Deep Excavation in Hong Kong PDF
Deep Excavation in Hong Kong PDF
Deep Excavation in Hong Kong PDF
Abstract: This paper is presented in three parts. The first part discusses the design of deep
excavations including limit states, partial safety factors and ease of construction. The
second part discusses recent trends in numerical analysis with particular emphasis on
reliable displacement prediction while maintaining control over the input parameters. The
third part discusses the potential benefit of using the observational method and discusses
how the benefits of this method can be employed in the highly regulated regime that operates
within Hong Kong. A common theme in throughout this paper is the successful application
of design and construction methods that maximize economy while ensuring safety.
INTRODUCTION
Due to the shortage and therefore high cost of land in Hong Kong the use of deep excavations
has been widespread for over 30 years. While their design and construction has generally
been successful there have been some notable failures (GEO 2002). It is clearly important
that the design process meets the proper balance between safety and economy. The design
process must understand what may go wrong and that the effects on neighbouring buildings
and structures are adequately predicted.
This paper presents an overview of current and developing practice of the design of deep
excavations in Hong Kong. It is in three parts and addresses the following aspects
• Design considerations.
• Numerical modelling.
• The observational method.
For a deep excavation where there is soft marine clay under recent fill, particular care is
required to ensure the stratigraphy and undrained strength of the clay is adequately
understood. It is tempting to characterise the clay material as being drained and using
drained strength parameters but this is not a conservative assumption especially when
assessing the passive resistance provided by the soft clay beneath the excavation.
Fill
Passive well
CDV
MDV
Ensuring Buildability
The review of past cases of failures and excessive displacements of deep excavations in Hong
Kong (GEO 2002) showed that most problems have arisen because of inadequate sheet pile
penetration or because of removal of some (or even all) of the temporary props supporting the
wall. In many cases this was the result of the design being difficult to construct. While it is
clear that that having a sheet pile retaining wall penetrating into very strong material is good
from a toe stability perspective it is not a good design if it is virtually impossible to achieve
this penetration. Admittedly the penetration could probably be achieved with pre-boring but
this expensive procedure needs to be specified and costed into the project at the time of tender
if the contractor is to be realistically expected to carry out it out on any but a very minor scale.
Many levels of propping are also shown to be a potential problem. While using multiple
levels of propping at close vertical centres of 2m or so will reduce the lateral displacement of
the wall it causes major problems to the excavation contractor as he attempts to fit plant and
machinery within the props when removing the material from the base of the excavation.
Inevitably the contractor will tend to over-excavate and even temporarily remove props to be
able to manoeuvre his equipment. Experience has shown that the use of more substantial
and more widely spaced props provides a more workable and therefore safer solution. In
certain circumstances where a TBM is to be launched from within the deep excavation a clear
working height of over 10m may be necessary and diaphragm wall combined with a propping
spacing of the order of 5 to 6m has been found to be effective. Top down excavations, where
the propping is provided by the permanent floor slabs, also prove to be effective in this
regard.
The computer programs could also be used for back analysis of field results and it was quite
usual to find that the stiffness of the ground was significantly greater, say 3 to 9 times greater,
than measurements on small test specimens, (Endicott 1984).
In the mid 1980’s computer programs were made generally available on a commercial basis
and continuum models incorporating 2-D finite element or finite difference methods provided
the opportunity to model ground movements generally. For example, FLAC, (ITASCA
1987), has been used for retaining wall design in Hong Kong, since 1986, (Endicott and
Cheung 1991).
It was found that predictions of ground movement based on simple models with linear
elastic/plastic solutions, when compared with field monitoring, show that predicted settlement
is affected by the width of the numerical model. This effect can be overcome. When soil
stiffness is varied, high stiffness for small strains and small stiffness for large strains, a closer
match is obtained. The effect can be achieved by using sophisticated constitutive models
such as “Bricks on strings” (Simpson 1992), although such models are tools for specialists.
For practical purposes there is a simple approximation for retaining walls. One can
manually assign high stiffness at a distance from the wall where the strains are small and
small stiffness close to the wall. By this means one can closely model behaviour observed in
the field.
In practice, deep excavations are seldom carried out in 2-D. Usually there are other factors
such as staged construction, or the excavations are of limited length. Many basements are
nearly square in plan. These effects are likely to restrain movements to less than those
obtained when using basic parameters and 2-D analysis. Likewise back analysis of
monitored movements using 2-D analyses result in pseudo parameters, i.e. parameters for use
with 2-D analyses for excavations with end effects, and not fundamental soil parameters.
Currently, 3-D codes are available and desk top processors are increasingly powerful with
growing memory size. As a consequence 3-D analyses are coming into use for current
projects.
Tsim Sha Tsui Station is one of the first underground stations in Hong Kong and was opened
in 1979. Ongoing improvements to the system in 2002 included a requirement for increased
entrance capacity of Tsim Sha Tsui Station. It was proposed to extend the concourse
southwards by cut and cover construction over the existing running tunnels. The work
involved removing 9m to 10m of overburden leaving only 1.5 m of cover over twin bored
Figure 2: Layout plan showing the extension to the Tsim Sha Tsui Station concourse
constructed below busy Nathan Road
Figure 3: Typical cross section showing the ground conditions and the proposed works
The project was an unusual one. Therefore one could not rely on previous experience and
the project would have to be designed from first principles. It was complex because of the
configuration and phasing. At the planning stage, a trial analysis making a simple
conservative approximation and using 2-D analysis predicted that removal of overburden
would result in a rise of the tunnel crown of the order of 35mm, well in excess of the limit of
20mm. Additional measures, such as extensive ground improvement would have been
uneconomical. Therefore a preliminary design was carried out taking account of the phasing
and 3-D effects by adopting two orthogonal 2-D finite element models located longitudinally
and transversely. The longitudinal model was used to derive out of plane stiffness to add to
the transverse model to represent longitudinal effects. Coupling the interaction between the
models was carried out by manual iteration. This showed that predicted movements would
be reduced by taking into account the sequence of construction. Subsequently a 3-D model
was set up using FLAC 3D and the full sequence of construction was included.
Maximum lateral
deflection of about 22 mm
Z
Figure 4: View of the 3D FLAC model developed showing the lateral wall deflections along
the length of the excavation
Figure 5: Plan view from the FLAC 3D model showing the displacement of the tunnel linings
along the length of the excavation
Figure 6: Wall deflections and tunnel deformations obtained from the FLAC 3D model at a
section about 40 m from the southern end
Subsequently the strutting sequence and pre-stress loads were changed and a re-analysis was
carried out. As a consequence the predicted deflections were reduced compared to the
preliminary design. It was able to be demonstrated that deflections could be controlled
economically to within the required limits, and the project was considered feasible and could
go ahead.
The results of some of the monitoring are shown in Figures 7, 8 and 9. These show that the
vertical deformation (rise) of the tunnels was controlled to less than the Alert Limit of 10mm
and the inwards movements of the walls were less than 10mm.
It may be concluded that the modelling in 3-D was sufficiently realistic. Certainly a back
analysis exercise to calibrate the model by comparison with the monitoring data would
validate its adoption in the future as a predictive design tool.
3
5
2 6
1 7
Facing Admiralty
20
16
12
8
4
0
-4
-8
-12
-16
-20
10 Dec 02 09 Jan 03 08 Feb 03 10 M ar 03 09 Apr 03 09 M ay 03 08 Jun 03 08 Jul 03 07 Aug 03 06 Sep 03 06 Oct 03 05 Nov 03 05 Dec 03 04 Jan 04
Displacement (mm)
-30 -25 -20 -15 -10 -5 0 5 10 15 20 25 30
0.00
5.00
10.00
Depth (m)
15.00
20.00
INC01
INC02
25.00 INC03
30.00
Figure 9: Wall deflection measurements from inclinometers installed along the eastern wall
extensometers at Section Dl K at the middle of the excavation
The state-of-the-practice of geotechnical engineering in Taiwan and Hong Kong 145
Problems With Numerical Modelling
With proper use, numerical modelling is a useful tool for the designer and for the constructor.
However there are many pitfalls and mistakes are common.
Numerical modelling is often likened to use of a “Black Box”. The imagery is one of
putting numbers in and getting numbers out without understanding what goes on inside.
This can arise especially when a commercial program is provided with default values for
input parameters. In such a case the user can achieve instant gratification, the program
operates and provides an answer but the risk is that the parameters are other than those
required for the design. Intelligent users can understand how a program operates but this is
not easy where the code is not accessible and the user can only find out by trial and error.
Despite education and training, silly mistakes can he made. For example a variable soil
profile beneath a symmetrical building was modelled using an axi-symmetrical model. The
mistake was to put the axis of symmetry outside the building. Another common mistake is
to forget to zero the displacements after setting up the original stress field. Including bulk
excavation for the removal of a hill, the response of applying a modest load on a foundation
can be output as a net rise or ground heave!
It is common to use average parameters. This yields mid range predictions, not maximum
and minimum predictions. Adopting a material factor can grossly distort the results and in
some programs even small changes in parameters can greatly change the results. Reliance
on a one-off analysis without a parametric study or a good understanding of the program can
be very risky.
The one single dominant factor which bridges the Protective and Commercial considerations
is the risk arising due to uncertainty as to the actual conditions in the ground, whether it be
fundamental soil characteristics, soil/rock stratification, performance of groundwater,
foundation conditions of existing buildings and structures or obstructions etc.
Civil and geotechnical engineers are accustomed to putting to use the considerable reported
experience and increasingly advanced analytical tools to refine designs for underground
excavations. Nevertheless significant divergence of views will commonly (or invariably)
arise concerning how to attain a satisfactory balance of the protective and commercial aspects
of a design solution. Such divergent views may arise from many factors amongst which
might be;
• preferences for different analytical tools; for example variations in design actions
derived using different software for wall analysis may commonly be seen to be 10% or
more,
• views on whether disturbance due to field sampling significantly affects parameters
derived from lab testing,
• considerations of whether to adopt parameters based on an average of values derived
from site investigation or some more or less conservative basis,
• views as to whether certain parameters may themselves change as work progresses,
such as might occur as a result of gradual desiccation of soft clays in the proximity of
cementitious grout.
Undoubtedly the process toward agreeing design parameters before excavations proceed is
arduous and the outcome is rarely satisfactory to all parties concerned.
In another example of a recent project undertaken using the observational approach, it was
possible to omit the lowest layer of struts where excavation was being carried out in very soft
clay and where significant doubt existed initially about the potential effectiveness of a jet
grouted buried strut installed before excavation commenced (see Figure 10). This particular
example illustrates quite effectively how the observational method gave access to a solution
which would be unattainable in any other way, i.e. sampling of the jet grouted ground before
excavation commenced showed quite patchy results as to the degree of strength improvement
resulting from jet grouting. Based on this sampling, it was necessary to adopt conservative
Figure 10: Section through excavation and view of the desiccated marine clay
As the excavation was completed and the grouted ground exposed the reason became evident.
There was a very obvious desiccating effect of the grout which had caused the untreated
ground above, below and, more importantly, interspersed between grout patches to harden
quite dramatically. The desiccating effect was possibly also augmented by a pozzolanic
effect. Both effects would have been time dependent, hence the inability to identify the true
behaviour of the grouted soil before excavation commenced.
There are few examples other than these, and it is probably true to say that use of the
observational approach is still a rarity in Hong Kong.
The first two points are being addressed in modern instrumentation systems such as those
applied at excavations recently carried out for KCRC’s East Rail Extensions in Hong Kong.
The system employed on these projects uses an internet web based reporting interface which
is directly connected to automatically reporting devices such as load cells, strain gauges,
piezometers and the like. Every reading taken by these instruments, whether it is hourly,
daily or weekly is instantaneously uploaded to the system and available immediately for
enquiry (see Figure 11 for example). The system can be set up with predefined alert and
alarm limits that automatically trigger messages to relevant parties via email or via cell
telephone. Time lag is therefore eliminated and the on-line reports, both current and
archived, are available to all parties concerned without the possibility of manipulation thereby
giving assurance as to their reliability.
The more difficult issue to address is how to retain a satisfactory level of authority control
which gives the flexibility necessary for the observational approach to work. The following
describes how this could be achieved whilst working within the framework of the procedures
adopted for projects under the authority of the Buildings Department or other authorities such
Figure 11: Example of online report of strut load measurements, reported two hourly, showing
how trends are instantly visible (equally, anomalies would be immediately evident)
The CIRIA Guide C580 (Gaba et al 2003) defines two approaches to selection of soil
parameters for design as follows:
Approach A: Use parameters selected from a “Moderately Conservative” assessment of
available field and lab measurements.
Approach C: Use the “Most Probable” parameters defined as those which have a 50%
probability of being exceeded. These parameters are recommended only to be
used in conjunction with the Observational Method.
To date, it seems, there remains a tendency still to use “Moderately Conservative” parameters
even when the Observational Approach is to be used, perhaps because no clear strategy for
authority control is put into place. With the strategies described in outline below it should be
a straightforward matter to move on to the use of “Most Probable” values.
By using a more favourable set of parameters it is possible to show that 4 levels of strutting
would be safe if soil conditions actually were to match these parameters (see “Design B” in
Figure 12). The Observational Approach is to be used to substantiate that the 4 level scheme
can be followed, hence in the submission made for approval it is necessary to identify a test
Figure 12: Schematic diagram illustrating the two design approaches to allow the
observational approach to be applied in Hong Kong
In this case it is identified that the earliest meaningful test condition from which one could
reliably conclude, based on observations of strut load and deflection, that the more favourable
parameters were applicable is for excavation to a specified level below the third strut level.
Effectively two complete sets of design have been submitted and approved before
commencement of the work, “Design A” and “Design B”. A test condition has been defined
and approved in advance. In this case the test condition would be that observed strut loads
in the first three levels and deflections were below the values predicted for “Design B”.
Assuming that the test condition is satisfied the 5th level of struts could be omitted.
The most important aspect of this procedure, as illustrated in the very simple case above, is
that it has been unnecessary before commencement of the excavation to prove to the
approving authority that the more favourable parameters are applicable. Only the parameters
adopted for “Design A” need be accepted at that stage. Whether the more favourable
parameters are applicable, which might be considered a matter of conjecture before
commencement of excavation, becomes a matter of fact once observations have proven the
test condition to be satisfied.
The decision to follow the “Design B” route should be the responsibility of the supervising
engineer, the Registered Structural or Geotechnical Engineer in the case of private building
work. There would be no need for a formal hold point requiring the submission of a report
to the Authority for acceptance that the test condition has been met provided that the
instrumentation reporting system has be configured to provide a consistently up to date
archive of all observations to the Authority.
Conceptually, there is no need to interpret the term “Most Probable” in the way defined in
C580 as the statistical median value of parameters measured from ground investigation, as to
do so may still constrain the proponent of the design to seeking consensus from all sides on
what constitutes a valid measured parameter. In many circumstances the proponent (usually
Equally it must be acknowledged that there will be circumstances where the observational
approach must be used with more caution, for example where there are what might be termed
“brittle” elements affecting the stability of the excavation such as, for example, unfavourably
oriented relic joints in dense soils.
CONCLUSIONS
For the design of deep excavations the limit state design methodology combined with partial
safety factors applied to soil strength is preferred to global factors of safety and will lead to a
more consistently reliable and economic design.
The CIRIA C580 approach of ensuring the structural elements can withstand the ultimate
limit state design checks is also recommended. Interestingly the approach can show that the
most economical solution may not be that with the minimum wall toe penetration below the
excavation level.
It is important that groundwater conditions in the passive zone beneath the excavation is
considered and controlled especially when small toe penetrations are used.
Reviews of past failures show that most problems are caused by buildability problems being
implicit in the design and this aspect needs to addressed at an early stage of the design and
construction process.
2-D finite element numerical modelling is now tried and tested and is in regular use as a
design and construction management tool.
Sophisticated 3-D finite element numerical programs are available and can be used for
modelling more realistic geometry and sequences of construction. However sophisticated
modelling is a specialist activity and, for regular work, constitutive modelling should be kept
as simple as possible.
Where time is critical, as is often the case in the private sector projects in Hong Kong, it is
better to use a simplistic design, a conservative approach with prescriptive parameters. If the
project is unusual and tricky and the client plans well ahead then more detailed modelling
may be warranted.
There are large potential benefits from using an observational approach as the construction
proceeds. This paper outlines an example procedure showing how the observational
approach can be applied in the highly regulated environment of Hong Kong.
ACKNOWLEDGEMENTS
Information contained in this paper has been obtained by the kind permission and assistance
of the Mass Transit Railway Corporation, Hong Kong. Dr. Suraj De Silva assisted in the
preparation of the numerical analysis part of this paper.