Tunnelling and

Underground Space
Technology
incorporating Trenchless
Technology Research
Tunnelling and Underground Space Technology 21 (2006) 119–132
www.elsevier.com/locate/tust

Tunnel stability and arching effects during tunneling in soft clayey soil
a,*
C.J. Lee , B.R. Wu b, H.T. Chen a, K.H. Chiang a

a
Department of Civil Engineering, National Central University, No. 300, Jung-da Rd., Chungli, Taoyuan 32054, Taiwan
b
National Science and Technology Center for Disaster Reduction, 3F., No. 106, Sec. 2, HoPing E. Rd., Taipei 106, Taiwan

Received 5 October 2004; received in revised form 2 April 2005; accepted 5 June 2005
Available online 8 August 2005

Abstract

A series of centrifuge model tests and numerical simulations of these tests were carried out to investigate the surface settlement
troughs, excess pore water pressure generation, tunnel stability and arching effects that develop during tunneling in soft clayey soil.
The two methods were found to provide consistent results of the surface settlement troughs, excess pore water generation, and the
overload factors at collapse for both single and parallel tunneling. The arching ratio describes the evolution of the arching effects on
the soil mass surrounding tunnels and can be derived from the numerical analysis. The boundaries of the arching zones for both
single tunneling and parallel tunneling were determined. In addition, the boundaries of the positive and negative arching zones were
also proposed.

2005 Elsevier Ltd. All rights reserved.

Keywords: Arching effect; Tunnel stability; Centrifuge modeling; Numerical modeling

1. Introduction tion of tunneling-induced ground movements during

excavation of soft ground tunnel has been carried
Tunneling in soft clayey soils has become very pop- out using various methods, including empirical meth-
ular in recent years because it is one of the best con- ods derived from field observations (Peck, 1969;
struction methods for building mass rapid transit Clough and Schmidt, 1981) and centrifuge modeling
systems and sewage collection systems in densely pop- (Mair et al., 1981; Wu and Lee, 2003; Lee et al.,
ulated cities. As the face of a tunnel is advanced, a 2004), or numerical and analytical methods (Lee and
means of supporting the ground close to the face Rowe, 1991).
may be needed; without such support, collapse might Terzaghi (1943) explained how stress transfer from
occur due to gross plastic deformation of the soil. yielding parts of a soil mass to adjacent non-yielding
Moreover, tunneling inevitably induces varying de- parts leads to the formation of an arching zone. This
grees of ground movement towards the tunnel opening problem has two modes of displacement, depending
and results in detrimental effects on nearby facilities, on whether the trap door is translated into the soil
such as shallow foundations, piles, existing tunnels (passive mode) or away from it (active mode). The
and other pipeline systems. Taking appropriate mea- passive mode can be used for the evaluation of the
sures to protect nearby facilities before excavation is uplift force of anchors, or of any buried structure that
an important part of engineering practice. The predic- can be idealized as an anchor. The active mode can be
used to study the gravitational flow of granular mate-
* rial between vertical walls (the silo problem) or the
Corresponding author. Tel.: +886 3 4227151x34135; fax: +886 3
4252960. ground pressure on tunnel liners. Ladanyi and Hoy-
E-mail address: [email protected] (C.J. Lee). aux (1969) performed a series of model trap-door tests

0886-7798/$ - see front matter
2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.tust.2005.06.003
120 C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119–132

under 1 g conditions in order to check the validity of 2. Centrifuge and numerical modeling
the classic bin theory. Handy (1983) analyzed soil
arching action behind retaining walls, and Wang and 2.1. Centrifuge modeling
Yen (1973) carried out this analysis for slopes. Nakai
et al. (1997) performed a series of physical model tests The basic principle of centrifuge modeling is to recre-
under 1 g conditions and carried out numerical analy- ate the stress conditions that are present in full-scale
sis of these tests to investigate the arching effect. They constructions in models of greatly reduced scale. The
found that the results obtained from the model tests full-scale system modeled with a centrifuge model (with
were in good agreement with those obtained from dimensions N times larger than those of the model if it is
the numerical analysis. Park and Adachi (2002) per- tested in an acceleration that is N times earth gravity) is
formed model tests under 1 g conditions to simulate referred to as the prototype. It is intended that the pro-
tunneling events in unconsolidated ground with vari- totype should include all the important characteristics of
ous levels of inclined layers. They found that remark- the field situation of interest. Centrifuge modeling pro-
able non-symmetrical distributions of the earth vides an opportunity to study for example the ground
pressure arose when a tunneling event took place in responses due to tunneling before and after collapse; col-
inclined layers with 60 of inclination. Stone and New- lapse is of course not permitted to occur in the field.
son (2002) presented the results of a series of centri- This experimental study was undertaken in the geo-
fuge tests designed to investigate the effects of technical centrifuge at the National Central University.
arching on soil–structure interaction. Koutsabeloulis The NCU centrifuge has a nominal radius of 3 m and
and Griffiths (1989) implemented a finite element is capable of accelerating a 1 tonne model package to
method to investigate the trap-door problem. The con- 100 g and 0.55 tonne to 200 g. In the single-tunnel model
cept of soil arching was recently adopted in the anal- tests, one model tunnel, 60 mm in diameter, was embed-
ysis of the mobilization of resistance from passive pile ded at various depths specified by the cover-to-diameter
groups subjected to lateral soil movement (Chen and ratio (C/D). In the parallel-tunnel model tests, two mod-
Martin, 2002). el tunnels (60 mm in diameter) were separated by a spec-
When tunneling is conducted in the vicinity of exist- ified center-to-center distance (d) (as shown in Fig. 1),
ing pile foundations, the axial load transfer mechanism and buried at various depths specified by the cover-to-
and failure mode on existing piles vary depending on diameter ratio (C/D). All the model tests reported in this
the distance between the existing piles and the new driv- study were carried out under a centrifugal acceleration
ing tunnel and relative elevation of the piles with respect of 100 g in order to model a prototype tunnel with
to the centerline of the tunnel (Lee and Chiang, 2004). a diameter of 6 m embedded at depths with the tested
These behaviors result from the complicated redistribu- C/D ratios.
tions of stress around tunnels during tunneling. Hence, The soil used in the model tests had a plasticity index
the stress distribution in the vicinity of a tunnel or of of 18 and was classified as CL in the Unified Soil Clas-
several tunnels stacked closely in an underground sta- sification System. The soil slurry was remolded at about
tion needs to be established before appropriate protec- twice its liquid limit in a mixer and poured into the con-
tion measures for nearby existing piles can be solidometer. Consolidation pressure was applied in five
implemented. By deepening our understanding of the stages, with a final pressure of 196 kPa. Further details
arching effect in various geotechnical problems, we can of the soil bed preparation can be found in Wu and
improve the design of the protection measures required
for existing underground structures nearby new
tunneling. LVDT
Both centrifuge and numerical modeling were used in
the study. The stability of a tunnel, the movements of
soil mass, the evolution of stress on the soil mass around C
a tunnel, and the boundaries of the arching zone during
tunneling in clayey soils are investigated and discussed. D
480 mm

Firstly, a series of centrifuge model tunnel tests was con- (60 mm)
ducted. A finite difference program (FLAC) was then
chosen for numerical analysis of the system described PPT
d
by the centrifuge model to provide insight into the arch- Marked Tunnel
spaghetti Tunnel deformation gauges
ing mechanism and the boundaries of the arching zone
during tunneling. Finally, the results from the numerical (Dimensions are in model scale)
modeling and the measurements from the centrifuge
820 mm
modeling were compared in order to assess their
predictions. Fig. 1. Setup of test package for two parallel tunnels (model scale).
 C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119–132 121

Lee (2003). The basic properties of the prepared soil bed tion of pore water pressure (or changes of effective
are listed in Table 1. On completion of the consolida- stress) during these two stages may be less than 1% in
tion, the soil bed, which had an undrained shear average. A torvane apparatus was used for determining
strength profile of 30–40 kPa, was lifted and placed in the undrained shear strength at various depths on the
a strong box. The set-up of the test package for the par- side of the soil bed before and after the test. The test re-
allel-tunnel model is shown in Fig. 1. Five PPTs were in- sults show that no obvious changes in the undrained
serted at selected positions to monitor changes in the shear strengths (less than 2 kPa) at the same depths were
pore water pressure. The pore pressure transducer found. Therefore, we assumed that there was minimal
(PPT) was inserted into a pre-drilled hole and then the migration of pore water pressure and hence that the soil
hole was fully back-filled with thick slurry. Eight was subjected to undrained shearing. The changes in the
LVDTs were fixed on top of the strong box to record pore water pressures, the tunnel deformations, and the
the transverse surface settlements. surface settlements induced by tunneling were measured
The test package was first spun at an acceleration of continuously. After each model test, the soil bed was
100 g for 5 min so that any voids generated during the cautiously excavated to expose the implanted spaghetti.
installation of the PPTs and the assembly of the test Further, six undisturbed samples were taken from the
package might be filled. After the centrifuge was soil bed at selected depths for use in unconfined com-
stopped, one or two 60 mm diameter model tunnels were pression tests. The undrained shear strength (su) and
cut manually, depending on the test conditions, and the secant YoungÕs modulus (E50), for each soil bed
then rubber bags were inserted into the tunnels. Tunnel was determined from the average of the results for the
deformation gauges consisting of four thin cantilevers six samples, as listed in Table 2.
made from stainless steel were installed inside the rubber The surrounding soil squeezes into the tunnel as the
bags to measure the deformations at the crowns, inverts, supporting air pressure is gradually reduced, which fi-
and side-walls of the tunnels. The test package shown in nally causes tunnel collapse. The overload factor (OF),
Fig. 1 was prepared for further tunnel collapse tests by as defined below, is a useful index for describing tunnel
connecting air pressure lines to the rubber bags. stability throughout the entire test process.
The air pressure in the rubber bags was carefully reg- rvo
pi
ulated to balance the overburden pressure at the tunnel OF ¼ ; ð1Þ
su
center during the reacceleration of the model up to a
centrifuge acceleration of 100 g. The tunneling event where rvo is the overburden pressure at the tunnel cen-
was simulated by simultaneously reducing the air pres- ter, and pi is the supporting pressure. Fig. 2 shows plots
sure inside the tunnels and eventually down to zero. This of the surface settlements at the tunnel axis versus OF
air-pressure method of simulating tunnel excavation was for the single-tunnel test, Test7 (C/D = 3), and of those
adapted from Mair (1979). The air-pressure method was at the symmetrical axis of the settlement trough for the
chosen to support the tunnel during the accelerating parallel-tunnel test, Twin3 (C/D = 3, d/D = 1.5). This
stages because measuring the tunnel deformation was
needed and the waterproofing of this measuring device
Table 2
was difficult if an incompressible heavy fluid-pressure
Test configurations
method was used. The method used in the study may
Test no.a C/D d/D su (kPa)
cause smaller surface settlements and larger settlement
trough widths at the corresponding supporting pressures Test11 0.5 – 31.00
Test12 0.5 – 35.12
but no difference in the supporting pressure at collapse
Test5 1 – 36.90
compared to the incompressible heavy fluid-pressure Test8 1 – 37.90
method. Twin4 1 1.5 33.00
No more than 15 min were spent under 100 g prior to Twin5 1 1.5 39.10
collapse (including the accelerating stage). The dissipa- Test3 2 – 30.25
Test9 2 – 35.79
Twin1 2 3 48.70
Table 1
Twin2 2 1.5 41.00
Basic properties of the prepared soil bed
Twin9 2 1.5 39.52
Specific gravity, Gs 2.67 Twin12 2 2 35.10
Liquid limit, LL 40 Test6 3 – 33.30
Plastic limit, PL 22 Test7 3 – 34.00
Plasticity index, PI 18 Twin3 3 1.5 36.10
Unit weight, c (kN/m3) 18.1 Twin6 3 1.5 32.90
Compression index, Cc 0.28 Test10 4 – 32.17
Swell index, Cr 0.0275 Twin10 4 1.5 34.25
Coefficient of consolidation, Cv (cm2/s) 0.010524 Twin11 4 1.5 32.83
Permeability, k (m/s) 4.5 · 10
9 a
Test3–Test12: single tunnel; Twin1–Twin11: two parallel tunnels.
122 C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119–132

Overload factor, (OF) FLAC2D (Cundall et al., 1993). A plane-strain model

0 1 2 3 4 5 6 7 8 9 10 11 12 with large-strain formulation was used to simulate the
Settlementin prototype scale, (m)

0.0
(OF ) c =3.96 deformation behavior of the ground surrounding un-
(OF )c =3.2 lined tunneling. In order to compare the results of the
0.5
numerical and centrifuge modeling, the boundary condi-
tions and soil properties used in the numerical model
1.0 were chosen to be the same as those studied in the cen-
trifuge model tests. The numerical analysis considered a
1.5 mesh with a width of 82 m and a height of 48 m as
shown in Fig. 3, which are the exact dimensions of the
2.0 Twin3 soil bed used in the centrifuge model tests. The left
Test7
and right boundaries were fixed in the x-direction, and
2.5 the bottom boundary was fixed in the x- and y-
Fig. 2. Definition of overload factor at collapse, (OF)c. directions. The grid size around the tunnels was
0.5 m · 0.5 m in the prototype and was enlarged by a
factor of 1.08 as the distance to tunnel center increases.
figure shows that the surface settlements, S, increase The soil bed was treated as an isotropic and elastic per-
dramatically once OF exceeds a critical value. Extending fectly plastic continuum following the Mohr–Coulomb
the straight-line portions of the first and second parts of failure criterion (/ = 0). A total of six numerical models
the S vs OF curves to intersect at the points shown in were analyzed. The model conditions and the mechani-
Fig. 2, the horizontal ordinate of each of these critical cal properties of soil bed measured from the unconfined
points is defined as the overload factor at collapse, compression tests used in the numerical analysis are
(OF)c. In engineering practice, the tunnel must be sup- shown in Table 3.
ported against collapse during tunneling. Thus, the load The numerical modeling was commenced at a state of
factor (LF) is regarded as the reciprocal of the safety geostatic equilibrium, and allowed to come to numerical
factor equilibrium under the force of gravity. This step pro-
rvo
pi OF vides an estimate of the in situ stress in the soil prior
LF ¼ ¼ ; ð2Þ
rvo
ðpi Þc ðOF Þc
where (pi)c is the measured supporting pressure at col-
lapse in the centrifuge model tests. The value of LF var-
ies from 0 to 1, which corresponds to variation of the
tunnel stability from stable to critical.
A total of 19 model tests were performed in the study:
nine single-tunnel tests and ten parallel-tunnel tests, as
listed in Table 2. The C/D ratio varied from 0.5 to 4
for the single-tunnel tests, and from 1 to 4 for the paral-
lel-tunnel tests. The d/D ratios of the parallel-tunnel
tests were 1.5, 2, and 3 for C/D = 2, and the d/D ratio
was 1.5 for the other C/D ratios.

2.2. Numerical modeling

The numerical ‘‘experiments’’ were carried out with a

two-dimensional explicit finite difference program, Fig. 3. Geometry used in the numerical analysis (C/D = 3, d/D = 1.5).

Table 3
Geometric conditions and mechanical properties of the numerical models
Test no. C/D d/D su (kPa) E50 (kPa) c (kN/m3) m
Ntest8 1 – 37.90 3500 18.1 0.49
Ntwin4 1 1.5 33.00 3500 18.1 0.49
Ntest9 2 – 35.79 3500 18.1 0.49
Ntwin2 2 1.5 41.00 4000 18.1 0.49
Ntest7 3 – 34.00 3500 18.1 0.49
Ntwin3 3 1.5 36.10 3500 18.1 0.49
 C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119–132 123

to tunneling. Secondly, the elements representing the can provide. Therefore, integrating and comparing the
excavated soil in the tunnel were nulled and a uniform results derived from the numerical and centrifuge mod-
supporting pressure equal to the overburden pressure els provides improved understanding of the deformation
at the center of tunnel was applied to the interior bound- behavior and the arching mechanism during tunneling.
ary of the circular opening to keep the tunnel stable. The
supporting pressure was then reduced by a decrement of
10 kPa per step in order to simulate the decrease in the 3. Comparison of the results from centrifuge and
supporting pressure that was applied in the centrifuge numerical modeling
model tests.
For FLAC, the value of maximum nodal unbalance 3.1. Tunnel stability
force is used to determine if a simulation having reached
equilibrium. In the current study, an equilibrium state The tunneling event was simulated by reducing the
was regarded as having converged when the maximum supporting pressure inside the tunnels in both the centri-
unbalance force of every node in the mesh was less than fuge modeling and the numerical experiments. Fig. 5
10 N (the ratio of the maximum unbalance force to the shows the relations between the C/D ratio and (OF)c
overburden pressure on the element was about 0.005%) measured in the single-tunnel and the parallel-tunnel
in the simulation of per step. The simulation was then model tests (solid symbols), and the values of (OFnum)c
moved to the next step (reducing the supporting pres- calculated from the numerical modeling (hollow sym-
sure by a decrement of 10 kPa in the study). The sup- bols). The relationships between the C/D ratio and the
porting pressure was reduced further until an error lower bound of the overload factor, (OF)L, for single-
message indicating bad geometry of mesh appeared. tunnel (Lee et al., 1999), and the upper bound of over-
The message of the bad geometry of mesh implies a dra- load factor, (OF)L for single-tunnel and parallel tunnels
matic increase in the displacement within the mesh, (Wu and Lee, 2003), are also displayed in Fig. 5. The
therefore, the supporting pressure at this step is defined test results from Mair (1979) are also included in this fig-
as the collapse supporting pressure, (pnum)c, which is ure. The increase in (OF)c with the C/D ratio illustrates
determined from the numerical modeling. By substitut- that the stability of the tunnel improves if the tunnel is
ing (pnum)c into Eqs. (1) and (2), the overload factor, embedded more deeply. In addition, Fig. 5 also shows
(OFnum)c, at collapse and the load factor (LF)num, that the stability of parallel-tunnel is worse than that
respectively, can be determined. The relationship be- of single tunnel; lower overload factors at collapse were
tween the overload factor and the maximum surface set- found in the parallel-tunnel model.
tlement obtained from the centrifuge modeling and that The overload factors at collapse (solid symbols) ob-
computed with the numerical modeling for the single- tained from the centrifuge model for both the single-tun-
tunnel and parallel-tunnel models are in good agreement nel and parallel-tunnel models have nearly the same
before tunnel collapse, as shown in Fig. 4. values as those derived from the corresponding numeri-
Numerical modeling can be used to examine defor- cal model (hollow symbols), as shown in Fig. 5. The val-
mation more precisely at small strain levels and at more ues of (OF)c and (OFnum)c for the single-tunnel model
locations than can be achieved with centrifuge modeling, are all confined by upper and lower bounds. The values
but is not as precise as the measurements of the failure of (OF)c and (OFnum)c for the parallel-tunnel model are
and post-failure behavior that the centrifuge modeling

6
Overload factor at collapse, (OF)c

Overload factor, (OF) Upper bound

(single - tunnel)
0 1 2 3 4 5 6 7 8 9 10 11 12 5 (Wu and Lee, 2003)
0.0
Settlement in prototype, (m)

4
0.5
3 Upper bound
(parallel- tunnel)
1.0 (Wu and Lee, 2003)
2 Single - tunnel
Parallel -tunnel
1.5 Lower bound centrifuge model of single - tunnel (Mair, 1979)
Test7 (C/D=3) 1 (single - tunnel) Numerical solution (single - tunnel)
NTest7 (C/D=3) (Lee et al. , 1999) Numerical solution (parallel- tunnel)
2.0 Twin3 (C/D=3, d/D=1.5) 0
NTwin3 (C/D=3, d/D=1.5) 0 1 2 3 4
2.5 Cover-to-diameter ratio, C/D

Fig. 4. Relations between the overload factor and maximum surface Fig. 5. Comparison of the (OF)c values obtained from numerical and
settlement derived from the numerical and centrifuge modeling. centrifuge modeling.
124 C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119–132

also confined by the upper bound, but no lower bound ing load factors. The shape of the settlement trough
solution has yet been derived. The centrifuge and for the single tunnel approximates closely to that of
numerical modeling provide consistent evaluations of the error function. An empirical approach derived
the tunnel stability. from centrifuge modeling has been proposed for calcu-
lating the surface and subsurface settlement troughs
3.2. Surface settlement troughs for various ground losses due to tunneling in soft clay
and in sandy soils (Wu and Lee, 2003; Lee et al.,
Figs. 6 and 7 compare the computed (represented 2004). In addition, these researchers proposed a super-
by lines) and measured surface settlement troughs imposition method for estimating the surface settle-
(represented by symbols) at selected load factors for ment troughs caused by parallel tunneling from the
the single-tunnel model (C/D = 2) and for the paral- parameters obtained for single tunneling. Their meth-
lel-tunnel model (C/D = 1, d/D = 1.5), respectively. odology for predicting the settlement troughs based on
The distances, X, offset from the tunnel center-line the ground loss was verified by comparison with 12
(or from the center-line of the two tunnels in the par- sets of monitored field data (Wu and Lee, 2003).
allel-tunnel tests) and the surface settlements, S, are
both normalized with respect to the tunnel diameter, 3.3. Comparison of the excess pore water pressures
D. The computed settlement troughs compare reason- obtained from centrifuge and numerical modeling
ably well both in shape and magnitude with those
measured in the centrifuge models at the correspond- In an undrained system, the stress changes
(Dr1, Dr2, Dr3) due to tunneling would generate excess
pore water pressure within the soil mass. With the ap-
proach suggested by Henkel, the changes in the pore
Normalized distance from centre-line, (X/D) water pressure, Du, can be determined with the equation
-8 -6 -4 -2 0 2 4 6 8
0.00 1 2
Du ¼ ðDr1 þ Dr2 þ Dr3 Þ þ a½ðDr1
Dr2 Þ
3
Settlement ratio, (S/D)

0.02 þ ðDr2
Dr3 Þ2 þ ðDr3
Dr1 Þ2 1=2 ; ð3Þ

where a is HenkelÕs pore water pressure parameter, and
0.04 NTest9 (C/D=2) Test9 (C/D=2)
Dr1, Dr2, Dr3 are the changes in the major, intermediate,
LF=0.567 LF=0.570 and minor principle stresses, respectively. A comparison
0.06 LF=0.646 LF=0.651
LF=0.730 LF=0.733 of a and SkemptonÕs parameter, A, derived from triaxial
LF=0.814 LF=0.814 compression tests gives
0.08 LF=0.893 LF=0.896
LF=1.0 LF=1.0  
1 1
0.10 a ¼ pffiffiffi A
. ð4Þ
2 3
Fig. 6. Comparison of surface settlement troughs at various load
factors obtained from numerical and centrifuge modeling for a single This definition is useful because it enables the predic-
tunnel. tion of the Du associated with loading conditions un-
der plane strain conditions if we assume that the soil
bed is an isotropic and elastic-plastic material. By
Normalized distance from centre-line of parallel tunnels, (X/D) using a value of a of 0.3 (determined from the triaxial
-8 -6 -4 -2 0 2 4 6 8 compression tests) and Dr2 = 1/2(Dr1 + Dr3) for the
0.00 plane strain and undrained conditions, the developed
excess pore water pressure can be estimated after the
Settlement ratio, (S/D)

0.02 changes in total stress are calculated in the numerical

simulations.
0.04 NTwin4(C/D=1, d/D=1.5) Twin4(C/D=1, d/D=1.5) In the centrifuge modeling, the stress changes on a
LF=0.369 LF=0.374 soil element during tunneling cannot be measured,
0.06 LF=0.493 LF=0.503 whereas the changes of pore water pressure can be di-
LF=0.616 LF=0.615
LF=0.739 LF=0.738
rectly measured. The numerical analysis in terms of total
0.08 LF=0.862 LF=0.880 stress approach can give the changes of total stress but it
LF=0.985 LF=1.0
cannot give the changes of pore water pressure. How-
0.10 ever, the changes of pore water pressure can be esti-
Fig. 7. Comparison of surface settlement troughs at various load
mated using Eq. (3) after the total stress changes were
factors obtained from numerical and centrifuge modeling for two obtained from the numerical analysis. In this study,
parallel tunnels. the excess pore water pressure ratio, (Du/rvo), defined
 C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119–132 125

as the excess pore water pressure normalized with the to- pore water pressures were measured at the points repre-
tal overburden pressure of the measured point, is used to sented by solid circle symbols in Figs. 8 and 9 (i.e., A0,
follow the variations of the excess pore water pressure A3, B1, C1, and C3 in Fig. 8(a); A0, B2, C1, D1, and D2
during the tunneling simulations. The method for com- in Fig. 8(a)).
parison of the measured and calculated (Du/rvo) is de- Figs. 10 and 11 compare the variations in the mea-
scribed as follows. sured (represented by lines) and computed (repre-
The effects of tunneling on the total stress within the sented by hollow circle symbols) Du/rvo with the
soil mass around a tunnel can be estimated from the overload factor at the corresponding locations for
computed responses at the grids of points shown in Test7 (C/D = 3) and for Twin3 (C/D = 3. d/D = 1.5),
Fig. 8(a) (Test7, C/D = 3), Fig. 8(b) (Test9, C/D = 2), respectively. There is very reasonable agreement be-
and Fig. 8(c) (Test8, C/D = 1) for the single-tunnel tests tween the measured and computed excess pore water
and from those shown in Figs. 9(a) (Twin3; C/D = 3, pressures before tunnel collapse. The small discrep-
d/D = 1.5) and 9(b) (Twin2; C/D = 2, d/D = 1.5) for ancy between them may result from pore water pres-
the parallel-tunnel tests. In the centrifuge model, the sure dissipation caused by partial drainage due to
the appearance of micro-cracks around the small holes
into which the pore water pressure transducers were
inserted. Results of similar consistency were also
Unit: m
Line A Line B Line C Line D Line E x found in the tests with different C/D and d/D ratios.
1.78 Thus the stress states computed from the numerical
A5 experiments can be and are used to investigate the
B5 C5 D5
Negative 5.89
arching zone
arching behavior of the soil mass around the tunnel
A4 in the next section, although the stress measurements
18 B4 C4 D4 E4
Positive 6.04 in the centrifuge model tests were impossible.
arching zone
A3
B3 C3 D3 E3
A2 Plastic zone 7.29
B2 C2 D2 E2
C.L.
6 Spring line B1 D1
C1 E1 Unit: m x
7.29 Line A Line B Line C Line D Line E Line F

1.78
A0
B0 C0 D0 A4 B4 C4 D4
z 5.89
(a) 7.5 3 3 5.5 12
A3 B3 C3 D3 E3
Negative Positive 6.04
Unit: m Arching zone Arching zone
Line A Line B Line C Line D Line E
x
3.44 A2 B2 C2 D2 E2 F2

A3 C.L. 7.29
Negative B3 C3 D3
12 arching zone 4.23 Plastic zone
A2 C1
B2 C2 D2 E2 A1 Spring line D1 E1 F1
Positive 7.42
Plastic zone arching zone 7.33
A0
6 C.L. Spring line
B1 C1 D1 E1 z
(a) 4.5 7.5 6 7 8
(b) z 7.5 4.5 4.5 6.5
Unit: m
Line A Line B Line C Line D Line E x
Unit: m A3 1.24
Line A Line B Line C Line D x B3 C3 D3
1.71 Negative 6.43
A2 12 arching zone
Negative B2 C2 D2
6 arching zone A2
B2 C2 D2 E2
Positive 7.29 Positive
arching zone
Plastic
arching zone 7.33
zone
6 Spring line 6 A1 Spring line
C.L. B1 C1 D1 C1 D1 E1
C.L. 7.5
z z
(c) 7.5 6 5.5 A0

(b) 4.5 7.5 6 7

Fig. 8. (a) Locations of reference soil elements and the arching zone
for Test7 (C/D = 3). (b) Locations of reference soil elements and the Fig. 9. (a) Locations of reference soil elements and the arching zone
arching zone for Test9 (C/D = 2). (c) Locations of reference soil for Twin3 (C/D = 3, d/D = 1.5). (b) Locations of reference soil
elements and the arching zone for Test8 (C/D = 1). elements and the arching zone for Twin2 (C/D = 2, d/D = 1.5).
126 C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119–132

0.4 0.4

Test7 (C/D = 3) B1 0.3 C1

0.3
Measured Measured
0.2 0.2 Calculated
(∆u / σ vo )

Calculated

(∆u / σ vo )
0.1 0.1

0.0 0.0

-0.1 -0.1
(OF)c (OF)c
-0.2 -0.2
0 2 4 6 8 10 12 0 2 4 6 8 10 12
(a) Overload factor, (OF ) (b) Overload factor, (OF )

0.4 0.4
C3 A3
0.3 0.3
Measured Measured
0.2 Calculated 0.2 Calculated
(∆u / σ vo )

(∆u / σ vo )
0.1 0.1

0.0 0.0

-0.1 -0.1
(OF)c (OF)c
-0.2 -0.2
0 2 4 6 8 10 12 0 2 4 6 8 10 12
(c) Overload factor, (OF ) (d) Overload factor, (OF)

0.4
A0
0.3
Measured
0.2 Calculated
(∆u / σ vo )

0.1

0.0

-0.1
(OF)c
-0.2
0 2 4 6 8 10 12
(e) Overload factor, (OF)

Fig. 10. Comparison of measured and calculated excess pore water pressures for Test7 (C/D = 3).

4. Mechanism of arching around tunnels Drv

ARð%Þ ¼  100; ð5Þ
rvo
4.1. Evolution of arching effect
where Drv is the change in the vertical stress during tun-
The effects of tunneling on the total stress within the neling and rvo is the total overburden pressure. In this
soil mass around a tunnel can be estimated from the study, we used the arching ratio to describe the arching
computed responses at the grids of points shown in behavior quantitatively at various locations.
Fig. 8(a) (Test7, C/D = 3), Fig. 8(b) (Test9, C/D = 2), An element that receives higher load transfers from
and Fig. 8(c) (Test8, C/D = 1) for the single-tunnel tests adjoining yielding or flexible elements will generate a
and from those shown in Figs. 9(a) (Twin3; C/D = 3, larger positive arching ratio. Conversely, a negative
d/D = 1.5) and 9(b) (Twin2; C/D = 2, d/D = 1.5) for arching ratio will arise if an element shifts load to
the parallel-tunnel tests. A more detailed understanding non-yielding parts or to more rigid elements. Fig. 12
of the stress transfer in a tunneling problem from shows a plot of the arching ratio versus overload factor
moving parts of the soil (settle more) to adjacent parts for NTest7 (C/D = 3) at the grid of points shown in
(settle less) can be achieved by considering the vertical Fig. 8(a). As the overload factor increases, all of the ele-
stress redistributions in the soil mass above the spring ments on the vertical center-line (Line A) experience a
line. The arching ratio is defined as decrease in vertical stress (negative AR) but the elements
 C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119–132 127

0.4 0.4
Twin3 (C/D=3, d/D=1.5) C1 D1
0.3 0.3
Measured Measured
0.2 calculated 0.2 calculated
(∆u /σvo )

(∆u / σvo)
0.1 0.1

0.0 0.0

-0.1 -0.1
(OF)c (OF)c
-0.2 -0.2
0 2 4 6 8 10 12 0 2 4 6 8 10 12
(a) Overload factor, (OF) (b) Overload factor, (OF)

0.4 0.4

0.3 D2 0.3 B2
Measured Measured
0.2 0.2
(∆u /σvo )

calculated calculated

(∆u /σvo )
0.1 0.1

0.0 0.0

-0.1 -0.1
(OF)c (OF)c
-0.2 -0.2
0 2 4 6 8 10 12 0 2 4 6 8 10 12
(c) Overload factor, (OF) (d) Overload factor, (OF)

0.4

0.3 A0

0.2 Measured
(∆u /σvo )

calculated
0.1

0.0

-0.1
(OF)c
-0.2
0 2 4 6 8 10 12
(e) Overload factor, (OF)

Fig. 11. Comparison of measured and calculated excess pore water pressures for Twin3 (C/D = 2, d/D = 1.5).

on the spring line (B1–E1) experience an increase in ver- ative as the overload factor increases for the elements on
tical stress (positive AR), as shown in Fig. 12(a). Simi- Line A (Fig. 13(a)).
larly, as can be seen in Fig. 12(b) and (c), the Similarly, Fig. 14 summarizes the arching ratio as a
magnitude of the arching ratio is also related to the dis- function of the overload factor for the elements on the
tance offset from the tunnel center and to the overload spring line (B1–E1) and on the vertical center-line
factor (comparing the arching ratios on Lines A, B, C, (A2–A3) shown in Fig. 8(b) for NTest9 (C/D = 2).
D, and E). Fig. 13 summarizes the changes in the Fig. 15 summarizes the arching ratio as a function
arching ratio for three overload factors (OF = 1, 3, and of the overload factor for the elements on the spring
(OFnum)c) for the elements on Lines A to E for NTest7 line (B1–D1) and on the vertical center-line (A2)
(C/D = 3). The magnitude of the positive arching ratio shown in Fig. 8(c) for NTest8 (C/D = 1). The evolu-
increases with increases in the overload factor for the tion of the arching ratio in the shallower tunneling
elements along Line B but declines rapidly once the ele- shown in these two figures is similar to those pre-
ment is yielding (Fig. 13(b)). The element at C1 experi- sented in Figs. 12 and 13 but smaller positive and neg-
ences the largest positive arching ratio (about a 9% ative arching ratios are obtained at the corresponding
rise). The elements located at a greater distance from points for deeper tunneling (NTest7, C/D = 3). Shal-
the tunnel center experience smaller positive arching lower tunneling imposes a larger arching effect on
ratios. In contrast, the arching ratio becomes more neg- the surrounding soil mass.
128 C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119–132

10 (C/D = 3, d/D = 1.5). Fig. 17 shows the changes in the

arching ratio for the elements on Lines A–F (Fig. 9(b))
NTest7 (C/D=3)
5 for NTwin2 (C/D = 2, d/D = 1.5). A comparison of
the changes in the arching ratios shown in Figs. 13
Arching ratio, (%)

and 16 indicates that the arching ratios in parallel tun-

0
neling evolve in nearly the same way as in single tunnel-
ing, except at Element A1, which is located on Line A
-5 A0 and on the spring line. At this point, a larger positive
A3 arching ratio develops initially, but its value rapidly de-
A4 clines to near zero once the stress state becomes yielding,
-10
A5
B1
in the same manner as the other elements (Fig. 16(a)).
C1 This part of the soil mass takes the load that is trans-
-15
D1 ferred from the two compressive arches above the tun-
E1 (OF)c nels due to parallel tunneling, and behaves like a
-20 pillar. After examining in detail the changes in the arch-
0 1 2 3 4 5
ing ratio shown in Figs. 13–15 for the single-tunnel tests
(a) Overload factor, (OF)
and in Figs. 16 and 17 for the parallel-tunnel tests, we
10
B0 reached the same conclusion for parallel tunneling as
NTest7 (C/D=3)
B1 obtained for single tunneling, namely that shallower
8 B3 tunneling imposes a larger arching effect on the sur-
B4
rounding soil mass because of the higher arching ratio.
Arching ratio, (%)

B5
6 C0
In the centrifuge modeling, the excess pore water
C1 pressure ratio will increase on an element in the soil
C3 bed that receives load transfers from adjoining yielding
4
C4 or flexible elements. Conversely, the excess pore water
C5
pressure ratio will decrease if an element shifts load to
2 non-yielding parts or to more rigid elements. Therefore,
the changes in the measured excess pore water pressure
0 ratio can also be used to track the load transfers among
the elements during tunneling simulation in the centri-
(OF)c
-2
fuge modeling. For example, the excess pore water pres-
0 1 2 3 4 5 sure ratios shown in Figs. 14 and 15 initially rise with
(b) Overload factor, (OF) increases in the overload factor, but later fall as the ele-
10 ments in non-yielding states progress to yielding states
NTest7 (C/D=3) during increases in the overload factor.
8
D0
4.2. Boundaries of the arching and plastic zones in the
D1 tunnel collapse stage
Arching ratio, (%)

6
D3
D4 As discussed in the previous section, the outer bound-
D5
4 aries of the arching and plastic zones expand outward
E1
E3
from the excavated area as the overload factor increases.
2 E4 Knowledge of the boundaries of the arching and plastic
zones in the tunnel collapse stage is thus crucial for engi-
neering practice, and is now discussed.
0
Fig. 18 presents the variations in the shear stress ra-
(OF)c tio, q/su, at (OFnum)c for the soil elements on the selected
-2 lines at the elevations of 3.5, 7.29, 11, and 16 m above
0 1 2 3 4 5
the spring line (NTest7, C/D = 3). Here q =
(c) Overload factor, (OF )
1/2(r1
r3). The arrows shown in Fig. 18 indicate the
Fig. 12. Arching ratio versus overload factor at various locations positions of the outermost boundary of the plastic zone
(NTest7 C/D = 3). at the elevations where the elements have stress states of
su/q = 1 and are regarded as yielding. As can be seen in
Fig. 16 summarizes the changes in the arching ratio the figure, the plastic boundaries extend from the tunnel
for the elements on Lines A–F (Fig. 9(a)) for three over- center-line as far as 12 m along the spring line (about
load factors (OF = 1.5, 2.5, and (OFnum)c) for NTwin3 twice the tunnel diameter) and but only to 10 m at an
 C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119–132 129

25 25 25
OF=1 NTest7 (C/D=3) Line A NTest7 (C/D=3) Line B NTest7 (C/D=7) Line C
20 20 OF=1 20 OF=1
Distance from spring line, (m)

OF=3
OF=3 OF=3
(OFnum)c
15 15 (OFnum)c 15 (OFnum)c

10 10 10

5 5 5
Spring line
0 0 0

-5 -5 -5

-10 -10 -10

-20 -15 -10 -5 0 5 10 -20 -15 -10 -5 0 5 10 -20 -15 -10 -5 0 5 10
(a) Arching ratio, (%) (b) Arching ratio, (%) (c) Arching ratio, (%)

25 25
NTest7 (C/D=3) Line E
OF=1 NTest7 (C/D=3) Line D OF=1
20 20
Distance from spring line, (m)

OF=3 OF=3
(OFnum)c (OFnum)c
15 15

10 10

5 5

Spring line
0 0

-5 -5

-10 -10
-20 -15 -10 -5 0 5 10 -20 -15 -10 -5 0 5 10
(d) (e) Arching ratio, (%)
Arching ratio, (%)

Fig. 13. Changes in arching ratio in various overload factors along; (a) Line A; (b) Line B; (c) Line C; (d) Line D; (e) Line E (NTest7 C/D = 3).

20 35

15 NTest9 (C/D=2) 30

10 25
Arching ratio, (%)
Arching ratio,(%)

5 20
B1
0 15 C1
-5 D1
10
A2
B1
-10 5
C1
-15 D1
0
E1
-20 A2 -5
A3
-25 -10
(OF)c (OF)c
-30 -15
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 0 1 2 3 4
Overload factor, (OF) Overload factor, (OF)

Fig. 14. Arching ratio versus overload factor at various locations Fig. 15. Arching ratio versus overload factor at various locations
(NTest9 C/D = 2). (NTest8 C/D = 1).

elevation 3.5 m above the spring line. Therefore, the out- dashes. Fig. 8(b) and (c) also display the outer bound-
er boundary of the plastic zone in the tunnel collapse aries of the plastic zone for the single-tunnel models
stage can be inferred by examining all the elements re- (C/D = 1, 2).
garded as yielding (not just those considered in Similarly, Fig. 19 presents the variations of the shear
Fig. 18), and is depicted in Fig. 8(a) as a thick line of stress ratio, q/su, at (OFnum)c for the soil elements on the
130 C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119–132

20 20 20
NTwin3 Line A NTwin3 Line B
(C/D=3, d/D=1.5) OF=1.5 OF=1.5 NTwin3 Line C
(C/D=3, d/D=1.5)
Distance from spring line, (m)

15 15 15 (C/D=3, d/D=1.5)
OF=2.5 OF=2.5
(OFnum)c (OFnum)c
10 10 10

5 5 5

spring line spring line spring line

0 0 0
OF=1.5
-5 -5 -5 OF=2.5
(OFnum)c
-10 -10 -10
-30 -20 -10 0 10 20 -30 -20 -10 0 10 20 -30 -20 -10 0 10 20
(a) Arching ratio, (%) Arching ratio, (%) Arching ratio, (%)
(b) (c)
20 20 20
NTwin3 Line D NTwin3 Line F
(C/D=3, d/D=1.5) NTwin3 Line E
Distance from spring line, (m)

15 15 (C/D=3, d/D=1.5) 15 (C/D=3, d/D=1.5)

10 10 10

5 5 5

spring line spring line spring line

0 0 0
OF=1.5 OF=1.5 OF=1.5
-5 OF=2.5 -5 OF=2.5 -5 OF=2.5
(OFnum)c (OFnum)c (OFnum)c
-10 -10 -10
-30 -20 -10 0 10 20 -30 -20 -10 0 10 20 -30 -20 -10 0 10 20
(d) Arching ratio, (%) (e) Arching ratio, (%) (f) Arching ratio, (%)

Fig. 16. Changes in arching ratio in various overload factors along; (a) Line A; (b) Line B; (c) Line C; (d) Line D; (e) Line E; (f) Line F (NTwin3
C/D = 3; d/D = 1.5).

selected lines at elevations of 3.5, 7.29, 11 and 16 m parallel tunnels only generate a slightly wider arching
above the spring line (NTwin3, C/D = 3, d/D = 1.5). zone for the same burial depth. The extents of the arch-
As can be seen in the figure, the boundaries of the plastic ing zones for a single-tunnel and parallel-tunnels embed-
zone extend from the center-line of the two tunnels as ded at various depths (C/D = 1, 2, 3) are shown in
far as 15 m along the spring line and 8 m at an elevation Fig. 20(a) and (b), respectively. The curves representing
of 3.5 m above the spring line. The outer boundary can the outer bounds of the arching zones were obtained as
also be determined with the procedure described in the follows:
previous paragraph, as shown in Figs. 9(a) (Twin3)    
C
and 9(b) (Twin2) (thick lines of dashes). z ¼ 0.1 exp 0.305x
0.045 x
D
The non-yielding elements in the arching zone receive
a load transfer from the elements in the plastic zone. The for single
tunnel; ð6aÞ
   
area lying between the outer boundary of the plastic C
zone and the boundary at which the elements have an z ¼ 0.105 exp 0.27x
0.0375 x
D
arching ratio larger than 1% is regarded as the arching
for two parallel
tunnels ðd=D ¼ 1.5Þ; ð6bÞ
zone. Thus the outer boundaries of the single-tunneling
arching zones can easily be determined, and are shown in which x is the distance offset from the tunnel center
as thick lines in Fig. 8(a)–(c) for various C/D ratios. for the single-tunnel and the distance offset from the
The arching mechanism in the case of two parallel- center-line of the two parallel-tunnels, and z is the depth.
tunnels is similar to that for a single-tunnel, so the same
procedure can be used to infer the boundaries of the 4.3. Positive and negative arching zones
arching zone. They are depicted in Fig. 9(a) and (b)
for Twin3 and Twin2. After examining the variations in the arching ratios
Tunneling at different burial depths can result in of the elements surrounding the tunnel (or tunnels), as
arching and plastic zones of different extents. Deep sin- shown in Figs. 13, 16, and 17, it was found that each
gle tunneling will result in a wider arching zone, but two arching zone can be divided into two zones, the positive
 C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119–132 131

15 15 15
NTwin2 Line A NTwin2 Line B NTwin2 Line C
(C/D=2, d/D=1.5) (C/D=2, d/D=1.5) (C/D=2, d/D=1.5)
Distance from spring line, (m)

10 10 10
OF=0.97
OF=1.95
5 (OFnum)c 5 5

spring line spring line spring line

0 0 0
(OF)=0.97
OF=0.97
OF=1.95
-5 -5 -5 OF=1.95
(OFnum)c
(OFnum)c

-10 -10 -10

-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30

(a) Arching ratio, (%) (b) Arching ratio, (%) (c) Arching ratio, (%)

15 15
NTwin2 Line D NTwin2 Line E
Distance from spring line, (m)

(C/D=2, d/D=1.5) (C/D=2, d/D=1.5)

10 10

5 5

spring line spring line

0 0

OF=0.97 OF=0.97
-5 OF=1.95 -5 OF=1.95
(OFnum)c (OFnum)c
-10 -10
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30

(d) Arching ratio, (%) (e) Archinf ratio, (%)

Fig. 17. Changes in arching ratio in various overload factors along; (a) Line A; (b) Line B; (c) Line C; (d) Line D; (e) Line E (NTwin2 C/D = 2;
d/D = 1.5).

1.2 1.2
Ratio of shear stress, (q/s u )

NTest7 (C/D=3)
spring line NTwin3 (C/D=3, d/D=1.5)
Ratio of Shear stress,q/s u

1.0 1.0 spring line

3.5m above spring line
3.5m above spring line
0.8 7.29m above spring line
0.8 7.29m above spring line
11m above spring line
0.6 16m above spring line 11m above spring line
0.6
16m above spring line
0.4 0.4
0.2
0.2
0.0
0.0
0 10 20 30 40
0 10 20 30 40
Distance from centerline of tunnel, (m)
Distance from center line of two parallel tunnels, (m)
Fig. 18. Variations in the shear stress ratio at (OFnum)c for soil
elements along lines at various elevations above the spring line for a Fig. 19. Variations in the shear stress ratio at (OFnum)c for soil
single tunnel (NTest7 C/D = 3). elements along lines at various elevations above the spring line for two
parallel tunnels (NTwin3 C/D = 3 d/D = 1.5).

arching zone (with positive AR) and the negative arch-

ing zone (with negative AR). The boundaries (dash– can be determined from the upper bound solution using
dot lines) of the two zones derived from the numerical the collapse mechanism proposed by Wu and Lee
analysis are shown in Fig. 8(a)–(c) for the single-tunnel (2003). The sliding wedges are reasonably consistent
with various C/D ratios and in Fig. 9(a) and (b) for with the boundaries of the negative arching zones. A pile
two parallel-tunnels with C/D = 2, 3 and d/D = 1.5. embedded in the negative arching zone would partially
The shaded rectangular regions in these figures corre- lose both end bearing capacity and skin friction on the
spond to the half-width of the sliding wedge, which pile body due to the reduction of the vertical stresses
132 C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119–132

Distance from tunnel center, (m) Distance from centerline of two parallel tunnels, (m)
0 10 20 30 0 10 20 30
0 0

5 5
Single-tunnel Paralle-tunnel
Depth, (m)

Depth, (m)
10 10
C/D=1 C/D=1, d/D=1.5
15 C/D=2 15
C/D=2, d/D=1.5
C/D=3 C/D=3, d/D=1.5
20 Fitting curve 20
Fitting curve
z=0.1exp{0.305x-0.045(C/D)x] z=0.105exp[0.27x-0.0375(C/D)x]
25 25
(a) (b)

Fig. 20. Extents of the arching zones for single-tunnel and parallel-tunnel (d/D = 1.5).

and would experience a large amount of settlement dur- Chen, C.Y., Martin, G.R., 2002. Soil–structure interaction for landslide
ing new nearby tunneling. stabilizing pile. Computer and Geotechnics 29 (5), 363–386.
Cundall, P.A., Coetzee, M.J., Hart, R.D., Varona, P.M., 1993. FLAC
UserÕs Manual. Itasca Consulting Group, USA.
Handy, R.L., 1983. The arch in soil arching. Journal of Geotechnical
5. Summaries and conclusions Engineering, ASCE 111 (3), 302–318.
Koutsabeloulis, N.C., Griffiths, D.V., 1989. Numerical modeling of the
A series of centrifuge model tests were carried out to trap door problem. Geotechnique 39 (1), 77–89.
investigate the surface settlement trough, excess pore Ladanyi, B., Hoyaux, B., 1969. A study of the trap-door problem in a
granular mass. Canadian Geotechnical Journal 6 (1), 1–11.
water pressure generation, and tunnel stability of the
Lee, C.J., Wu, B.R., Chiou, S.Y., 1999. Soil movements around a
tunnels with various C/D ratios (single-tunnel) and d/ tunnel in soft soils. Proceedings of the National Science Council,
D ratios (two parallel-tunnels). Numerical analysis was Part A: Physical Science and Engineering 23 (2), 235–247.
also conducted to evaluate the tunnel stability and arch- Lee, C.J., Chiang, K.H., 2004. Load transfer on single pile near new
ing effects that develop during tunneling in soft clayey tunneling in sandy ground. In: Matsui, Tanaka, Mimura (Eds),
Proceedings of the International Symposium on Engineering
soil. The centrifuge and numerical modeling produced
Practice and Performance of Soft Deposits (IS-OSAKA 2004),
consistent results in their predictions of the surface set- pp. 495–506.
tlement trough, excess pore water generation, and the Lee, C.J., Chiang, K.H., Kou, C.M., 2004. Ground movement and
overload factors at collapse for both single tunneling tunnel stability when tunneling in sandy ground. Journal of the
and parallel tunneling. An arching ratio derived from Chinese Institute of Engineers 27 (7), 1021–1032.
Lee, K.M., Rowe, R.K., 1991. An analysis of three-dimensional
the numerical analysis was defined to describe the evolu-
ground movements: the thunder bay tunnel. Canadian Geotechni-
tion of the arching effect in the soil mass surrounding the cal Journal 28 (1), 25–41.
tunnels. The boundaries of the arching zones for both Mair, R.J., 1979. Centrifugal modeling of tunnel construction in soft
single tunneling and parallel tunneling were determined. clay. Ph.D Thesis, University of Cambridge, UK.
In addition, the boundaries of the positive and negative Mair, R.J., Gunn, M.J., OÕReilly, M.P., 1981. Ground movements
around shallow tunnels in soft clay. In: Proceedings of the 10th
arching zones were also proposed. Construction engi-
International Conference on Soil Mechanics and Foundation
neers can easily locate the boundaries of the positive Engineering, pp. 323–328.
and negative arching zones and then take appropriate Nakai, T., Xu, L., Yamazaki, H., 1997. 3D and 2D model tests and
measures to mitigate possible damages to underground numerical analyses of settlements and earth pressures due to tunnel
structures due to new nearby tunneling. excavation. Soils and Foundations 37 (3), 31–41.
Park, S.H., Adachi, T., 2002. Laboratory model tests and FE analyses
on tunneling in the unconsolidated ground with inclined layers.
Tunneling and Underground Space Technology 17, 181–193.
Acknowledgments Peck, R.B., 1969. Deep excavation and tunneling in soft ground. In:
Proceedings of 7th International Conference on Soil Mechanics
The financial support provided by the National Science and Foundation Engineering, Mexico, State of the Art Volume, pp.
Council, Taiwan, under Grants NSC 87-2211-E-008-024 225–290.
and NSC 91-2211-E-008-026 is gratefully acknowledged. Stone, K.J.L., Newson, T.A., 2002. Arching effects in soil–structure
interaction. In: Phillips, Guo, Popescu (Eds.), Physical Modeling in
Geotechnics: ICPMG Õ02, pp. 935–939.
Terzaghi, K., 1943. Theoretical Soil Mechanics. Wiley, New York.
References Wang, W.L., Yen, B.C., 1973. Soil arching in slopes. Journal of the
Geotechnical Engineering, ASCE 100 (1), 61–78.
Clough, G.W., Schmidt, B., 1981. Design and performance of Wu, B.R., Lee, C.J., 2003. Ground movements and collapse mecha-
excavation and tunnels in soft clay. In: Soft Clay Engineering. nisms induced by tunneling in clayey soil. International Journal of
Elsevier, Amsterdam, pp. 600–634 (Chapter 8). Physical Modelling in Geotechnics 3 (4), 13–27.