Documente Academic
Documente Profesional
Documente Cultură
Instn
Civ. Engrs
in clay
Geotech. Engng,
2000, 143, Jan.,
43±55
R. J. Grant, BEng, PhD and R. N. Taylor, MA, PhD, CEng, MICE Paper 11971
Written discussion
& A series of plane strain centrifuge model x horizontal distance from the tunnel closes 31 June 2000
tests of single tunnels in moderately sti centre line in the transverse direction
clay has been conducted at City Univer- z vertical distance below the ground Manuscript received
sity, London. Ground movements were surface 16 March 1999;
measured using conventional displacement z0 depth of tunnel axis below the ground revised manuscript
transducers at the ground surface and by surface accepted 15 October
1999
digital image-processing techniques below z vertical distance above the tunnel axis;
ground level. The relatively simple tests z z0 ÿ z
are shown to result in ground movements eh horizontal ground strain
which are consistent with ®eld measure- ev vertical ground strain
ments, but the tests produced considerably o angular velocity (rad/s)
more high-quality data, which have been
analysed to improve predictions of both
surface and subsurface movements in Introduction
the plane transverse to a tunnel in clay. The worldwide use of tunnelling techniques to
The measured data are assessed against minimize the environmental impact of infra-
commonly used empirical predictive structure development has generated increased
techniques, indicating some deviations activity in the investigation of tunnelling-
from current practice. A procedure for induced ground movements. Ground response
predicting horizontal movements as a due to tunnelling is an area of ongoing
function of the vertical settlement pro®le research interest at City University, London,
is suggested. and as part of this work a series of seven plane
strain centrifuge model tests has been con-
Keywords: models (physical); subsidence; ducted to investigate the development of move-
tunnels & tunnelling ments transverse to a tunnel in moderately sti
clay.
2. Sti clay is a common tunnelling medium
Notation throughout the world as construction is rela-
D tunnel diameter tively straightforward. For example, where R. J. Grant,
g gravity possible the recent Jubilee Line extension was Richard Davies
H distance below the horizon of interest to constructed within the sti London Clay. Gen- Associates,
the level at which the projection of a erally, excavation can proceed with limited Consulting
vector of movement intersects the ver- groundwater seepage problems and the risk of Geotechnical
tical centre line of the tunnel collapse in the short term is small. However, Engineers, Bradford
H distance below the horizon of interest to some ground movement at the tunnel boundary on Avon, formerly
the intersection of the tangent to the is inevitable during the construction process. City University,
London
distribution of i with depth and the These movements propagate through the soil
vertical centre line of the tunnel mass and so can aect buildings and buried
hm height in model services.
hp height in prototype 3. While underground construction can help
i distance from the tunnel centre line to lessen the environmental impact of infrastruc-
the point of in¯exion for a Gaussian ture development it is also necessary to mini-
distribution mize the eects of construction on existing
K settlement trough width parameter structures. This includes settlement damage to
N gravity scaling factor both surface and subsurface structures which
r radius from centre of rotation exist adjacent to the tunnel and it is the
Sh horizontal ground movement assessment of such damage that has driven the R. N. Taylor,
Sv settlement need for more accurate prediction of tunnelling- Professor of
Sv max maximum settlement at the tunnel induced ground movements. Notwithstanding Geotechnical
Engineering,
centre line the fact that the presence of existing structures
Geotechnical
VL tunnel volume loss (as a percentage of will modify the development of ground move- Engineering Research
the excavated volume per metre length ments depending on the proximity of the Centre, Department
of tunnel) structures to the tunnel, it is important to of Civil Engineering,
Vs volume of the settlement trough (per understand the development of tunnelling- City University,
metre length of tunnel) induced ground movements on a green®eld site London
43
GRANT AND TAYLOR
Horizontal movements
9. From inspection of ground displacements
in centrifuge model tests by Mair, 4 O'Reilly and
New 2 suggested that the vectors of movement
near the ground surface were directed towards
the tunnel axis. This resulted in a simple
relationship between vertical ground move-
ments, Sv , and horizontal ground movements,
Sh , at the ground surface as follows:
x
Sh Sv
7
z0 In addition, recent advances in digital image Fig. 2. Principle of
and, extrapolating to the subsurface region, processing at City University have increased centrifuge model
the quantity and quality of measurement data testing
x
Sh Sv
8 that can be obtained from centrifuge model
z0 ÿ z tests. Considerable insight into ground move-
O'Reilly and New 2 correctly stated that this is ments throughout the soil depth can now be
consistent with the condition of constant achieved in plane strain models.
volume for undrained behaviour in clay (pro-
vided K is constant with depth).
10. Following the work of Mair et al., 3 Plane strain centrifuge model tests on tunnels
Taylor 5 stated that the condition of constant 12. Figure 3 illustrates the sources of move-
volume and the variation of the parameter K ment around a tunnel heading (after Mair and
with depth de®ned by equation (5) imply that Taylor 6 ). While it is true that tunnelling-
the displacement vectors focus on a point on the induced ground movements are three-dimen-
vertical centre line of the tunnel 0175z0 =0325 sional in nature, many previous investigations
below the tunnel axis. This yields an alterna- have shown that useful insight can be gained
tive distribution of horizontal movements as from two-dimensional idealizations of tunnels
follows: (e.g. Mair 4 ). With the exception of movement
towards the tunnel face, the face loss, all the
Sv x
Sh
9 sources of movement identi®ed in Fig. 3 are in
1 0175=0325z0 the plane perpendicular to the tunnel and so it
Equation (9) predicts a signi®cantly smaller is reasonable that movements transverse to
proportion of horizontal to vertical displace- long tunnels may be represented under con-
ment than equation (8) and therefore less ditions of plane strain. (Note that consolidation
potentially damaging horizontal ground strain. movements are not considered in the work
The condition of constant volume is a necessary presented.)
requirement as without it vertical and horizon- 13. To investigate the ground movements
tal displacements cannot be related. In fact, the around single long tunnels in moderately sti
constant-volume condition and distribution of clay a series of plane strain centrifuge model
K as a function of depth are not sucient tests has been conducted. A schematic diagram
without the assumption that the displacement of a typical model is presented in Fig. 4.
vectors focus on a single point. This is investi- It consisted of a block of overconsolidated
gated in detail later. kaolinite clay contained within a model con-
tainer (a strongbox). A circular tunnel cavity Fig. 3. Sources of
Centrifuge model tests lined with a latex membrane and supported by movement at a tunnel
11. Centrifuge model testing is a powerful compressed air pressure was located in the heading (after Mair
tool for investigating geotechnical problems. It clay, extending the full width of the model from and Taylor 6 )
provides a means of conducting well-controlled
1 Face loss
eective-stress-path scale-model tests using
2 Passage of the shield
real soil. The increase in inertial acceleration at 3 Closure of the tail void
N times gravity results in stress similitude with 2
1 4 Lining deflection
2
5
5
a prototype of linear dimensions N times that of 3
5 Consolidation 3
the model. The principle is illustrated in Fig. 2.
It is relatively simple to vary the geometry of
the scale model to explore a variety of proto-
type situations and this leads to substantial 4 4
bene®ts over ®eld measurements in terms of
cost, time, variability and experimental control.
45
GRANT AND TAYLOR
Up to 304 mm
front face of the model in order to determine
subsurface patterns of movement; this is Marker beads
described brie¯y in the following section. A
total of seven tests were conducted and the
geometric details of each are given in Fig. 4 and
m
m
Table 1. The test geometries were chosen to
0
20
minimize the eects of the side wall boundaries.
550 mm
All the tests were conducted at an acceleration
of 100 times gravity (the gravity scaling factor
N was 100), at which the tunnel cavity of Pore pressure Tunnel cavity (50 mm dia.)
diameter D 50 mm represented a 5 m dia. transducers supported by compressed air pressure
prototype tunnel with the depth to the tunnel
axis varying from 10 m to 22´5 m. Fig. 4. Schematic diagram of a typical plane strain centrifuge model (not
to scale)
The centrifuge facility at City University,
London
14. The centrifuge facility has an Acutronic Table 1. Centrifuge model tests conducted*
661 geotechnical centrifuge with a radius to the
platform of 1´8 m and the capacity to test Test reference Total depth to tunnel axis
models weighing up to 200 kg at 200g. The
z0 (model scale): mm z 0 /D
normal payload volume is approximately
500 mm by 700 mm by 500 mm high. Fig. 5 is a RJG1 225 4´5
schematic diagram of the centrifuge and control RJG9 225 4´5
room. The model swing is shown at rest but as RJG14 100 2
the acceleration increases it rotates upwards in RJG15 175 3´5
line with the radial acceleration ®eld. Clearly, RJG17 150 3
the model apparatus must be designed to TH1 175 3´5
operate remotely and under high accelerations. CK1 125 2´5
15. An important feature for this research is * RJG1 was a preliminary test in which equilibrium conditions were not achieved
the digital image-processing capability. A small at 100g before reducing the tunnel pressure. In the other tests the water table was
solid-state charge-coupled device (CCD) camera 25 mm below ground level, except for TH1, in which it was 5 mm below ground
mounted on the centrifuge swing views the level. Measurement from image-processing techniques was successful from test
front face of the models in ¯ight through the RJG15 onwards.
Test procedure
16. Preparation of the models began by
mixing kaolinite clay with water to form a
slurry with a moisture content of approximately
120%. The clay slurry was then placed inside a
model container and one-dimensionally precon-
solidated in a press to produce a vertical
eective stress of 500 kPa throughout the soil.
Once this had been achieved the sample was (b)
However, in other soils such as natural clays 180 Ground surface for RJG15 and TH1
collapse may occur very suddenly and at
signi®cantly lower volume losses. In such 160
Ground surface for RJG17
cases, the range of precollapse volume losses
over which the trough widths were constant
tion (5) implies that the ground movement of horizontal movement, and these are compared Fig. 18. Vertical and
vectors focus on the point where the distribu- in Fig. 18 with normalized horizontal move- horizontal movement
tion of i with depth intersects the vertical ments at various horizons and for a range of at dierent subsurface
centre line of the tunnel. Horizontal ground volume losses in test RJG17. The vertical move- levels for test RJG17
movements can then be described as a function ments are also shown, with the best-®t Gaussian with best-®t curves:
of the vertical movements as in equation (9). curves. In the near-surface region the horizontal (a) 10 mm below
Assuming constant-volume conditions, for a movements are not well described by assuming ground level (140 mm
non-linear distribution of i with depth it can be an average vector focus but at all other subsur- above tunnel axis);
shown that if a vector focus exists for each face horizons shown the agreement is very good. (b) 30 mm below
horizon it will also be at the point where the 28. Thus there is evidence that assuming a ground level (120 mm
tangent to the distribution of i with depth vector focus at individual horizons may be an above tunnel axis);
intersects the vertical centre line of the tunnel; appropriate way of predicting horizontal move- (c) 70 mm below
a detailed derivation is given in Appendix 1. ments. In Fig. 19 the data from test RJG17 are ground level (80 mm
The idea is illustrated in Fig. 17 and shows how compared with predictions based on equations above tunnel axis);
the vector focus may vary with depth. (5) and (9), which are shown by the solid lines. (d) 100 mm below
27. Calculating the position of the projection The vertical movements are well predicted at ground level (50 mm
of ground movement vectors from the tests onto all horizons. The horizontal movements tend to above tunnel axis)
the vertical centre line of the tunnel produced be overestimated but, with the exception of the
considerable scatter at any horizon. No trend in near-surface horizon, the predictions are
the position was apparent except in the region reasonable and probably more than adequate
close to the free ground surface, where the for most design situations.
position of the vectors projected onto the verti- 29. It should be noted that in the tests
cal centre line of the tunnel tended to move up described there was no constraint on the upper
with increasing horizontal distance from the soil surface and it has therefore been referred to
tunnel axis, that is, the proportion of horizontal as a free ground surface. It is likely that the
to vertical movements increased with x. deviation from the linear distribution of i with
However, the average position of the vector depth given by equation (5) in the vicinity of
focus can be used to give average distributions the soil surface is associated with this free
52
TUNNELLING-INDUCED GROUND
MOVEMENTS IN CLAY
ground surface. In urban areas such a situation conditions apply, the movements must Fig. 19. Vertical and
is rare and it is possible that even the presence focus on the intersection of the tangent to horizontal movement
of a thin pavement layer may provide sucient the distribution of i with depth and the at dierent subsurface
restraint to negate the free-surface eect. vertical centre line of the tunnel levels for test RJG17
(d ) assuming an average vector focus at any with predicted curves:
Conclusions given horizon results in a good prediction (a) 10 mm below
30. The centrifuge model tests described of horizontal movements despite there ground level (140 mm
have provided a considerable body of data to being considerable scatter in the back- above tunnel axis);
examine the patterns of precollapse ground analysed location of the focal point (b) 30 mm below
movements caused by tunnelling in clay. The (e) except in the vicinity of a free ground ground level (120 mm
empirical relations currently used for predic- surface, adequate predictions of horizontal above tunnel axis);
tions have been examined against the compre- movements can be made by using the linear (c) 70 mm below
hensive measurements and their limitations distribution of i with depth described by ground level (80 mm
highlighted. The main ®ndings are Mair et al., 3 equation (5), and assuming that above tunnel axis);
the ground movement vectors focus at the (d) 100 mm below
(a) both surface and subsurface settlement point where the tangent to the distribution ground level (50 mm
troughs are well represented by Gaussian of i with depth intersects the vertical centre above tunnel axis)
distributions except within about 05D of line of the tunnel, equation (9).
the tunnel crown
(b) the distribution of i with depth suggested
by Mair et al., 3 equation (5), adequately Appendix 1. Focus of vectors of ground
predicts the distribution of vertical move- movement due to tunnelling
ments except within 05D of the tunnel 31. The following analysis assumes that
crown, where steeper troughs may be constant-volume conditions apply, which is
anticipated (see also (a) above), or within consistent with the development of short-term
the vicinity of a free ground surface, where movements in clay. The geometric terms used
wider troughs may be anticipated are de®ned in Fig. 20. It is convenient to
(c) if a focal point for ground movements rede®ne the distance from the tunnel axis to the
exists at any horizon and constant-volume horizon of interest as z ; relating this to the
53
GRANT AND TAYLOR
ev e h 0
14 z z dK
1
24
H K dz
The vertical strain in the ground is given by
34. As shown in Fig. 21, i varies with depth
dSv and the distribution may be considered to be
ev
15
dz linear except in the vicinity of the tunnel or a
Noting that K f
z , we obtain from equation free surface. The tangent to the distribution of i
(11) with depth intersects the vertical centre line of
the tunnel at a distance H below the horizon of
dSv Vs ÿx2 1 1 dK interest. Since i Kz and K f
z ,
p exp ÿ ÿ
dz
2p 2K 2 z2 Kz2 K 2 z dz
2
2 di dK
1 ÿx 2x 2x2 dK K z
25
exp dz dz
Kz 2K 2 z2 2K 2 z3 2z2 K 3 dz
and
16
so 1 di z dK
1
26
Vs ÿx2 1 K dz K dz
ev p exp
2p 2K 2 z2 Kz2 Therefore, substituting equation (24),
2
x x2 dK z dK z 1 di
ÿ 1 ÿ
17
27
K 2 z2 K 3 z dz K dz H K dz
54
TUNNELLING-INDUCED GROUND
MOVEMENTS IN CLAY
Free surface Centre line was developed for the centrifuge largely by
members of the Engineering Surveying
Research Centre at City University. This was
also funded by the EPSRC, under grant award
i GR/J74022. Particular thanks are due to Dr
Linear Stuart Robson, Dr Jin Chen and Professor Mike
region Cooper.
38. Some of the tests reported were con-
z
* ducted by visitors and students at the Geo-
technical Engineering Research Centre at
City University. The eorts of Dr Toshiyuki
H
* Hagiwara and Caesar Kerali are gratefully
A distribution of i
with depth acknowledged.
References
1. P E C K R. B. Deep excavations and tunnelling in
soft ground. Proceedings of the 7th International
x
Conference on Soil Mechanics and Foundation
Engineering, Mexico, 1969, state of the art
Fig. 21. De®nitions for the intersection of volume, 225±290.
tangents to a distribution of i with depth and the 2. O'R E I L L Y M. P. and N E W B. M. Settlements above
vertical centre line of the tunnel tunnels in the United KingdomÐtheir magnitude
and prediction. Proceedings of Tunnelling '82
Symposium, London, 1982, 173±181.
or 3. M A I R R. J., T A Y L O R R. N. and B R A C E G I R D L E A.
Subsurface settlement pro®les above tunnels in
di Kz i clays. GeÂotechnique, 1993, 43, No. 2, 315±320.
28
dz H H 4. M A I R R. J. Centrifugal Modelling of Tunnel Con-
so at any particular z , H H . struction in Soft Clay. PhD thesis, University of
Cambridge, 1979.
35. Thus, assuming that the vertical ground
5. T A Y L O R R. N. Tunnelling in soft ground in the
settlement pro®les are of Gaussian form, that UK. Proceedings of the International Symposium
constant-volume conditions apply and that the on Underground Construction in Soft Ground
vectors of ground movement at a given horizon (Fujita K. and Kusakabe O. (eds)). Balkema,
above the tunnel focus on a single point on the Rotterdam, 1995, 123±126.
vertical centre line of the tunnel, the point of 6. M A I R R. J. and T A Y L O R R. N. Bored tunnelling in
focus lies at the intersection of the tangent to the urban environment. Proceedings of the 14th
the distribution of i with the vertical centre line International Conference on Soil Mechanics and
of the tunnel. Foundation Engineering, Hamburg, 1997, 4,
Theme Lecture.
7. T A Y L O R R. N., R O B S O N S., G R A N T R. J. and
Acknowledgements
K U W A N O J. An image analysis system for deter-
36. Most of the work described was funded mining plane and 3-D displacements in centrifuge
by the Engineering and Physical Sciences models. Proceedings of the International Confer-
Research Council (EPSRC) under grant award ence Centrifuge 98 (Kimura T., Kusakabe O. and
GR/J53010. Takemura J. (eds)). Balkema, Rotterdam, 1998,
37. The digital image-processing system 73±78.
55