Professional Documents
Culture Documents
net/publication/306963316
The calculation of target face pressures for TBMs in soft and mixed ground
conditions
CITATION READS
1 2,657
2 authors, including:
Nick Shirlaw
WSP-Golder
131 PUBLICATIONS 836 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Nick Shirlaw on 26 August 2016.
Yan Zhen
Shanghai Tunnel Engineering Company, Singapore
ABSTRACT: As part of Contract 855 of the Circle Line in Singapore, two slurry shield TBMs had to
be driven from One North to Farrer Road Station, passing through Buona Vista and Holland Village
Stations. The tunnels passed directly beneath a number of buildings. The ground consisted of a con-
stantly varying face of mixed soil and rock grades of weathered Jurong Formation (sedimentary rock)
and Bukit Timah Granite. The granite over part of the drives had been metamorphosed into massive
gneiss. The boundary between the rock and soil weathering grades in the Granite and Gneiss varied, but
could be abrupt. The completely weathered (grade V) granite or gneiss at the boundary was granular
and permeable in nature, while the overlying grade VI had a low permeability and was cohesive. The
tunnelling conditions therefore varied widely. The basis of the face pressure calculations is presented.
With the high groundwater levels characteristic of Singapore, there was a very small margin between
the minimum pressure required to maintain face stability and the maximum pressure required to control
the risk of slurry escape to the surface.
1 INTRODUCTION
Contract 855 of the Circle Line in Singapore includes the construction of Holland Village, Buona Vista
and Kent Ridge. One North Station was built by others between Buona Vista and Kent Ridge Stations.
Tunnels were driven using two slurry shields from One North Station, through Buona Vista and Holland
Village to Farrer Station. In the other direction, tunnels were driven using two earth pressure balance
(EPB) shields through Kent Ridge Station to West Coast Station (Figure 1).
The running tunnels between the stations are formed with 5.8m internal diameter segmental precast con-
crete lining. The two tunnels are termed the inner and outer bound respectively. Although the tunnels are
generally parallel, between Holland Village and Farrer Station the tunnels separate by up to 95m.
The slurry shields that drove from One North to Farrer Road Stations had drive lengths of 2976m (2126
rings) and 2890m (2065 rings) for the outer and inner bound respectively. The EPB drives had a length
of 1222m (873 rings).
All four of the TBMs were built by Herrenknecht AG. The slurry machines were Mixshields, which use
a bubble of compressed air to control the slurry pressure. The diameter excavated by the machines was
6.63m. All of the machines were equipped with disk cutters, for rock, and scrapers, for soil, as shown in
Figure 2. The shields were equipped with simultaneous tail void grouting systems.
Figure 1. The route of C855 Figure 2. One of the shields used for C855
This paper discusses the methods used to assess the target face pressures used for TBM operation and
during interventions into the excavation chamber. The ground and groundwater conditions will be dis-
cussed, followed by the methods used to establish the target pressures. Finally, observations will be
made on the effectiveness of the methods used, based on the results of the tunnelling work. The methods
were adapted and developed during the tunnelling, and what is presented here is the final development,
which was used for the majority of the tunnelling.
The ground conditions for the EPB drives consisted of weathered rock of the Jurong Formation.
The Jurong Formation consists of interbedded sedimentary rocks, primarily siltstone, mudstone and
sandstone. Limestone was also present in the area of Kent Ridge Station, and an intrusion of Gab-
bro/Dolorite was encountered close to the west end of Buona vista Station over a length of about 100 m
of each drive.
The weathering grades of both the Bukit Timah granite and the Jurong Formation are based on Ap-
proach 2 in BS 5930:1999, and detailed in CP2004:2003. Six weathering grades are used for classifica-
tion, with Grade I to III being rock grades and IV to VI being soil grades. The Geotechnical Interpreta-
tive Report (Babtie, 2005) further subdivided the soil grades by SPT-N value and general nature (sandy
or clayey). From that report, the recommended geotechnical design parameters for the soil grades are
given in Table 1.
Table 1. GIR recommended parameters for grades VI and V weathered rock
Source rock Grade SPT-N Description γ Cu c’ Φ’
3
kN/m kPa kPa Degrees
Granite VI <20 Sandy 20 5N 0 30
Granite VI <20 Clayey 20 5N 5 28
Granite VI 20-50 20 5N 5 30
Granite V 20-100 20 4N 5 32
Granite V >100 20 4N 10 35
Jurong VI <20 Sandy 20 5N 0 32
Jurong VI <20 Clayey 20 5N 5 25
Jurong VI 20-50 Sandy 20 5N 0 33
Jurong VI 20-50 Clayey 20 5N 10 25
Jurong V 20-100 Sandy 20 5N 5 35
Jurong V 20-100 Clayey 20 5N 15 25
Jurong V >100 Sandy 20 4N 15 35
Jurong V >100 Clayey 20 4N 20 27
The maximum Uniaxial Compressive Strength of the rock measured in the site investigation was 221
MPa for the granite and 142MPa for the Jurong Formation. In practice the Fresh Gneiss encountered
over a part of the tunnels was typically extremely strong, with a uniaxial compressive strength higher
than measured in the site investigation..
In the valleys, soft soils of the Kallang Formation and fill were present over the tunnels. The interpreta-
tion, based on the original site investigation, indicated that the minimum cover of weathered rock over
the tunnels would be about 2m, and that the tunnels would not encounter the Kallang Formation.
The cover over the crown of the tunnels varied from 8.5m to 29m, with most of the tunnelling having a
cover of 15 to 20m.
Figure 3. Retaining wall near Holland Road Figure 4. Retaining wall at Leedon Heights.
There were a limited number of piezometers installed prior to tunnelling. Typically, the spacing between
piezometers at tunnel level was between 100m and 500m. The measured groundwater levels, based on
piezometric pressure measured at tunnel level, were between ground level and 6m below ground level.
The ground water levels were lowest (relative to ground level) at the top of the hills and in the middle of
the valleys. In the middle of the valleys, the groundwater level generally reflected the conditions prior to
the placement of fill, due to the presence of drainage canals or deep drains running along the valleys.
Groundwater levels at ground level, or slightly artesian pressures, were encountered at the base of sud-
den changes in ground elevation. At the base of the retaining wall shown in Figure 2 the groundwater
level was slightly artesian with respect to the invert of the drain at the base of the wall, and at about the
level of the ground adjacent to the drain.
After the start of tunnelling an additional 51 piezometers/dewatering wells were installed between
Cornwall Gardens and Farrer Road, a distance of approximately 500m. These additional piezome-
ters/wells were installed as part of a programme of local dewatering introduced to aid tunnelling under a
number of properties, in an area where the tunnels were driven through a mixed face of extremely strong
Gneiss and completely weathered Gneiss. The additional piezometers provided a greater level of detail
on groundwater levels than was available previously.
The interpretative report for the project assigned both drained and undrained parameters for the soil
grades of the weathered Bukit Timah granite and Jurong Formation. The report divided each soil weath-
ering grade into ‘sandy’ and ‘clayey’ subdivisions. However, the gradation of the weathered rocks var-
ied such that it was not possible to predict at any point whether the face would be ‘sandy’ or ‘clayey’
material. The range of permeability from the original site investigation, 3 x 10-9 to 6x 10-7m/s. This sug-
gested that almost all of the soil grades of the weathered rocks would behave in an undrained manner
during active tunnelling. However, it was known from previous experience that:
The residual soil (Grade VI) of both the Bukit Timah granite and the Jurong Formation was
generally of low permeability, and behaved in an undrained manner during tunnelling. Some
earlier tunnelling, using open face shields or SCL methods, had been carried out in the residual
soil without a supporting pressure at the face
The majority of the completely weathered (grade V) granite and Jurong Formation behaved as
undrained materials during tunnel excavation. However, some of the grade V had a relatively
high permeability and behaved in a drained manner during tunnel excavation. In particular, the
weathered sandstone of the Jurong Formation, and the soil of the weathered granite that was
closest to the rock grades behaved like sand. Later testing, for the design of dewatering wells,
confirmed that there was a zone, typically 1m to 2m thick, just above granite rockhead. In this
zone the permeability of the weathered granite was in the range of 10-5 to 10-6m/sec, The SPT-N
value of the material in the zone was typically >100, but the material was friable and easily
eroded.
It was decided to calculate the face pressure based on both effective and total stress parameters, and
use the higher face pressure from the two sets of calculations. For the weathered soils it was as-
sessed that it was unlikely that the soil would behave in a drained manner during tunnelling unless:
For the weathered Jurong Formation: the soil was described as ‘sandy’, rather than ‘clayey’
For the weathered Bukit Timah Granite: the Grade V and IV weathering grades
For these soils the lowest friction angle from the geotechnical interpretative report was 32O. After
applying a factor of 1.2 on Tan φ’, the calculations were based on a friction angle of 28O.
For the total stress calculations, the undrained cohesion was taken as 5 x SPT-N.
The calculation methods used were only slightly different for the EPB and slurry shield tunnels; the
outline given below is for the slurry shield drives.
3.2 Ground water level
As a default value, the groundwater level in the calculations was generally set at 1.5m below ground
level. For areas at the foot of retaining walls the groundwater level was taken at ground level, except
where piezometric measurements showed a lower level. Examples of retaining walls along the alignment
are shown in Figures 4, 5 and 6, at Leedon Road, Holland Road and the crossing under the KTM (Ma-
laysian) railway, respectively. The groundwater level was taken at ground level for the first two exam-
ples, and at 4m below ground level in the last example, based on data from piezometers located at each
retaining wall.
Figure 5. The KTM railway. The sandbags show where slurry leaked
to the surface up an old borehole
3.3 Surcharge
For the total stress calculations a general surcharge of 20kPa was applied. Most of the tunnel alignment
was under roads, gardens and parking lots, and this surcharge was considered conservatively high for
these areas. The tunnels also passed under a number of buildings, including two storey houses in the
Leedon Park area, and shophouses in Holland Village. In all of the cases where the tunnelling was under
buildings the target face pressure was based on the effective stress calculations, which gave a higher
face pressure than the total stress calculations at these locations.
For the effective stress calculations no surcharge was considered in the calculations. Based on Atkinson
and Mair (1983), the factor on surcharge for a friction angle of 28O would be 0.02 and 0.06 at a cover
to diameter ratio (C/D) of 3 and 2 respectively. The impact of a surcharge of 20kPa on the target face
pressure would thus be between 0.4 kPa and 1.2kPa, which would be negligible.
For the effective stress parameters, the ground most likely to behave in a drained manner was assessed
as having a friction angle of 35O and a c’ of 10kPa. As discussed above, the friction angle of 28O adopt-
ed implied a factor of 1.2 on Tan φ’.
For the total stress parameters a maximum allowable stability number of 3 was used. For a C/D ratio of
1.5 to over 3, and a P/D ratio of 2, the stability number at failure would be between 4.5 and 6. The use
of a target N of 3 was equivalent to applying a partial factor of between 1.5 and 2 on the undrained
shear strength.
The water pressure was assessed on the basis of the highest likely value at tunnel level, and a factor of 1
was used.
Target face pressure = Pressure due to water + Pressure due to soil + Pressure due to surcharge + Al-
lowance for variation in pressure
The pressure due to water at axis level = hw.γW, where hw is the head of water at tunnel axis level and
γW. is the unit weight of water. This assumes no drainage into the chamber or (for the EPB machines) up
the screw conveyor. Little or no seepage should occur in the soil grades of the weathered rock using a
slurry machine with adequate face pressure. For the EPB shields, the effect of any drainage would be to
reduce the water pressure, but to cause a seepage pressure. Based on the charts in Anagnostu and Ko-
vari (1996), the net effect of seepage would be to reduce the water pressure by a larger value than the
seepage pressure induced, so assuming no drainage was conservative in terms of the total face pressure
required.
Where:
FO, F1, F2 and F3 are factors derived from charts in the Anagnostu and Kovari (1996), and depend on
φ´ and the ratio C/D.
γ´ is the submerged unit weight of the soil or weathered rock
D is the diameter of the TBM
Δh is the loss of head due to drainage into the excavation chamber of the TBM
As Δh was taken as 0 (no drainage), and c’ was taken as zero, the pressure reduced to FOγ´D. All of the-
se terms were constants in the calculations, with FO taken as 0.23, γ´ as 10 kN/m3 and D as 6.7m.
As discussed above, the pressure due to surcharge was considered negligible for the effective stress cal-
culations. The allowance for variation in pressure was +/-0.1 or +/-0.2 bars, depending on location.
The calculated pressure then had to be adjusted to obtain the target pressure at the control sensor, locat-
ed 2.35m above tunnel axis level. For the slurry machines, this was done by reducing the calculated
pressure from equation 1 by h1γSL, where h1 is the height of the sensor above tunnel axis level and γSL is
the unit weight of slurry. For the calculations γSL was taken as 11.5 kN/m3, an average of the unit
weight of the fresh slurry (10.4 kN/m3) and the maximum operating unit weight of the slurry (11.5
kN/m3).
Based on the terms discussed above, the target face pressure, as measured on the sensor, was derived
from the equation:
= 3.12 bars
It can be seen that the effective stress required to support the soil skeleton is the smallest term in the
equation, and represents only about 5% of the calculated pressure in this example. The water pressure is
the dominant term in the equation. This is consistent with the field measurements of volume loss over
EPB TBMs in weathered granite reported by Shirlaw et al. (2003). A face pressure slightly higher than
the in-situ water pressure resulted in low volume loss, typically <1%; conversely, a face pressure only
slightly lower than the in-situ water pressure could result in a major loss of ground.
Where γ is the average unit weight of the soil over the tunnel, ZO is the depth to tunnel axis level, q is the
surcharge, N is the desired stability number (in this case 3) and CU is the undrained shear strength. This
was then compared with the target face pressure from the effective stress analysis, and the higher of the
two values was used. Typically, the total stress calculations showed that no face pressure was needed
above an SPT-N value of about 40. The face pressure from the total stress analysis was only greater
than that from the effective stress analysis when the SPT-N value was less than about 20 blows/300mm.
Only short stretches of the tunnelling were in Grade VI (residual soil of the weathered granite or Jurong
Formation) with an SPT value lower than 20. For the majority of the tunnelling the target face pressure
was governed by the effective stress calculation.
Most of these incidents were due to a blockage developing between the excavation and plenum chambers
of the TBM. As discussed in Shirlaw & Hulme (2008), modifications were carried out to the TBM to
reduce the risk of blockages causing a sudden spike in the pressure in the excavation chamber. After
these modifications, loss of slurry to the surface was limited to locations where an old borehole or ge-
otechnical instrument was close to the tunnel, providing a direct path for the slurry to escape. Checks
were made on all known boreholes and instruments to ensure that they were sealed. However, the tun-
nels run under an old urban area, and a number of undocumented boreholes/instruments were encoun-
tered. As a result, a check was made on the target face pressure, to assess the risk of slurry being lost up
an open path. This was done by calculating:
PSL = (ZO-h1)γSL
Where PSL is the pressure required to expel slurry up an open path, and γSL is the unit weight of the
slurry. In the calculations the unit weight of the slurry was taken as 11.5 kN/m3, representing the lower
bound of the slurry density during normal operation.
If (Ptarget + v) was greater than PSL then slurry would be expelled up any open path. In practice, it was
found that slurry would always be expelled to surface up an open path in areas where the cover over the
tunnel was low or the groundwater was particularly high, i.e. at or very close to ground level. It was de-
cided that reducing the target face pressure to reduce the risk of slurry loss was not practical, as reduc-
ing the target face pressure would risk a major loss of ground. It was therefore decided to mitigate the
potential loss of slurry by maintaining a continuous watch, at the ground surface, for slurry leakage.
Measures were then taken to control the leakage as and when this occurred. In Figure 5 some sandbags
can be seen. These were placed when slurry leaked up an old borehole, which had probably been drilled
as part of the investigation for the old bridge on the left of the photograph. Containment of the slurry
prevented it from running down the slope to the adjacent railway tracks, which would have contaminated
the ballast.
Long lengths of tunnelling through extremely abrasive ground resulted in some leaks developing be-
tween the plenum and excavation chambers. These allowed compressed air, from the pressure control
bubble, to leak into the excavation chamber. If an old borehole was encountered with compressed air in
the excavation chamber, this acted like an airlift, causing a fountain of slurry at the surface (Figure 7).
This effect was driven by the compressed air, and was not related to the target operating pressure of the
machine. Mitigation measures for this effect were to seal leaks during interventions, and to bleed off
compressed air from the excavation chamber on a regular basis.
The tunnel manager used the ‘look ahead’ report as the basis for daily instructions to the TBM opera-
tors.
In these conditions wear on the cutting tools, particularly the gauge cutters, was extreme. In places the
gauge cutters were wearing down before the TBM had advanced a single ring. In the area of Cornwall
Gardens and Leedon Heights the tunnels were driven through almost continuous mixed face conditions
(Figure 9).
Figure 9. Mixed face profile between Cornwall Gardens and Leedon Heights
As a result multiple interventions in the head were required. The dewatering helped to reduce the com-
pressed air pressure during interventions and increase the factor of safety during TBM advance. The ba-
sis of the face pressure calculations was changed for the dewatered zone, due to the variability of the re-
sulting piezometric pressures.
Typically, settlements during TBM operation were in the range of 5mm to 10mm, confirming that the
face pressures set were appropriate to the ground and groundwater conditions.
Figure 10. Bent mixing paddle on one of the slurry shields
Larger settlements were recorded locally, generally where the tunnel was in mixed face conditions. The
larger settlements were particularly associated with areas where there were multiple interventions into
the excavation chamber over a short distance of advance, and/or long interventions. Long interventions,
where the TBM was halted for several days or weeks, were required occasionally for major maintenance
due to the abrasion and damage caused by driving through the mixed ground conditions. An example of
the damage sustained to one of the slurry shields is shown in Figure 10. The figure shows one of the
mixing arms, built of 40mm thick steel, bent at more than a right angle. The mixing arms were essential
to the operation of the TBM in mixed ground. The arms helped to break up nests of boulders that oth-
erwise built up in the base of the excavation chamber, blocking the passage of material into the plenum
chamber (Figure 11). As the rock crusher was located in the plenum chamber, passage of boulders
through this opening was essential to allow further advance of the machine.
During long interventions in compressed air, the filter cake of bentonite dried out and started to peel off
the ground. The exposed ground then also dried out and started to ravel. Figure 12 shows a chimney de-
veloping, in weathered granite, over the tunnel while in compressed air.
Figure 12. Chimney developing over one of the slurry TBMs dur-
ing a long intervention in compressed air
The face was occasionally sprayed with bentonite slurry to refresh it, but this did not entirely eliminate
the problem. One of the advantages of the dewatering discussed above was that the dewatering allowed
a reduction in the compressed air pressure, shortening the time required for interventions.
Many short interventions were necessary over short distances where the TBM was operating in a mixed
face of rock and weathered rock, to check and change the gauge cutters. Although these interventions
were generally short in duration, the repeated changes between slurry and compressed air support to the
face had an adverse effect on ground stability. It is difficult to control the face pressure during these
changes as accurately as during active tunnelling, and the regular removal and reintroduction of the ben-
tonite may have caused some erosion of the friable grade V granite.
The problems on the slurry shield drives, discussed above, were largely due to the extreme contrast be-
tween the extremely strong rock and the friable grade V granite when tunnelling in a mixed face. This
caused damage to the cutters and the machine and, in places, required numerous interventions in com-
pressed air, with some of the interventions lasting for several days or even weeks. Locally higher settle-
ments were recorded where multiple and/or long interventions were required. When the TBM could ad-
vance regularly, with limited interventions, surface settlements were kept to a low value (typically under
10mm), even in the extreme mixed face conditions.
The EPB drives experienced few significant problems. The sedimentary rocks of the Jurong Formation
are generally significantly weaker than those of the Bukit Timah granite. The rock/soil grade interface in
the Jurong Formation does not have a similar high permeability, friable material just above rockhead to
that found in the granite. Although regular interventions for tool changes and maintenance were re-
quired, these were much less numerous than required in the granite.
6 CONCLUSIONS
The slurry and EPB TBMs on C855 of the Circle Line in Singapore were used to drive tunnels through
varying rock and weathered rock conditions. The target face pressures were determined from simple,
published, equations that were set up in spreadsheets. Calculations were carried out at over 700 sections
along the tunnels. The spreadsheets allowed rapid updating of the calculations to reflect new infor-
mation, such as that on piezometric pressures at tunnel level. The methods used proved effective, based
on the measured ground movements. Although there were some areas of greater ground movement, these
movements were typically associated with very frequent and/or long interventions.
The spreadsheet calculations were changed significantly during the early stages of the tunnelling. In ret-
rospect, some general changes would have improved the presentation of the calculations. In particular:
The partial factors of safety could have been presented explicitly, rather than implicitly
The surcharge loads could have been refined to reflect varying conditions along the tunnel
alignment
An explicit, rather than indirect, check could have been made for uplift
In practical terms, these refinements would have had little or no effect on the calculated target pressures
set for the tunnelling. However, the refinements would have provided a clearer and more consistent basis
for the calculations.
The additional piezometers installed over the later part of the slurry shield drives proved very useful da-
ta for the key term in the calculations. The tunnels were completed successfully in other areas where
there was less information on groundwater levels and how they changed along the alignment. Where
there was limited information, conservative assumptions had to be made on groundwater levels. If there
had been shorter intervals between the piezometers, the target face pressure calculations could have been
refined.
REFERENCES
Atkinson, J.H. and Mair, R.J. (1981). Soil mechanics aspects of soft ground tunnelling. Ground Engi-
neering, July, pp20-26
Anagnostu, G. and Kovari, K. (1996). Face stability in slurry and EPB shield tunnelling. Proceedings
of the Symposium on Geotechnical Aspects of Underground Construction in Soft Ground, Lon-
don, pp 379-384.
Shirlaw, J.N. and Hulme, T.W. 2008. Risk mitigation for slurry TBMs, Tunnels and Tunnelling Inter-
national, April 2008, 26-30