Baugrundtagung 2008, Dortmund

Foundation design for an immersed tunnel with CDM and SCP soil improvement of soft
marine clay

T. Kasper*1, P.G. Jackson1, R.W.M.G. Heijmans2 & G. Meinhardt2

1
COWI A/S, Parallelvej 2, 2800 Kongens Lyngby, Denmark
2
ARCADIS Infra BV, Piet Mondriaanlaan 26, 3800 AE Amersfoort, Netherlands

Abstract

A 3.2 km long immersed road tunnel forms part of the Busan-Geoje Fixed Link in South Korea, which
connects Koreas second largest city Busan with the island of Geoje. To ensure the water tightness of the
joints of the segmental tunnel structure and to limit the forces in the shear keys, the differential
settlements have to be limited. This is achieved by extensive soil improvement of the marine clay below
the tunnel with cement deep mixing (CDM) and sand compaction piles (SCP). This paper is concerned
with the characteristics of the marine clay, the basic design principles, layout and constructional issues
of the soil improvement as well as the numerical settlement and stability analyses.

1. Introduction

The tunnel of the Busan-Geoje Fixed Link [1] consists of two short cut & cover tunnel access ramps and
a 3.2 km long immersed tunnel (Figure 1). The 180 m long concrete tunnel elements of the immersed
tunnel have a 26.5 m wide and 10 m high rectangular cross-section and are cast in a dry dock on shore.
They are floated out to the sea, immersed to their final positions on a gravel bed in pre-excavated
trenches in water depths of up to 50 m and connected to each other. After immersion, the tunnel is
protected by backfill material at the sides and on top. Each tunnel element consists of eight 22.5 m long
segments. The joints between the segments and between the tunnel elements are equipped with rubber
gaskets and concrete shear keys. The displacement behaviour of the segmental tunnel structure in
varying ground conditions represents a complex soil-structure interaction problem. In order to ensure
the water tightness of the joints and to limit the forces in the shear keys, the differential longitudinal
settlements have to be limited. Without soil improvement of the marine clay below the tunnel, long-term
tunnel settlements of up to 35 cm and too large differential settlements were predicted by calculations in
an early stage of the project. It was therefore decided to carry out extensive soil improvement with
cement deep mixing (CDM) and sand compaction piles (SCP).

Figure 1: Longitudinal section of the tunnel.

* Corresponding author. Phone: +45 45972211. Fax: +45 45972212. Email: [email protected].
Website: www.cowi.com.
At a length of 2.2 km, where the tunnel is placed in a trench, the clay below has a depth of up to 15 m
and is improved with partial depth CDM columns (Figure 2, left). The tunnel alignment also requires a
shorter underwater embankment in the western part, where the clay below has a depth of up to 30 m and
is improved with full-depth SCP (Figure 2, right).

Figure 2: Typical layout of the CDM foundation at tunnel elements TE3 to TE14 (left) and the SCP foundation at
tunnel elements TE15 to TE17 (right).

2. Ground investigations

As the soft marine clay requires special attention and necessitates special foundation solutions,
extensive ground investigations and laboratory testing programmes have been carried out [2]. The field
tests comprised 50 boreholes with soil sampling, field vane and SPT tests and 60 CPT soundings with
pore pressure measurements. Classification tests have shown that the clay exhibits high to extremely
high plasticity. The determined range of the liquid limit was 68 - 117 % (average 96 %) and the range of
the plasticity index was 30 - 85 % (average 63 %). The saturated unit weight was found to vary between
11.5 and 16.2 kN/m3 (average 14.7 kN/m3) and the void ratio varies between 1.99 and 3.24 (average
2.44). The strength properties of the clay were assessed from the results of CPT and triaxial tests. Based
on the SHANSEP approach, the undrained strength can be described as cu = 0.26σ'v0((σ'v0+∆σ)/σ'v0)0.76,
with the preconsolidation stress ∆σ of the marine clay typically ranging between 10 and 30, max. 65
kPa. The effective strength parameters φ' = 25º and c' = 3 kPa were derived from consolidated undrained
triaxial tests and due consideration of the general experience from the area. Oedometer tests have shown
ranges of the primary compression index Cc and recompression index Crc of 0.83 - 1.82 (average 1.25)
and 0.041 - 0.133 (average 0.091), respectively. The hydraulic permeability of the clay is in the order of
10-9 m/s.

3. Foundation solutions with soil improvement

The normal foundation concept for immersed tunnels is a simple foundation with a gravel bed on the
underlying soil. However, this foundation concept was found to be insufficiently robust mainly due to
the sensitivity to changes in load caused by construction tolerances and due to creep related settlements,
together with the uncertainties in their prediction. To make the foundation more robust, soil
improvement with cement deep mixing (CDM) and sand compaction piles (SCP) has been chosen. The
reason for using SCP at the embankment section is to allow quick drainage and thus to achieve quick
strength gain of the marine clay. This avoids stability problems of the embankment and allows
sufficiently quick embankment construction between typhoon seasons.

3.1 Cement deep mixing (CDM)

Where the tunnel is placed in a trench (from tunnel element TE3 to TE14, cp. Figure 1) the clay is
strengthened with walls formed by contiguous columns of mixed cement and clay (Figure 2, left). This
form of foundation has previously been used in Korea and is in wide-spread use in Japan [3] and in
northern Europe for settlement control of structures on soft clay deposits. The CDM columns are
constructed using the wet mixing method, where cement slurry is mixed with the clay. The mixing
shafts shown in Figure 3 (left) are first drilled with the rotating blades into the soil to the desired depth.
The cement slurry is injected in the withdrawal stage, where the vertical speed of the machine and the
flow rate of the cement slurry are kept constant. Four 1 m diameter columns with overlaps of 10 cm
form a 1.9 m square column pattern. To form continuous walls, these square columns are placed with 10
cm overlaps in rows (Figure 2, left). The offshore CDM production is illustrated in Figure 3 (right). The
CDM walls reach depths of up to 60 m below sea level (cp. Figure 1). The heaved soil in the trench due
to CDM construction is dredged away before tunnel construction takes place.

Figure 3: Mixing blades of the deep mixing machine (left) and offshore CDM production barge (right).

The equivalent unit weight of the tunnel is only some 12 % higher than water. The treatment of the clay
is concentrated where the loads are largest and thus the treatment ratio of 35 % beneath the centre of the
tunnel is less than the treatment ratio of 66 % beneath the backfill and the edges of the tunnel. The
chosen partial depth treatment of the marine clay enables a load share between the soil and the CDM
columns. The major part of the foundation loads is transferred to a lower level in the marine clay where
its strength and stiffness properties are significantly better. An important advantage of the load-sharing
concept is that the contact pressure below the foundation can be controlled, thereby assuring shearing
resistance in the case of lateral static, dynamic or cyclic loading of the tunnel. Another benefit of load-
sharing is that the foundation system becomes “self-adjusting”, whereby a gradual transfer of load
between the clay and the stabilizing elements takes place until equilibrium is achieved. The depth of
treatment can be adapted along the tunnel alignment to accommodate for differences in geotechnical
conditions and specific requirements at transitions between the different foundation types. A full-depth
solution would induce a high risk of uneven load distribution and stress peaks in the CDM walls, which
would in turn lead to undesired local stress concentrations in the bottom slab of the tunnel.
3.2 Sand compaction piles (SCP)

Sand compaction piles [4] are used where the tunnel is constructed in an embankment (from tunnel
element TE15 to TE17 according to Figure 1). Sand compaction piles have been chosen in order to
allow quick consolidation and strength gain of the clay and thus sufficiently quick embankment
construction between typhoon seasons, and to introduce a favourable ductile behaviour of the
foundation under earthquake impact. The risks of embankment construction on the marine clay are
illustrated in [5], where failure of a breakwater close to the construction site of the immersed tunnel is
discussed.
The design and optimization criteria of the SCP foundation are
• Sufficient preloading of the SCP by a preloading embankment
• Sufficient safety against failure of the preloading embankment during construction
• Sufficient safety against failure in the final situation considering the effect of wave and
earthquake impact
• Control of the tunnel settlements
• Control of the irrecoverable displacements of the embankment / tunnel due to earthquake impact
An SCP machine with vertical vibrating compaction technique has been used for the project. As
illustrated in Figure 4 (left), the casing pipes are driven into the ground by vertical vibratory excitation
from the vibro-hammer on top of the casing pipe. During the penetration, the casing pipes are filled with
SCP material. After reaching the prescribed depth, the casing pipe is pulled about 1 m to feed the sand
into the ground with the help of compressed air. The sand is compacted by the vertical vibratory
excitation and the degree of compaction is controlled such that the diameter of the sand pile becomes the
designed value. This procedure continues until the sand piles are completed. The depth of the casing
pipes and the sand levels in the casing pipes are continuously measured. The piles have a diameter of 2
m and reach depths of up to 65 m below sea level (cp. Figure 1). The offshore SCP production barge is
shown in Figure 4 (right). The chosen treatment ratio is 40 %. However, an increased treatment ratio of
61 % is applied below the edges of the tunnel to support it against downdrag resulting from wall friction
due to placement of the fill material.
Positioning
Penetration finished Construction finished

Construction of
Penetration sand pile
Sand
Vibrator Bohrrohr

Sand level
gauge
Retrieval Penetration

Sand bed

Trace of pipe bottom

Figure 4: SCP execution procedure according to [4] (left) and offshore SCP production barge (right).
4. Modelling

The self-weight of the tunnel and backfill and the downdrag resulting from friction between the backfill
and the walls of the tunnel represent the permanent, settlement inducing loads. One of the reasons why a
tunnel forms part of the Fixed Link is to accommodate heavy ship traffic to a large nearby harbour.
Sunken ship impact has been considered as a load of 30 MN applied uniformly over one tunnel segment,
i.e. 30 / 22.5 = 1.33 MN/m. The design was supported and optimized by a large number of settlement
predictions and stability analyses with elasto-plastic finite element calculations. Settlements and
equivalent soil springs are determined by 2D plane strain analyses of tunnel cross-sections with the
finite element program Plaxis [6].

Lift 1, 2 and 3 of preloading embankment

Figure 5: A typical FE model for the trench section with CDM foundation (left) and FE model for the highest
embankment section with SCP foundation (right).

Figure 5 shows two examples of finite element models, one for the tunnel in the trench section on the
left hand side and one for the tunnel at the highest embankment section on the right hand side. The
bottom of the models is assumed to be fixed in both directions, while the vertical model boundaries are
fixed in horizontal direction. To account for the actual 3D geometry, the sand compaction piles are
modelled with equivalent widths. Higher-order 15-node triangular elements are used to achieve a good
quality of the results. The marine clay and the sand compaction piles are modelled with an advanced
elasto-plastic model called Hardening soil model [7]. The parameters for the clay have been derived
from the results of laboratory tests, while the parameters for the sand compaction piles are best estimates
for dense sand. The material parameters are summarised in Table 1.

Table 1: Material models and parameters.

Marine clay Alluvium CDM SCP Gravel bed Fill + Armour

Material model Hardening Mohr-Coulomb Linear elastic Hardening Mohr-Coulomb Mohr-Coulomb
soil model soil model
γ (kN/m3) 14.7 20 20 20 20 20
ref
E50 (MPa) 1.2 50 80 39 50 50
ν (-) 0.2 0.25 0.2 0.3 0.2 0.2
ref
E oed (MPa) 0.6 33
Eref
ur (MPa) 8 120
m (-) 1 0.5
pref (kPa) 100 100
c' (kPa) 3 0 0 0 0
φ' (º) 25 35 33 36 40
k (m/s) 1.10-9 1.10-5 1.10-9 1.10-4 1.10-1 1.10-1
The construction sequence is modelled in several stages:
• Initial state. Initial stresses in the clay and alluvium are generated based on self-weight and
Jaky's formula K0 = 1 - sin φ'. Pore pressures are initialised based on the sea level and the unit
weight of water.
• In the trench section: Modelling of trench excavation by deactivation of the respective finite
elements.
• To consider the installation of CDM and SCP, the material properties are changed in the
respective finite elements. This simplified approach neglects the effect of CDM and SCP
installation (remoulding and compaction) on stresses and consequently strength and is
considered to be conservative.
• In the embankment section: Modelling of the preloading embankment construction in 4 phases
(lift 1, lift 2, lift 3 and subsequent partial excavation) by activation and deactivation of the
respective finite elements.
• Modelling of tunnel construction by activation of finite elements and change of material
properties.
When elements are (re-)activated, their stresses are initialised to 0. Depending on the purpose of the
analysis and the result of interest (e.g. instantaneous settlements, final settlements or development of
settlements in time), the individual stages are modelled as undrained, drained or consolidation
processes. The displacements are reset to zero at the beginning of the last stage in order to separate the
tunnel settlements from the settlements in the previous stages. Also in the last stage, interface elements
representing a wall friction angle of δ = 2/3 φ' are considered around the tunnel. Additional stages are
calculated at the end to consider loads such as trench sedimentation, sunken ship or dragged anchor
impact or to determine equivalent subgrade reaction moduli (load-displacement curves) of the tunnel
foundation. Furthermore, strength reduction analyses [8] are carried out for the different stages to
determine the corresponding safety factors. A satisfying safety factor of 2.1 has been determined for the
slope stability of the tunnel protection material in the trench section. The safety factor for the stability of
the highest embankment section on SCP is 1.5 (undrained, just after tunnel construction) and 1.9 (in the
final situation after full consolidation). The trench section is exposed to possible impact from large
waves. The stability of the open trench during construction with consideration of wave impact has been
investigated in [9], while the stability of the tunnel under wave impact after construction has been
investigated in [10]. The predicted average final tunnel settlements are in the order of 9 cm at the trench
section and 5 cm at the embankment section (cp. examples in Figure 6 and Figure 7). Based on a large
number of cross-sections and sensitivity studies, possible differential settlements were carefully
evaluated and have been found acceptable. The finite element analyses allowed to optimize the CDM
and SCP layout (i.e. position, number and depth of the reinforcing elements) and to investigate the
effect of soil parameter and geometry variations. The analyses were used to verify the loading of the
reinforcing elements and further allowed to design the geometry as well as construction sequence and
periods of the preloading embankment on SCP.

Figure 6: Example of settlements due to tunnel construction in the trench section (left) and vertical effective stresses
(right).
Figure 7: Example of settlements due to tunnel construction at the embankment section (left) and vertical effective
stresses (right).

The equivalent subgrade reaction moduli of the tunnel foundation and their variations as obtained from
the Plaxis calculations have been used as input for soil springs in a 3D structural model of the tunnel.
This model, which also considers the joints of the segmental tunnel structure, has been used to verify
that the behaviour of the tunnel and the resulting joint openings for different loading situations are
satisfactory.

5. Conclusions

Too large total and differential settlements were predicted for the segmental structure of the immersed
tunnel of the Busan-Geoje Fixed Link in case of a simple foundation on the marine clay. It was therefore
decided to carry out soil improvement of the clay with CDM and SCP. The soil improvement limits the
tunnel settlements and provides the necessary robustness of the foundation. Numerical models have
been applied as a complement to traditional design methods and have been found to be a very useful
tool for the design, optimization and verification of the chosen foundation solution. The soil
improvement was finished last year and the first tunnel elements have been immersed earlier this year.

6. References

[1] Odgaard, S.S., Jensen, O.P., Kasper, T., Yoon, Y.H., Chang, Y., Park, R.Y., 2006. Design of
long immersed tunnel for highway in offshore conditions - Busan-Geoje Fixed Link. Tunnelling
and Underground Space Technology 21(3-4), Special Issue: Safety in Underground Space (CD-
ROM Proceedings of the ITA-AITES 2006 World Tunnel Congress, Seoul, Korea).
[2] Steenfelt, J.S., Jackson, P.G., Christensen, C.T., Lee, J-S., Ha, Y-B., 2008. Ground
investigations for the Busan - Geoje Immersed Tunnel. In: Huang, A-B., Mayne, P.W. (Eds.):
Geotechnical and Geophysical Site Characterization. Taylor & Francis, London, 1431-1436.
[3] Coastal development institute of technology (CDIT), Japan, 1999. The deep mixing method.
Principle, design and construction. Balkema, Rotterdam.
[4] Kitazume, M., 2005. The Sand Compaction Pile Method. Taylor & Francis, London.
[5] Chung, S.G., Kim, S.K., Kang, Y.J., Im, J.C., Nagendra Prasad, K., 2006. Failure of a
breakwater founded on a thick normally consolidated clay layer. Géotechnique, 56(6), 393-409.
[6] Plaxis b.v., the Netherlands: User's manuals, Plaxis version 8. www.plaxis.nl.
[7] Schanz, T., Vermeer, P.A., Bonnier, P.G., 1999. The Hardening-Soil Model: Formulation and
verification. In: Brinkgreve, R.B.J. (Ed.): Beyond 2000 in Computational Geotechnics.
Balkema, Rotterdam, 281-290.
[8] Brinkgreve, R.B.J., Bakker, H.L., 1991. Non-linear finite element analysis of safety factors. In:
Beer, G., Booker, J.R., Carter, J.P. (Eds.): Computer Methods and Advances in Geomechanics:
1117-1122. Rotterdam: Balkema.
 [9] Kasper, T., Jackson, P.G., 2008. Stability of an underwater trench in marine clay under ocean
wave impact. Proc. of the Conference on Geotechnical Aspects of Underground Construction in
Soft Ground, Shanghai, 2008.
[10] Kasper, T., Steenfelt, J.S., Pedersen, L.M., Jackson, P.G., Heijmans, R.W.M.G., 2008. Stability
of an immersed tunnel in offshore conditions under deep water wave impact. Coastal
Engineering, in press.