Shin, J. H., Addenbrooke, T. I. & Potts, D. M. (2002). Géotechnique 52, No.

6, 391–403

A numerical study of the effect of groundwater movement on long-term

tunnel behaviour
J. H . S H I N
, T. I . A D D E N B RO O K E { a n d D. M . P OT T S {

Soft ground tunnelling beneath the water table induces La construction de tunnels dans des sols aquifères pro-
pore water pressure changes and alters the hydraulic duit des changements dans la pression de l’eau intersti-
regime in the ground. There are various factors that tielle et change le régime hydraulique dans le sol. Divers
control the tunnel/soil interaction behaviour during the facteurs contrôlent l’interaction entre le tunnel et le sol
equilibration period after tunnel construction through to pendant la période d’équilibrage après la construction du
the long-term steady state. Through the presentation of tunnel et jusqu’à ce que celui-ci parvienne à un état
results from coupled non-linear finite element analyses, stable à long terme. En présentant les résultats obtenus à
this paper considers the differences in predicted ground partir d’analyses couplées non linéaires d’éléments finis,
and lining responses in the long term for different post- cet exposé considère les différences qui se manifestent à
tunnelling flow conditions. A 4 m diameter circular tun- long terme dans les réponses prédites du sol et des parois
nel constructed 20 m below ground level is adopted for pour diverses conditions d’écoulement après la construc-
the geometry, with the soil conditions based on those tion du tunnel. Nous avons adopté pour la géométrie un
found in London, UK, and Seoul, South Korea. The tunnel circulaire ayant un diamètre de 4 m et construit à
lining permeability is varied from fully permeable to 20 m en dessous du niveau du sol, les conditions étant
impermeable, and an approach to modelling finite lining celles qui existent à Londres en Angleterre et à Séoul en
permeability is presented and assessed. For the per- Corée du sud. La perméabilité des parois peut aller de
meable lining analyses two flow regimes are modelled, totalement perméable à imperméable et nous présentons
one in which the water table is maintained at its original et évaluons une méthode permettant de mettre en ma-
elevation and one in which the water table is drawn down quette la perméabilité finie de la paroi. Pour les analyses
towards the tunnel. The importance of giving careful de paroi perméable, nous avons mis en maquette des
consideration to lining permeability and hydraulic bound- régimes d’écoulement confinés et non confinés. Nous
ary conditions is highlighted. The effects of groundwater soulignons l’importance de tenir compte de la perméabi-
movements on numerical predictions of tunnel behaviour lité des parois et des conditions de limite hydraulique.
are also identified. Nous identifions également les effets généraux des mouve-
ments des nappes d’eau souterraine sur les prévisions
KEYWORDS: numerical modelling and analysis; tunnels; numériques du comportement d’un tunnel.
groundwater

INTRODUCTION and mechanical behaviour, which investigate the impact of

Soft ground tunnelling beneath the water table induces pore the controlling factors on numerical predictions of long-term
water pressure changes and alters the hydraulic regime in tunnel lining behaviour. The variants considered are:
the ground. There are various factors that control the tunnel/
(a) fully permeable and impermeable linings, and linings
soil interaction behaviour during the equilibration period
with finite permeability
after tunnel construction through to the long-term steady
(b) flow regimes that do and do not draw down the phreatic
state. The dominant factors are the permeability of the soil
surface (referred to as maintained phreatic surface and
and the boundary conditions to the new flow regime. At the
drawdown flow regimes respectively)
tunnel boundary itself the permeability of the lining will
(c) low-permeability and high-permeability soil profiles.
govern the flow regime. Above the tunnel the soil permeabil-
ity and the surface recharge conditions will govern the It is expected that, in clays, there will be a gradual
response of the phreatic surface. (The phreatic surface is the increase in earth pressure on a tunnel lining in the long
locus of points at zero pore water pressure, which when term. Guidelines for tunnel lining design have pointed out
horizontal is often termed the water table.) After setting out that the design procedures should include provision for the
the need for a sophisticated numerical investigation of this time dependence of soil behaviour (e.g. ASCE, 1984).
engineering problem, the paper presents methods for model- Decisions on the long-term lining loads to be used in design
ling the controlling factors in a finite element analysis and will depend on the life expectancy of the support system,
introduces a novel approach to modelling finite lining per- the permeability of the soil, and the new hydraulic regime.
meability. The paper goes on to present the results from In this paper, ground loading and lining loads and their
non-linear finite element analyses with coupled hydraulic dependence on the hydraulic regime are the main considera-
tion. On the basis of field measurements Peck (1969), Mair
& Taylor (1997) and Mair (1998) indicated that, in the long
Manuscript received 22 February 2001; revised manuscript accepted term, tunnelling-induced surface settlement troughs become
26 March 2002.
deeper and wider than those initially generated in the short
Discussion on this paper closes 2 March 2003, for further details
see p. ii. term. There is no evidence that this deepening and widening

Seoul Metropolitan Government of Korea; formerly Imperial is detrimental to overlying structures (New & O’Reilly,
College of Science, Technology and Medicine, London, UK. 1991), and it is not considered in this paper.
{ Imperial College of Science, Technology and Medicine, London, The timescales required for field-based research seriously
UK. limit the number of publications presenting measurements of

391
392 SHIN, ADDENBROOKE AND POTTS
long-term behaviour. One significant publication is that of Pore water pressure: kPa
0 100 200
Barratt et al. (1994). They presented long-term monitoring 0
of the earth pressure (ground loading) on, and the hoop Water table
thrust in, an expanded concrete lining installed on the
Jubilee Line beneath Regent’s Park in London in the 1970s.
The geometry and soil profile there are very similar to those Thames gravel
10
modelled in the analyses presented in this paper, and the London clay
data are presented for comparison with the numerical predic-

Z: m
tions in the section ‘Lining performance’ below. Two other
relevant publications are those of Ward & Pender (1981), Linear
20
who have reported pore pressure measurements that indi- permeability
cated London Underground tunnels in London clay acting as Hydrostatic
drains, and Bowers et al. (1996), who have shown that far-field profile
surface settlements increase between the short and long term Non-linear
above tunnels in London clay, which again indicates that the 30
permeability
tunnels in question were acting as drains.
In addition to the problems with field monitoring, the 4 m diameter tunnel
complexity of the problem means that there is little pub- with permeable lining
lished research into the analysis of long-term tunnel behav-
iour. For example, no publications have been unearthed that Fig. 1. Long-term pore water pressure profiles down axis line
analyse drawdown flow regimes coupled with the mechanical (Figure 4.55 of Addenbrooke, 1996)
behaviour of the ground and tunnel lining. Atkinson & Mair
(1983) provide an analytical method for determining the
pore pressure on a lining. Their analysis assumes that the
tunnel is deep enough to induce radial flow, that the phreatic the long term over much of the profile, and increases the
surface remains unaffected, and that no significant soil strain hydraulic gradient close to the tunnel.
occurs close to the tunnel. Numerical modelling of the For the analyses in this paper the log law permeability
problem (e.g. Gunn & Taylor, 1984; Atwa & Leca, 1994; model proposed by Vaughan (1989) was employed. The
Schweiger et al., 1997) has invariably been steady-state coefficient of permeability, k (which is subsequently referred
seepage analysis or sequential seepage analysis and stress to as the permeability), is defined as
analysis employing a constant permeability. With these ap-
proaches the full-time dependent lining/soil interactions fol- k ¼ k o e Bp9 (1)
lowing tunnel construction cannot be described, and the
dependence of permeability on stress level cannot be mod- where k o is the permeability at zero mean effective stress,
elled. It is therefore appropriate that soft ground tunnelling p9 is the mean effective stress, and B is a material property
beneath the phreatic surface be treated as a complex non- with the units m2 =kN. (This model assumes that the soil’s
linear problem, and be handled using coupled numerical volume compressibility is constant.)
analysis. Modelling of a drawdown flow regime must permit the
phreatic surface to move through the soil. This was achieved
using a model that produces a reduction in permeability
NUMERICAL MODELLING OF GROUNDWATER FLOW (orders of magnitude) as suction increases above a pre-
Permeability model for soil scribed value. This model is referred to as the suction switch
Permeability is the critical soil parameter governing the in this paper. The suction value at which the permeability
post-construction behaviour. Coupled finite element analyses begins to reduce can be equated to the air entry value of the
require soil permeability to be defined. It is usual to assume soil. The use of such a model for the analysis of drawdown
that soil permeability remains constant through an analysis. problems is described by Hsi & Small (1992) and Hsi et al.
However, it is known that permeability is dependent on void (1994). In the analyses in this paper, no account was taken
ratio, which is dependent on mean effective stress, and of the effect of desaturation if suctions rose above the air
hence seepage flow is non-linear (Lambe & Whitman, 1969; entry value.
Vaughan, 1989). Vaughan (1989) pointed out that, when
there are significant changes in effective stress along flow
paths, the variation of permeability with effective stress has Hydraulic boundary conditions
a dominant effect on the distribution of pore water pressure This section considers two of the key boundaries control-
at steady state. He indicated that predictions based on the ling the long-term behaviour of tunnels: the tunnel lining
conventional assumption of a constant permeability are itself, and the phreatic surface.
usually far from reality, and one of the main causes of Once complete, a tunnel excavation can act as a drain and
discrepancy is the effect of non-linearity in permeability. allow groundwater to flow in. The most extreme case is if
During the equilibration period after tunnelling the effective the tunnel lining is fully permeable, and provides no resis-
stresses are likely to be changing in the ground. It is there- tance to inflow. In such a case the rate of inflow is governed
fore appropriate to use a non-linear permeability model that by the permeability of the surrounding soil. It is possible,
allows the permeability to vary during an analysis in however, that the lining may provide some resistance to flow.
response to changes in effective stress. Addenbrooke (1996) In such a case the inflow and the pore water pressure acting
compared the long-term steady-state pore water pressure on the lining are governed by the relative permeability of
profiles above a 34 m deep permeable tunnel predicted with the lining to the soil. The third alternative is that the lining
an isotropic linear (permeability reducing with depth) and an is effectively impermeable and the tunnel boundary is a no-
isotropic non-linear permeability model (permeability vary- flow boundary.
ing with mean effective stress as defined in the next Hydraulic conditions in the ground are broadly governed
paragraph). Fig. 1 presents that comparison. The non-linear by the climate and the hydrological environment. In London,
model causes pore pressures to recover to near hydrostatic in buildings and roads usually intercept rainfall and prevent
 EFFECT OF GROUNDWATER MOVEMENT ON LONG-TERM TUNNEL BEHAVIOUR 393
evaporation, so eliminating most of the cyclic and transient than zero (i.e. suction) the boundary was maintained as a
effects associated with climate. Invariably the level of the no-flow boundary (q ¼ 0) at that location. This removed the
phreatic surface in the upper aquifer above the London clay unrealistic possibility of soil drawing water from within the
is governed by leaking water mains or local rivers. Geotech- tunnel.
nical analyses of urban situations are usually simplified to A lining with a finite permeability was modelled through
monotonic consolidation problems without attention to the introduction of a novel scheme combining structural
groundwater fluctuations in response to river levels or other beam elements, which give direct solutions for lining distor-
cyclic events. If the lining is permeable, and there is a tions, forces and moments, and thin quadrilateral solid
steady long-term inflow to the tunnel, there are two possible elements, which can have a prescribed permeability. The
flow regimes. If there is sufficient recharge the phreatic arrangement is shown in Fig. 3. Prior to lining installation
surface can be maintained at a roughly constant level. these solid elements behave like soil, but after the activation
Alternatively, if there is insufficient recharge to balance the of lining elements their permeability is changed to corre-
amount of water flowing into an excavation then the phreatic spond to the lining permeability, which is then kept constant
surface is lowered and a drawdown regime exists. This latter throughout the analyses into the long term. (Only the corner
behaviour can be expected for soils with a high void ratio nodes on the solid elements had pore water pressure degrees
and a low capacity for storing water in the voids, where the of freedom.) UMTA (1986) reported that the major paths of
permeability is too high and the recharge too low to main- water infiltration into a concrete tunnel lining are shrinkage
tain the tunnel inflow without a drawdown of the phreatic cracks in the lining, construction joints (where water-stop
surface. This effect is shown schematically in Fig. 2, in systems do not work or are not used), and channels and
which the depth of the tunnel crown beneath the initial ducts in concrete as the results of concrete deterioration and
phreatic surface is defined as H d and the depth of the tunnel structural failure. Similarly, the major paths of water infiltra-
axis beneath the ground surface as Z o . The consequence of tion into bolted segmental cast iron or precast concrete
drawdown will be a reduction in flow rate into the tunnel linings are through the joints, bolt holes, and any lining
with time. This trend will occur in association with any cracks or imperfections. Consequently, the evaluation of
changes in flow rate due to the non-linear model for per- lining permeability is very difficult, and has rarely been
meability. If the tunnel lining is impermeable, excess pore reported. The rate of flow into the tunnel and pore water
water pressures generated close to the tunnel must dissipate pressure on the lining will be dependent on the relative
to more remote drainage boundaries defined by free-draining permeability of the lining to the adjacent soil, so the rela-
geological strata or the ground surface. tive permeability of a lining, k l , to the soil, k s , is more
Consequently three cases of analysis can be considered as important.
practically possible situations:
(a) a permeable tunnel lining with a maintained phreatic
FINITE ELEMENT ANALYSIS OF LONG-TERM TUNNEL
surface flow regime
BEHAVIOUR
(b) a permeable tunnel lining with a drawdown flow regime
The Imperial College Finite Element Program (ICFEP)
(c) an impermeable tunnel lining with a maintained
was used for all the analyses presented in this paper. Eight-
phreatic surface flow regime.
node plane-strain isoparametric elements with reduced inte-
An impermeable lining is modelled by prescribing the gration were used to represent the soil. Three-node Mindlin
flow rate as on the excavation boundary zero, q ¼ 0, beam elements with selected reduced integration were used
throughout the analysis. A fully permeable lining is mod- to model the tunnel linings (Day & Potts, 1990). The beam
elled by prescribing the pore water pressure, p, on the elements were directly connected to the soil elements, so no
excavation boundary as zero: that is, p ¼ 0 throughout the slippage could occur between the two. An accelerated mod-
long-term analysis. However, the use of such a boundary ified Newton–Raphson scheme with a sub-stepping stress-
condition can be unrealistic in situations where, immediately point algorithm was employed to solve the non-linear finite
after tunnel excavation, suctions exist in the soil adjacent to element equations. A coupled consolidation formulation was
the tunnel. Prescribing p ¼ 0 in such a case will allow flow employed throughout.
from the tunnel into the soil. Consequently, to overcome this
shortcoming, a special boundary condition was used for
these analyses that monitored the pore water pressures in the
soil immediately adjacent to the tunnel. If the pore water
pressure in the soil at any point round the boundary was less

Phreatic surface

Zo Hd

Tunnel

Fig. 2. Flows into the tunnel. Z o , depth to axis; Hd , depth of

crown below phreatic surface Fig. 3. Modelling of a lining with finite permeability
394 SHIN, ADDENBROOKE AND POTTS
Analysis details

20 m
Figure 4 shows the soil profiles adopted for the two

50 m
permeability scenarios. Fig. 4(a) presents a typical profile
for London, with Thames gravel deposits overlying the
London clay, which in turn overlies the Lambeth Group clay
and sand. Fig. 4(b) shows a profile typical encountered in
Seoul during construction of Seoul Subway lines 3 and 4 200 m
(Shin & Yoo, 1985). In both cases the tunnel geometry was (a)
circular with diameter D ¼ 4 m, and the axis level was at a

0·3 m
depth Z o ¼ 20 m. The material properties given in Fig. 4
are bulk unit weight, ªt , initial coefficient of earth pressure
at rest, K o , strength parameters for the Mohr–Coulomb

4m
model (see the section ‘Material models’ below) of cohesion
c9, angle of shearing resistance 9, and angle of dilation ł9,
and permeability model parameters k o and B (see ‘Per-
meability model for soil’ above and ‘Material models’ (b)
below).
Figure 5 presents the finite element mesh used to repre- Fig. 5. (a) Finite element mesh; (b) mesh detail of tunnel
sent the London profile. There was a slight modification to
the mesh for the Seoul profile to correctly match the
material boundaries with element boundaries.
Initial excavation of the tunnel was modelled so as to analyses were carried out for just under 80 000 days (about
obtain relevant conditions for the beginning of the long-term 200 years) post-construction, and the Seoul analyses for
equilibration period. The approach taken was to excavate 500 days post-construction.
over a number of increments unlined so as to induce ground
movements representative of those typically measured in
London and Seoul, and then to complete excavation with the Boundary conditions
lining constructed. This process introduces thrusts and bend- Hydraulic conditions must be defined along all the bound-
ing moments into the tunnel lining before the end of aries to consolidating materials in a coupled analysis. The
excavation. The initial stresses in the soil prior to excavation vertical plane of symmetry through the tunnel’s axis was
were defined by the bulk unit weight of the soil and the K o prescribed as a no-flow boundary. For the maintained phrea-
ratio given in Fig. 4, with a hydrostatic pore water pressure tic surface analyses the pore pressures at the top of the
profile below a water table 2·5 m below the ground surface. London clay and at the top of the decomposed granite soil
Tunnel excavation was modelled over a number of incre- were fixed as equal to the initial values throughout the
ments representing a time of 15 days. The area of the analysis. Similarly the remote vertical boundary 200 m away
transverse surface settlement profile was controlled by the from the tunnel and the deep horizontal boundary 50 m
construction of the lining at the relevant increment (once below the ground surface were also modelled as constant
the lining is constructed there can be no further volume loss, pore water pressure boundaries defined by the initial values.
and any remaining unloading due to excavation is transferred For the drawdown analyses, to simulate the effect of insuffi-
into lining loads). The target volume loss was 2% of the cient recharge to maintain the phreatic surface, a horizontal
cross-sectional area of the tunnel. During excavation the hydraulic ‘cut-off’ (i.e. no-flow boundary) was modelled at
tunnel boundary was modelled as a no-flow boundary. This the level of the original water table at 2·5 m below ground
is reasonable for an open excavation in a very low per- level. This meant that water could be drawn into the solution
meability clay, but not representative of drained excavation domain only across the right-hand side and the bottom
in more permeable soil. From an analytical point of view, boundaries.
the purpose of preventing inflow during excavation was to During tunnel excavation the tunnel boundary was pre-
distinguish between the response during excavation itself and scribed to be a no-flow boundary. During the equilibration
the post-construction response to pore pressure equilibration period to the long term two alternative boundary conditions
to the new hydraulic boundary conditions. The London were prescribed. For the condition of a fully permeable or

EL–00·0 EL–00·0
Thames gravels (γt = 20 kN/m3, Ko = 0·5, φ = 35˚, ψ = 17·5˚) EL –02·5 Fill/alluvium (γt = 15·7 kN/m3, Ko = 0·5)
Z Z
EL–05·0 EL–05·0
Decomposed granite soil
Z1 (γ = 20·6 kN/m3, K = 0·45,
20 m

t o
c = 2 + 0·667Z1 kPa, φ = 23 + 0·622Z1,
London clay (γt = 20 kN/m3, ψ = 15·5˚, ko = 1·9 × 10–6 m/s,
Ko = 1·4, 5 < Z < 15 m B = 0·0043)
Ko = varies linearly from 1·4 to 1·1, 15 < Z < 25 m EL–20·0
D=4m Ko = 1·1, 25 < Z < 35 m,
50 m

Highly to moderately weathered granite

c′ = 5 kPa, φ′ = 23˚, ψ′ = 0˚ Z2 (γt = 20·6 kN/m3, Ko = 0·45,
ko = 1·0 × 10–10 m/s, B = 0·007) c = 10 + 7·5Z2 kPa, φ = 42˚,
ψ = 21˚, ko = 1·0 × 10–6 m/s,
B = 0·0043)
EL–35·0 EL–35·0
Lambeth Group clay (γt = 20 kN/m3, Ko = 1·0, c = 25 kPa, φ = 27˚, Slightly weathered to unweathered granite
ψ = 0˚ ko = 1·0 × 10–10 m/s, B = 0·007) Z3 (γ = 20·6 kN/m3, K = 0·3,
EL–40·0 t o
c = 100 + 500Z3 kPa, φ = 56˚, ψ = 28˚,
Lambeth Group sand (γt = 20 kN/m3, Ko = 1·0, ko = 3·5 × 10–7 – 0·139 × 10–7m/s,
c = 0 kPa, φ = 34˚, ψ = 17·5˚) B = 0·0043)
EL–50·0 EL–50·0

(a) (b)

Fig. 4. Soil profiles: (a) London clay (London, UK); (b) decomposed granite soil (Seoul, South Korea)
 EFFECT OF GROUNDWATER MOVEMENT ON LONG-TERM TUNNEL BEHAVIOUR 395
finite permeable lining a pore water pressure of zero was a predicted downward movement of the phreatic surface in
prescribed along the boundary, and flow into the tunnel was response to tunnel drainage.
permitted. (The special boundary condition described above For the analyses in which the tunnel lining had a finite
in ‘Hydraulic boundary conditions’ was used.) For the con- permeability, k l , this was defined through a ratio k l =k s ,
dition of a fully impermeable boundary no flow was per- where k s is the initial soil permeability at the tunnel level.
mitted across the tunnel boundary. The influence of finite permeability is demonstrated in this
The displacement boundary conditions permitted: no hor- paper through the results from analysis of the London
izontal displacement down the two vertical mesh boundaries; profile, for which this ratio was equal to 0·1.
no vertical or horizontal displacement along the bottom
boundary of the mesh; and freedom to displace along the
top boundary (ground surface) and the tunnel boundary. RESULTS
The nodes of the tunnel lining beam elements that lie on the Flow and ground behaviour during the equilibration period
vertical symmetry axis were not permitted to rotate. are considered in terms of pore pressures and soil stresses,
lining pressures and loads, and the surface settlement. Lining
performance is considered in terms of ground loading and
hoop thrust. Fig. 6 defines the ground loading,  n , and the
Material models
hoop thrust, H f , and presents them as percentages of the
Mechanical models. The soil strata for both the London
overburden stress at axis level,  vo , and the overburden load,
and the Seoul profiles were modelled as non-linear elastic
Fo , calculated from  vo . Bending moments are not consid-
perfectly plastic. The non-linear elastic model employed was
ered as they show little variation with time.
based on that described by Jardine et al. (1986). It defines a
shear stiffness to mean effective stress ratio that decays with
shear strain, and a bulk stiffness to mean effective stress ratio
London profile
that decays with volumetric strain. (The effect of stress path
Flow and ground behaviour. Figure 7 compares the pore
reversals is not accounted for in this model.) The relevant
water pressure distributions at the end of construction, and in
equations and the parameters used are given in Appendix 1.
the long term for both maintained phreatic surface and
The perfect plasticity was modelled using a non-associated
drawdown flow regimes. Defining the zone of influence as the
Mohr–Coulomb model. The yield surface is defined by the
lateral extent of any pore pressure variation from hydrostatic,
strength parameters cohesion, c9, and angle of shearing
it is evident that the zone of influence on completion of
resistance, 9, and the plastic potential is defined by an angle
excavation is much smaller than in the long term. Comparing
of dilation, ł9. The parameters, based on the authors’
the two long-term profiles, the zone of influence has an
experience of analyses of engineering in these soil types, are
extent of 2 H d in the maintained phreatic surface case, but is
given in Fig. 4.
greater than 4 H d in the drawdown case.
The tunnel linings were modelled as continuous elastic
Figure 8 presents the distributions of vertical total stress,
rings. For the London profile the lining represented a precast
pore water pressure and vertical effective stress along the
concrete lining, but the joints between segments were not
vertical axis line above the tunnel crown. Inspection of these
modelled. For the Seoul profile the model represented a fully
distributions identifies the long-term influence of ground-
hardened sprayed concrete lining. The linear elastic proper-
water movements. In the case of maintained phreatic surface
ties are given in Appendix 2.
flow (Fig. 8(a)) a significant reduction in pore water pressure
occurs during the equilibration period, which increases the
Permeability models. For the maintained phreatic surface effective vertical stress. Consequently the total vertical stress
analyses the Thames gravel and Lambeth Group sand for
London, and the fill/alluvium for Seoul, were modelled using
non-consolidating elements, so no permeability model was
required. These elements have no pore pressure degree of
Tunnel diameter = D
freedom at their nodes, so the pore pressures prescribed in
the initial stresses remain constant throughout the analysis.
All other materials were modelled using consolidating
elements with pore pressure degrees of freedom, so the
permeability is required to formulate the finite element
equations that couple mechanical and hydraulic behaviour.
The logarithmic permeability model introduced in ‘Per-
meability model for soil’ above was employed, and the
parameters are given in Fig. 4. Tunnel
For the drawdown analyses, the row of elements represent- depth = Zo
ing Thames gravel below the original water table were given
σn Ground loading
the same permeability model and parameters as the London
clay, but the mechanical model remained that of the gravel. Overburden pressure
The same approach was used for the fill/alluvium elements B
σvo = γtZo
below the original water table in the Seoul analyses: that is, Hf: hoop thrust
they were given the same permeability model and para- A C
meters as the decomposed granite soil. In addition to the γt = bulk unit weight of soil
logarithmic permeability model the suction switch was em-
Overburden load Fo = ZoDγt/2 Fo
ployed. This prescribed a reduction of permeability of two
orders of magnitude between a suction of 50 kPa and
100 kPa for the London soils, and from 0 kPa to 50 kPa Percentage ground loading = {σn/σvo} × 100(%)
suction for the Seoul soils. The combination of these per- Percentage hoop thrust = {Hf /Fo} × 100(%)
meability models and the hydraulic ‘cut-off’ boundary con-
dition referred to in ‘Boundary conditions’ above resulted in Fig. 6. Parameters to define lining performance
396 SHIN, ADDENBROOKE AND POTTS
kPa:
EL–02·5 0
down case, as the increase in effective stress is so much
greater than in the maintained phreatic surface case. Hsi et
100 al. (1994) similarly showed a larger surface settlement
EL–18·0
response to pumping from a point sink with drawdown flow
EL–22·0 200
than with maintained phreatic surface flow. In contrast, the
300 impermeable lining results in a reduction in the volume of
Scale
the surface settlement profile with time, caused by the
400 swelling of the soil immediately above the tunnel, as identi-
4m
fied in Fig. 8(c). The maintained phreatic surface permeable
(a)
and impermeable cases show no change in volume loss after
kPa:
EL–02·5 0
about 3000 days, whereas the drawdown problem shows
continued increases for much longer. (The changes in sur-
100 face settlement volume exhibited in Fig. 9 are associated
EL–18·0
with changes in maximum settlement and in settlement
EL–22·0 200
profile, but such considerations are beyond the scope of this
300 paper.) Bowers et al. (1996) report field measurements above
the Heathrow Express tunnel in London (a sprayed concrete
Scale 400
4m lining tunnel) that show an increase in settlement with time.
The numerical results presented here therefore suggest that
(b)
the Heathrow Express tunnel lining is permeable.
kPa:
EL–02·5 0
Lining performance. The variation of ground loading at the
100 tunnel crown with time is shown in Fig. 10(a). The variation
EL–18·0
with time of the permeable linings is not significant. Fig.
EL–22·0 200 10(b) shows that for these cases the long-term distribution of
300 ground loading round the tunnel is similar to the post-
Scale
construction distribution—about 40% of overburden pressure.
400 However, Fig. 10(a) shows that the ground loading at the
4m
crown on the impermeable lining increases from 170 kPa to
(c)
230 kPa. This takes place over a period of about 5000 days
Fig. 7. Pore water pressure distributions: (a) end of construc- (14 years), following which the long-term value is achieved.
tion; (b) maintained phreatic surface case (long term); (c) Fig. 10(b) reveals that the increase is fairly uniform around
drawdown case (long term) the tunnel. The ground loading on the impermeable lining
reaches slightly over 60% of overburden pressure in the long
term.
decreases slightly. Fig. 8(b) presents the results for the Figure 11 presents the ground loading results from the
drawdown flow condition. With the modelling method used analysis with finite lining permeability in the context of the
in this paper, the maximum suction developed is about fully permeable and impermeable results. The trend with
60 kPa. This suction above the phreatic surface increases the time at the crown lies between those of the fully permeable
vertical effective stress significantly. It is noteworthy that and impermeable linings (Fig. 11(a)), showing a slight in-
suction levels as low as this would not cause desaturation of crease with time. It takes more than 40 000 days (110 years)
London clay. Additionally, the boundary condition employed to reach a steady-state value, longer than the impermeable
at the clay/gravel interface results in a jump in suction from lining. Fig. 11(b) shows that the long-term ground loading is
60 kPa at the top of the clay to zero in the gravel, which is between those for the fully permeable and the impermeable
not unreasonable as a gravel stratum would not be able to linings, lying much closer to the former than to the latter.
sustain such suctions. For the analysis with the impermeable Figure 12(a) shows the variation of hoop thrust with time
lining (Fig. 8(c)) the initial hydrostatic pore pressure condi- at the tunnel crown. The hoop thrust in a permeable lining
tion is necessarily recovered in the long term. During varies little over time. It shows a slight reduction, which
equilibration, therefore, swelling takes place around the reaches a long-term value after about 10 000 days (27 years)
tunnel. A slight decrease in effective stress occurs, and with maintained phreatic surface flow, but is still reducing
consequently the total stress at the tunnel crown increases by after 40 000 days (110 years) with drawdown flow. Fig. 12(b)
30% from 175 kPa on completion of excavation to 230 kPa shows that the hoop thrust reduces all round the tunnel with
in the long term. maintained phreatic surface flow, but increases slightly at the
In response to changes in mean effective stress the per- springline with drawdown flow. However, Fig. 12(a) shows
meability of the soil changes. After tunnel construction the that, in an impermeable lining, the hoop thrust at the crown
permeability at axis level (k axis ) is 1:3 3 1010 m=s, a slight increases significantly, by 120 kPa (30% of the post-
increase from 1:0 3 1010 m=s in the initial state. Subse- construction value), over the 14-year equilibration period.
quently, for both the maintained phreatic surface and the Fig. 12(b) shows that the hoop thrust reaches 60% of the
drawdown regime with a permeable lining, k axis reduces to overburden load.
9:5 3 1011 m=s in the long term. In contrast, with the Figure 13 presents the results from the analysis with finite
impermeable lining, k axis in the long term increases to a lining permeability, again in the context of the fully per-
value greater than 1:4 3 1010 m=s. meable and impermeable results. The hoop thrust at the
One manifestation of the ground response to post-tunnel- crown shows an initial increase with time, below that of the
ling pore pressure equilibration is the change in volume (per impermeable lining (Fig. 13(a)). After 500 days the trend
metre run) of the surface settlement profile with time. This reverses, and follows a decreasing trend very similar to that
is plotted as volume loss in Fig. 9, showing the variation for the permeable lining, achieving steady state after
with time from the initial target value of 2%. It shows that, 40 000 days (110 years). Fig. 13(b) shows that the long-term
for a tunnel with a permeable boundary, there is an increase hoop thrust is between those for the fully permeable and the
in surface settlement. This increase is greater for the draw- impermeable linings, and as with the ground loading it lies
 EFFECT OF GROUNDWATER MOVEMENT ON LONG-TERM TUNNEL BEHAVIOUR 397
Initial
18 End of construction 18 18

Distance from tunnel crown: m

Distance from tunnel crown: m

Distance from tunnel crown: m

Steady state
15 15 15

12 12 12

9 9 9

6 6 6

3 3 3

0 0 0
0 100 200 300 400 0 100 200 300 400 0 100 200 300 400
Vertical total stress: kPa Pore water pressure: kPa Vertical effective stress: kPa
(a)

18 18 18
Distance from tunnel crown: m

Distance from tunnel crown: m

Distance from tunnel crown: m

15 15 15

12 12 12

9 9 9

6 6 6

3 3 3

0 0 0
0 100 200 300 400 –100 0 100 200 300 0 100 200 300 400
Vertical total stress: kPa Pore water pressure: kPa Vertical effective stress: kPa
(b)

18 18 18
Distance from tunnel crown: m

Distance from tunnel crown: m

Distance from tunnel crown: m

15 15 15

12 12 12

9 9 9

6 6 6

3 3 3

0 0 0
0 100 200 300 400 0 100 200 300 400 0 100 200 300 400
Vertical total stress: kPa Pore water pressure: kPa Vertical effective stress: kPa
(c)

Fig. 8. Stress distribution along the tunnel centre line above the tunnel crown: (a) permeable lining, maintained phreatic surface
case; (b) permeable lining, drawdown case; (c) impermeable lining

30
Fully permeable (maintained phreatic surface) The only data available for comparison with the numerical
Fully permeable (drawdown) predictions are Barratt et al.’s (1994) long-term measure-
24 Impermeable lining ments of hoop thrust, which are reproduced in Fig. 14.
(Their paper also reported measurements of ground loading,
Volume loss: %

18 but these are thought to be less reliable because they show

accelerating rates of increase 30 years after construction.)
12 The hoop thrust at both the crown and the axis increased
with time. In the long term the hoop thrust has reached
6 about 60% of overburden load at the tunnel axis and 40% at
the crown. The predicted short-term post-construction lining
0 load, which is the same for all analyses, is overestimated,
0 1000 2000 3000 4000 5000 6000 but increasing trends with time at both the axis and the
Time: days
crown were achieved with the impermeable lining, as shown
Fig. 9. Variation of volume loss with time in Fig. 14. The post-construction value of hoop thrust from
a two-dimensional analysis is entirely dependent on the
degree of unloading permitted prior to lining installation. It
much closer to the former than to the latter. (The relative is only the increments of unloading that follow lining
position of the long-term ground loading and hoop thrust construction that introduce stresses into the lining elements.
between the fully permeable and impermeable solutions will The methodology adopted in this study was to reproduce
depend on the ratio of k l to k s .) reasonable ground movements during tunnel excavation. It is
398 SHIN, ADDENBROOKE AND POTTS
300 75

Percentage of overburden pressure: %

Fully permeable (maintained phreatic surface)
Fully permeable (drawdown)
65
250 Impermeable lining
Ground loading: kPa

55

200

45

150
End of construction 35

100 25
10 102 103 104 105
Time: days
(a)

60

40

End of construction
Fully permeable (maintained phreatic surface)
Fully permeable (drawdown)
Impermeable lining
20 40 60 (%)

(b)

Fig. 10. Ground loadings on the tunnel lining: (a) variation of

ground loading with time (tunnel crown); (b) as a percentage of Fig. 11. Effect of finite permeable lining on ground loading: (a)
overburden pressure (óvo 400 kPa) variation of ground loading with time at the tunnel crown
(kl 2:0 3 10211 m=s, kl =ks 0:1); (b) distribution of ground
loading on the lining

clear therefore that if the methodology adopted was to give

post-construction hoop thrusts that matched the field data,
the lining would need to be installed on a much later
increment of excavation. The consequence of this would be Maintaining an impermeable lining into the long term
that the ground movements would be overpredicted, giving produces no change in pore pressure in the ground. Allowing
significantly larger pore water pressure changes due to flow into the tunnel during the equilibration period intro-
excavation. In the long term the magnitude of hoop thrust at duces a pore pressure response. Fig. 15 shows the long-term
the axis is closely matched, but at the crown it is over- distribution of pore water pressure. The zones of influence
predicted by the analysis with impermeable lining. are greater than predicted for London clay, and it is note-
worthy that, under drawdown flow conditions, the phreatic
surface is drawn down to the tunnel. It is arguable that in
Seoul profile reality such deep drawdown could occur during a very dry
The analyses show that the decomposed granite soil is season if there were no near-surface recharge from leaking
sufficiently permeable to give negligible excess pore water water pipes, but this is unlikely to be the case all year
pressure at the end of construction. The short-term zone round, particularly during the wet season.
of influence, defined in terms of pore pressure variation Differences in the long-term behaviour are highlighted by
from hydrostatic, is therefore also negligible. If inflow is comparing the ground loading on, and hoop thrust in, the
then permitted, the equilibration period is very rapid and tunnel lining. Fig. 16(a) compares ground loadings on an
long-term behaviour is reached within about 20 days of impermeable lining under long-term steady-state conditions.
completion of excavation. This time period would be At the end of construction both linings have very similar
reduced if flow into the tunnel were permitted during ground loads of just under 50% of the overburden pressure.
excavation. In the long term the ground loading at the tunnel spring line
 EFFECT OF GROUNDWATER MOVEMENT ON LONG-TERM TUNNEL BEHAVIOUR 399
700 86 600 74

Percentage of overburden load: %

Fully permeable (maintained phreatic surface)
Fully permeable (maintained phreatic surface) 550 Impermeable lining

Percentage of overburden load: %

Fully permeable (drawdown) Finite permeable (maintained phreatic surface)
65

Hoop thrust: kPa

600 Impermeable lining 74 500
Hoop thrust: kPa

450 56

500 62 400
End of construction 47
350

400 50 38
300
10 102 103 104 105
Time: days
(a)
300 38
10 102 103 104 105
Time: days
(a)
400

Fully permeable (maintained phreatic surface)

60 Impermeable lining 200
Finite permeable (maintained phreatic surface)

End of construction 40 0 200 400 (kN/m)

Fully permeable (maintained phreatic surface)
Fully permeable (drawdown)
Impermeable lining 20

0 20 40 60 (%)

(b)

Fig. 13. Effect of finite permeability of the lining on hoop

thrust: (a) variation of hoop thrust with time at the tunnel
crown (kl 2:0 3 10211 m=s, kl =ks 0:1); (b) distribution of
hoop thrusts in the lining
(b)

Fig. 12. Distribution of hoop thrust in the lining: (a) variation

of hoop thrusts with time ( Z o 20 m); (b) as a percentage of Crown cell (Barratt et al.)
overburden load (Fo 800 kN=m2 ) Axis cell ( " )
100 Axis cell ( " )
Crown (calculated for the impermeable lining)
Percentage of overburden load: %

Axis ( " )

75

in London clay reaches just over 60%, whereas in decom-

posed granite soil it is unchanged at less than 50%. This
difference occurs because at the end of construction, unlike 50
in the London clay, in the decomposed granite soil there is
no excess pore pressure to equilibrate. In the case where the
lining is permeable, no appreciable differences are seen 25
between the soil types, as shown in Fig. 16(b). It is
noteworthy that the long-term ground loads on the imperme-
able lining are only slightly greater than those on the 0
permeable lining in the decomposed granite soil. About 90% 1 10 102 103 104
Time: days
of the ground load on the impermeable lining is due to pore
pressure, so the effective stress acting on the lining is very
Fig. 14. Comparison of hoop thrusts with field data (data from
low. Regents Park after Barratt et al., 1994)
Figures 17(a) and 17(b) present the same relative compari-
sons, but for hoop thrust as a percentage of overburden load.
For an impermeable lining the long-term hoop thrust is a
maximum at the invert and the crown in London clay, and a behaviour during construction, not during the equilibration
minimum at the springline. In contrast, in decomposed period.
granite soil the maximum is at the springline, and the values
at the crown and invert are smaller. The same comparison
can be made for the permeable lining results in Fig. 17(b). CONCLUSIONS
Because the hoop thrust varies little with time for perme- Coupled non-linear finite element analyses simulating
able linings, this comparison is a consequence of different groundwater movements and predicting their influences on
400 SHIN, ADDENBROOKE AND POTTS
kPa:
0
(a) When the tunnel lining is permeable, surface settle-
EL–02·5
ments increase with time during the equilibration
100 period. In contrast, when the lining is impermeable,
EL–18·0
the surface settlements reduce with time. The limited
EL–22·0 200
available field evidence indicates that tunnelling-
300 induced settlements in London do increase with time
Scale
after construction, supporting field observations that
400 tunnels act as drains in the long term.
4m
(b) When the tunnel lining is permeable, changes in ground
(a) loading and hoop thrust with time are insignificant.
However, the ground loading on an impermeable lining
increases with time. The available field data, which
EL–02·5
kPa: again are limited, indicate that lining loads do increase
0 with time in London.
EL–18·0 (c) An increase in settlement with time and an increase in
100
EL–22·0
lining loads with time was achieved only with the finite
200 permeable lining analysis. Such a lining model is
therefore arguably more representative of reality than
Scale 300
4m
either of the two extremes. The magnitude of the
predicted post-construction settlement increase is less
(b) than that with the fully permeable lining, and the
magnitude of the predicted post-construction increase in
Fig. 15. Long-term distributions of pore water pressure (fully lining loads is less than with the impermeable lining.
permeable lining): (a) maintained phreatic surface case; (b)
drawdown case
(d ) A successful method for modelling linings with finite
permeability has been presented. The approach com-
bines structural beam elements, which model the
mechanical behaviour of the lining, with thin solid
ground and tunnel behaviour have been presented. Contrast- consolidating elements, which model the hydraulic
ing the results from different flow regimes and hydraulic behaviour of the lining.
boundary conditions has highlighted the care needed when (e) Under drawdown flow conditions, the drawdown of the
designing a long-term tunnelling analysis. Comparisons with water table causes larger surface settlements than under
field data to support the analyses have been limited by the maintained phreatic surface flow. In contrast, however,
scarcity of such data. Such comparisons have permitted the introduction of drawdown flow does not cause any
qualitative conclusions to be drawn, but the sensitivity of significant change in ground loading compared with
long-term tunnel behaviour to a number of controlling para- maintained phreatic surface flow.
meters, as revealed by this study, prevents the drawing of ( f ) With a stress-dependent permeability model, the per-
quantitative conclusions. meability of the soil adjacent to the tunnel reduces with
Analyses in the low-permeability London clay revealed time if the lining is permeable, but increases with time
that: if the lining is impermeable.

London clay London clay (maintained phreatic surface)

London clay (drawdown)
Decomposed granite soil
Decomposed granite soil (maintained phreatic surface)
70 Decomposed granite soil (drawdown) 70
60 60
50 50
40 40
30 30
20 20
10 10
10 20 30 40 50 60 70 (%)
10 20 30 40 50 60 70 (%)

(a) (b)

Fig. 16. Comparison of ground loadings, as a percentage of overburden pressure, on: (a) impermeable lining; (b) fully permeable
lining
 EFFECT OF GROUNDWATER MOVEMENT ON LONG-TERM TUNNEL BEHAVIOUR 401
London clay London clay (maintained phreatic surface)
London clay (drawdown)
Decomposed granite soil
Decomposed granite soil (maintained phreatic surface)
60
Decomposed granite soil (drawdown) 60
50
50
40
40
30
30
20
20
10
10
0 10 20 30 40 50 60 (%)
0 10 20 30 40 50 60 (%)

(a) (b)

Fig. 17. Comparison of hoop thrusts, as a percentage of overburden load: (a) impermeable lining; (b) fully permeable lining

Comparisons of the numerical results from the highly per- at Westminster (Higgins et al., 1996). Those for the decomposed
meable decomposed granite soil with those from the low- granite soil are as reported in Shin (2000).
permeability London clay highlighted the following: The non-linear equations for tangent moduli are (after Potts &
Zdravkovic, 1999)
(a) The general trends of ground behaviour are quite 1
similar; however, since the magnitude of the excess 3G X (c2 )
¼ C1 þ C2 cos[c1 X c2 ]  C2 :c1 :c2 : sin[c1 X c2 ] (2)
pore water pressure after construction is small and the p9 2:303
steady-state condition is reached quickly, time-depen- 1
K X (c4 )
dent effects are not significant. ¼ C4 þ C5 cos[c3 Y c4 ]  C5 :c3 :c4 : sin[c3 X c4 ] (3)
(b) The zone of pore pressure alteration (from far-field p9 2:303
values) in the long term is larger than for low- where
permeability soil. Under drawdown conditions the    
Ed v
phreatic surface is drawn down to the tunnel, whereas X ¼ log10 pffiffiffi ; Y ¼ log10
3 C3 C6
with low-permeability soil it remains above the crown
in the long term. (The analyses presented here did not G is the shear modulus, K is the bulk modulus, p9 is the mean
take account of the effects on the mechanical beha- effective stress, Ed is the deviatoric strain invariant used in ICFEP,
viour of desaturation above the phreatic surface in de- v is the volumetric strain, and C1, C2 , C3 , C4 , C5 , C6 , c1 , c2 , c3 , c4
composed granite soil. The suctions generated in are all coefficients. Ed is related to a (the axial strain observed in
the London clay analyses are too small to cause undrained triaxial tests) by the expression
desaturation.) pffiffiffi
E d ¼ 3 a (4)
(c) The ground loading on, and hoop thrust in, an
impermeable lining in highly permeable soil are not where
much higher than those for a permeable lining. In rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
contrast, an impermeable lining in a low-permeability Ed ¼ 2 [(1  2 )2 þ (1  3 )2 þ (2  3 )2 ]
6
soil attracts significantly higher ground loading and
hoop thrust than a permeable lining. with 1 , 2 and 3 being principal strains. The constants are obtained
from a fit to laboratory data from stress path tests. Throughout an
analysis the stiffness at a particular point is continually changing. It
depends on both the current strain and the current mean effective
ACKNOWLEDGEMENTS stress at that point. Until a specified minimum strain, Ed min or v min,
The first author was supported by the Seoul Metropolitan is exceeded the stiffness varies only with the mean effective stress.
Government, an Overseas Research Studentship Award from This condition also applies once a specified upper strain limit is
the University of London, and a British Chevening joint exceeded, Ed max or v max. Ed min , v min and Ed max, v max are
required ‘cut-offs’ because of the trigonometric nature of the
scholarship during his research at Imperial College. This
equations. The magnitude of the stiffness is prevented from falling
support is gratefully acknowledged. below specified minimum values (Gmin or K min ). The relevant
parameters are given in Tables 1 and 2 for London and Seoul
respectively.
APPENDIX 1. PRE-YIELD SOIL MODELS For the Seoul profile, the fill/alluvium and the slightly
The soil properties for London clay used in the analyses are based to unweathered granite were modelled as linear elastic pre-yield,
on those used for the analysis of the Jubilee Line Extension station defined by two independent elastic parameters (Table 3).
402 SHIN, ADDENBROOKE AND POTTS
Table 1. Material parameters for non-linear elastic model: London soils
Thames gravel London clay Lambeth clay Lambeth sand
Shear modulus
parameters
C1 1104 1400 1000 1300
C2 1035 1270 1045 1380
C3 : % 5 3 104 1 3 104 5 3 104 1 3 104
c1 0·974 1·335 1·334 1·22
c2 0·94 0·62 0·59 0·65
Ed min : % 8:8334 3 104 8:66025 3 104 13:8564 3 104 1:90526 3 104
Edmax : % 0·366 0·693 0·381 0·13
Gmin : kPa 2000 2666·7 2666·7 1000
Bulk modulus
parameters
C4 275 686 530 275
C5 225 633 460 235
C6 : % 2 3 103 1 3 103 5 3 104 5 3 104
c3 0·998 2·069 1·492 1·658
c4 1·04 0·42 0·68 0·54
v min : % 2:1 3 103 5:0 3 103 1:5 3 103 5:1 3 104
v max : % 0·2 0·15 0·16 0·3
K min : kPa 5000 5000 5000 5000

Table 2. Material parameters for non-linear elastic model: decomposed granite soil
Shear modulus parameters C1 C2 C3 : % c1 c2 Ed min : % Ed max : % Gmin : kPa
1515 1485 2 3 104 0·955 0·818 9:0 3 103 0·35 9706
Bulk modulus parameters C4 C5 C6 : % c3 c4 v min : % v max : % K min : kPa
475 465 2 3 104 0·848 0·872 5:0 3 103 0·5 6438

Table 3. Material parameters for linear elastic model NOTATION

Fill/alluvium Stiffness parameters
E9: kPa 1:47 3 104 E9 drained Young’s modulus
9 0·35 G shear modulus
K bulk modulus
Slightly to unweathered granite 9 drained Poisson’s ratio
E9: kPa 1:0 3 106
9 0·28 Stress and strength parameters
c9 cohesive strength
9 angle of shearing resistance
Ø9 angle of dilation
Ko coefficient of earth pressure at rest
p9 mean normal effective stress
q deviatoric stress in triaxial space
n ground loading on tunnel lining
APPENDIX 2. TUNNEL LINING MODEL  vo overburden pressure at axis level
The linear elastic parameters are listed in Tables 4 and 5.
Strain parameters
Ed deviatoric strain invariant
a axial strain in a triaxial test
Table 4. Precast concrete lining (London profile) v volumetric strain
Cross-sectional area: m2 =m 0·168 1 , 2 , 3 principal strains
Second moment of area: m4 =m 0·000 395 14
Permeability parameters
Young’s modulus: kPa 2:8 3 107
k coefficient of permeability
Poisson’s ratio 0·15
kaxis coefficient of permeability at tunnel axis level
kl coefficient of permeability of the tunnel lining
ko coefficient of permeability at zero mean effective stress
ks coefficient of permeability of the soil

Other parameters
Table 5. Shotcrete lining (Seoul profile) D tunnel diameter
Fo overburden load
Cross-sectional area: m2 =m 0·246 Hd depth to tunnel crown below phreatic surface
Second moment of area: m4 =m 0·0016 Hf hoop thrust
Young’s modulus: kPa 2:0 3 107 Zo depth to tunnel axis below ground level
Poisson’s ratio 0·2
Z depth below ground surface
 EFFECT OF GROUNDWATER MOVEMENT ON LONG-TERM TUNNEL BEHAVIOUR 403
REFERENCES Studies of the influence of non-linear stress–strain character-
Addenbrooke, T. I. (1996). Numerical analyses of tunnelling in stiff istics in soil–structure interaction. Géotechnique 36, No. 3,
clay. PhD thesis, Imperial College, University of London. 377–396.
ASCE (1984). Guidelines for tunnel lining design. O’Rourke (ed.). Lambe, T. W. & Whitman, R. V. (1969). Soil mechanics, SI version.
New York: ASCE. New York: John Wiley & Sons.
Atkinson, J. H. & Mair, R. J. (1983). Loads on leaking and water- Mair, R. J. (1998). Geotechnical aspects of design criteria for bored
tight tunnel linings, sewers and buried pipes due to groundwater. tunnelling in soft ground. Proceedings of world tunnel congress
Géotechnique 33, No. 3, 341–344. on tunnels and metropolis, Sao Paulo, pp. 183–199.
Atwa, M. & Leca, E. (1994). Analysis of groundwater seepage into Mair, R. J. & Taylor, R. N. (1997). Bored tunnelling in the urban
tunnels. Proceedings of International congress on tunnelling and environment. Proc. 14th Int. Conf. Soil Mech. Found. Engng,
ground conditions, Cairo, Egypt (ed. A. Salam), pp. 303–310, Hamburg 4, 2253–2385.
Rotterdam:Balkema. New, B. M. & O’Reilly, M. P. (1991). Tunnelling induced ground
Barratt, D. A., O’Reilly, M. P. & Temporal, J. (1994). Long term movements: predicting their magnitude and effect. Proc. 4th Int.
measurements of loads on tunnel linings in overconsolidated Conf. Ground Movements and Structures, Cardiff, 671–697.
clay, Proc. Tunnelling’94, London, 469–481. Peck, R. B. (1969). Deep excavations and tunnelling in soft ground:
Bowers, K. H., Hiller, D. M. & New, B. M. (1996). Ground state of the art. Proc. 7th Int. Conf. Soil Mech. Found. Engng,
movement over three years at the Heathrow Express Trial Mexico City, State of the Art Volume, 225–290.
Tunnel. In Geotechnical aspects of underground construction in Potts, D. M. & Zdravkovic, L. (1999). Finite element analysis in
soft ground (eds R. J. Mair and R. N. Taylor), pp. 647–652. geotechnical engineering: theory. London: Thomas Telford.
Rotterdam: Balkema. Schweiger, H. F., Schuller, H. & Pottler, R. (1997). Some remarks
Day, R. A. & Potts, D. M. (1990). Curved Mindlin beam and on 2D models for numerical simulation of underground con-
axisymmetric shell elements: a new approach. Int. J. Numer. struction with complex cross-sections. Proceedings conference
Methods Engng 30, 1263–1274. on computer methods and advances in geomechanics,
Gunn, M. J. & Taylor, R. N. (1984). Discussion on Atkinson & China, pp. 1303–1308, Rotterdam:Balkema.
Mair (1983). Géotechnique 35, No. 1, 73–75. Shin, J. H. (2000). Numerical analysis of tunnelling in decomposed
Higgins, K. G., Mair, R. J. & Potts, D. M. (1996). Numerical granite soil. PhD thesis, Imperial College, University of London.
modelling of the influence of the Westminster Station excavation Shin, J. H. & Yoo, T. S. (1985). A study on ground behaviour
and tunnelling on the Big Ben clock tower. In Geotechnical during tunnel excavation. J. Korea Geotech. Soc. 1, 31–46.
aspects of underground construction in soft ground (eds R. J. UMTA (1986). Water intrusion problems in transit tunnels (Final
Mair and R. N. Taylor), pp. 525–530. Rotterdam: Balkema. Report). Urban Mass Transportation Administration, US Depart-
Hsi, J. P. & Small, J. C. (1992). Simulation of excavation in a poro- ment of Transport.
elastic material, Int. J. Numer. Anal. Methods Geomech. 16, Vaughan, P. R. (1989). Non-linearity in seepage problems: theory
25–43. and field observations. De Mello Volume, pp. 501–516, Sao
Hsi, J. P., Carter, J. P. & Small, J. C. (1994). Surface subsidence Paulo: Editora Edgard Blucher Ltda.
and drawdown of the water table due to pumping. Géotechnique Ward, W. H. & Pender, M. J. (1981). Tunnelling in soft ground:
44, No. 3, 381–396. general report. Proc. 10th Int. Conf. Soil Mech. Found. Engng,
Jardine, R. J., Potts, D. M., Fourie, A. B. & Burland, J. B. (1986). Stockholm 4, 261–275.