure good control of the
figure also shows, a certain reduction from the elemental proper- positions of the soil-cement columns. It is also important to core
ties is required to take into account the effect of overlapping if and test samples that are located in the outer layer.
the test results of the elemental cores are to represent the mass
behaviour. Common overlapping used in practice would typically 5.4 Multiple overlapping columns
imply that the mobilised stiffness will see a reduction of at least In this section, soil-cement columns overlapping in an equilateral
30% from the elemental stiffness. triangular grid as shown in Figure 6(d) are studied. Two

22
Geotechnical Engineering Mass behaviour of embedded improved
Volume 164 Issue GE1 soil raft in an excavation
Yang, Tan and Leung

500 2·0
Inner 75 MPa; outer 300 MPa Overlap parameter, L ⫽ 0·779 m
L ⫽ 0·850 m
Inner 300 MPa; outer 75 MPa L ⫽ 0·880 m
1·5 L ⫽ 0·897 m
400 L ⫽ 0·900 m

Emax /Einner
1·0
Mass stress: kPa

300

0·5

200
0
0·125 0·25 0·5 1 2 4 8
Eouter /Einner
100
Figure 21. Normalised initial mass stiffness for one row of
overlapping columns

0
0 0·5 1·0 1·5 2·0
Mass strain: % 10
Volume ratio: outer layer 90%
Volume ratio: outer layer 70%
50 8 Volume ratio: outer layer 50%
Volume ratio: outer layer 30%
Volume ratio: outer layer 10%
Emass /Einner

6
40

4
Mass stiffness: MPa

30 2

0
20 0·125 0·25 0·5 1 2 4 8
Eouter /Einner

Figure 22. Normalised initial mass stiffness for multiple


10 overlapping soil-cement columns

0 expected, the variation in initial mass stiffness is the most


0 0·5 1·0 1·5 2·0
Mass strain: % significant for a volume ratio of 90% when the stiffness ratio
varies. It is difficult to separate the contribution of the outer layer,
Figure 20. Mass behaviour of one row of overlapping columns as the relative volume ratio of the outer layer also contributes to
with overlap parameter L ¼ 0.897 m the overall mass stiffness. To separate these two factors, the
initial mass stiffness for a volume ratio of 50% is normalised to
the material stiffness of both layers, as shown in Figure 23. The
parameters are examined: the stiffness ratio of the outer layer to variation of the normalised initial mass stiffness is smaller if it is
the inner layer, and the volume ratio of the outer layer to the normalised to the stiffness of the outer layer material. If the outer
whole soil raft. A volume ratio of 50% means that the inner and layer properties are known, there is less variation in the mass
outer layers have the same cross-sectional area. The results are properties than if the inner layer properties are known. Again,
shown in Figure 22. The initial mass stiffness is again normalised this observation emphasises that the outer layer is more important
by the stiffness of the inner layer material. For uniform soil- in evaluating the mass properties for an embedded improved soil
cement columns with stiffness ratio 1, overlapped in the grid as raft used for stabilising an excavation. Therefore, for core
shown in Figure 6(d), the corresponding mass stiffness must be sampling in the field, more attention should be paid to the outer
the same as the elemental stiffness. So all the curves in Figure 22 layer. Usually these columns are constructed prior to the excava-
pass through the same point (1, 1). The scatter in the initial mass tion, and no visual observation is possible during coring. With the
stiffness is greater as the stiffness ratio deviates from 1. As present understanding of the need to know the properties of the

23
Geotechnical Engineering Mass behaviour of embedded improved
Volume 164 Issue GE1 soil raft in an excavation
Yang, Tan and Leung

5
Normalised to outer layer stiffness columns. Overlapping the uniform soil-cement columns in a row
Normalised to inner layer stiffness helps to improve the mobilised initial mass stiffness of the soil
Normalised initial mass stiffness

4 raft. For a 30% overlap in terms of radius in the compressive


loading direction, 80% of the elemental material stiffness can be
3 achieved. Based on these observations, it is suggested that the
soil-cement columns be constructed to overlap, rather than just be
in contact with each other. The importance of overlap also calls
2
for good quality control when installing soil-cement columns in
the field.
1
Soil-cement columns often have layered properties, and the outer
0 layer plays a more important role in the mass behaviour of the
0·125 0·25 0·5 1 2 4 8
composite ground. Increasing the stiffness of the outer layer will
Eouter /Einner
have a greater impact on improving the mass stiffness than
increasing that of the inner layers. With the understanding gained
Figure 23. Normalised initial mass stiffness for multiple
from the present study, in determining the mobilised mass proper-
overlapping soil-cement columns: volume ratio 50%
ties, it is more important to know the properties of the outer
layer. A recommendation is therefore made that for sampling in
an embedded improved soil raft, at least two core samples located
less than a radius apart have to be taken to give a better idea of
outer layer better than those of the inner layer, a practical solution the layering profile of the in situ soil-cement columns. While this
is to take two samples from the centre of a column not more than is still not perfect, it will be able to provide a better idea of the
a radius apart. While this is not a perfect solution, it goes a long likely variation of properties within a column, and therefore of
way to alleviate the situation. the mobilised properties. The mobilised mass properties can then
be estimated by taking into account the geometric arrangements
6. Conclusions as well as the possible variation in the elemental properties.
In many deep excavations in soft soils in urban areas, a layer of
soil below the formation level is often improved to stabilise the REFERENCES
excavation. The improvement, usually by a deep cement mixing HKS (2003) ABAQUS User’s Manuals: Version 6.4. Hibbitt,
or jet-grouting method, will result in a row of overlapping short Karlsson and Sorensen, Inc., Pawtucket, RI.
columns being constructed. Soil investigation carried out will Hsi JP and Yu JBY (2005) Jet grout application for excavation in
usually provide information on vertical cores obtained from such soft marine clay. Proceedings of the 16th International
columns, but these columns are usually subjected to lateral (and Conference on Soil Mechanics and Geotechnical
usually horizontal) load imposed by the inward-moving retaining Engineering, Osaka. Millpress, Rotterdam, the Netherlands,
wall. It is the mobilised mass properties in the lateral direction vol. 3, pp. 1485–1488.
that are of importance in the design of the walls of such an Johnson KL (1985) Contact Mechanics. Cambridge University
excavation. In this paper, the mechanisms of how the mass Press, New York.
properties are mobilised and their relation to the elemental Kawasaki T, Saitoh S, Suzuki Y and Babasaki R (1984) Deep
properties, the variation of properties within a column and how mixing method using cement slurry as hardening agent.
the columns are arranged are examined in some detail. Proceedings of a Seminar on Soil Improvement and
Construction Techniques in Soft Ground, Singapore, pp. 17–
When a row of uniform soil-cement columns in point contact, 38.
with isotropic linearly elastic properties, is compressed laterally, Nakagawa S, Kamegaya I, Kureha K and Yoshida T (1996) Case
the mass behaviour is non-linear; and the mass stiffness is around history and behavioural analyses of braced large scale open
22% of the elemental material stiffness. This one single fact excavation in very soft reclaimed land in coastal area. In
highlights the important need to distinguish mobilised mass Proceedings of the International Symposium on Geotechnical
properties from the elemental properties. The simulation of a Aspects of Underground Construction in Soft Ground (Mair
reported case history also confirms that the mobilised mass RJ and Taylor RN (eds)). Balkema, Rotterdam, the
properties of soil-cement columns arranged just in contact with Netherlands, pp. 179–184.
each other in soft soils are only about 28% of the element O’Rourke TD and McGinn AJ (2006) Lessons learned for ground
material properties. The simulation further shows that if the outer movements and soil stabilization from the Boston Central
layer is less stiff in such a geometric arrangement, the mobilised Artery. Journal of Geotechnical and Geoenvironmental
mass stiffness can reduce to as little as 12% of the elemental Engineering 132(8): 966–989.
material stiffness, even with the contribution of the confining O’Rourke TD and O’Donnell CJ (1997) Field behavior of
effects from the in situ soils and the adjacent rows of soil-cement excavation stabilized by deep soil mixing. Journal of

24
Geotechnical Engineering Mass behaviour of embedded improved
Volume 164 Issue GE1 soil raft in an excavation
Yang, Tan and Leung

Geotechnical and Geoenvironmental Engineering 123(6): and Ground Treatment (Johnsen LF, Bruce DA and Byle MJ
516–524. (eds)). American Society of Civil Engineers, Reston, VA,
Ou CY, Wu TS and Hsieh HS (1996) Analysis of deep excavation USA, Geotechnical Special Publication No. 120, Vol. 1, pp.
with column type of ground improvement in soft clay. 198–217.
Journal of Geotechnical and Geoenvironmental Engineering Shirlaw JN (2003) Jet grouting soft clays for tunnelling and deep
122(9): 709–716. excavations: design and construction issues. In Grouting and
Pickles AR and Henderson TO (2005) Some thoughts on the use Ground Treatment (Johnsen LF, Bruce DA and Byle MJ
of numerical modelling in geotechnical design practice. (eds)). American Society of Civil Engineers, Reston, VA,
Proceedings of Underground Singapore 2005: Special Session USA, Geotechnical Special Publication No. 120, Vol. 1, pp.
Numerical Analysis in Geotechnical Engineering, Singapore, 257–268.
11 pp. Tan TS and Shirlaw JN (2000) Braced excavation: excavation in
Porbaha A (2000) State of the art in deep mixing technology. general. In Proceedings of the 3rd International Symposium
Part IV: Design considerations. Ground Improvement 4(3): on Geotechnical Aspects of Underground Construction in Soft
111–125. Ground, Tokyo (Kusakabe O, Fujita K and Miyazaki Y (eds)).
Porbaha A (2002) State of the art in quality assessment of deep Balkema, Rotterdam, the Netherlands, pp. 53–62.
mixing technology. Ground Improvement 6(3): 95–120, doi: Tanaka H (1993) Behavior of braced excavations stabilized
10.1680/grim.2002.6.3.95. by deep mixing method. Soils and Foundations 33(2): 105–
Porbaha A, Asada H and Fatemi MJ (1999) Deep mixing 115.
technology for deep excavation. In Proceedings of the 3rd Wen DZ (2005) Use of jet grouting in deep excavations. In
National Conference on Geo-Engineering for Underground Ground Improvement – Case Histories (Indraratna B and Jian
Facilities, Urbana-Champaign (Fernandez G and Bauer RA C (eds)). Elsevier, Amsterdam, pp. 357–370.
(eds)). American Society of Civil Engineers, Reston, VA, Wong IH and Poh TY (2000) Effects of jet grouting on adjacent
USA, pp. 1048–1059. ground and structures. Journal of Geotechnical and
Shibazaki M (2003) State of practice of jet grouting. In Grouting Geoenvironmental Engineering 126(3): 247–256.

WHAT DO YOU THINK?


To discuss this paper, please email up to 500 words to the
editor at journals@ice.org.uk. Your contribution will be
forwarded to the author(s) for a reply and, if considered
appropriate by the editorial panel, will be published as a
discussion in a future issue of the journal.
Proceedings journals rely entirely on contributions sent in
by civil engineering professionals, academics and students.
Papers should be 2000–5000 words long (briefing papers
should be 1000–2000 words long), with adequate illustra-
tions and references. You can submit your paper online via
www.icevirtuallibrary.com/content/journals, where you
will also find detailed author guidelines.

25

You might also like