Tunnelling and Underground Space Technology

Tunnelling and Underground Space Technology 21 (2006) 119132

incorporating Trenchless Technology Research

www.elsevier.com/locate/tust

Tunnel stability and arching eects during tunneling in soft clayey soil
C.J. Lee
a b

a,*

, B.R. Wu b, H.T. Chen a, K.H. Chiang

Department of Civil Engineering, National Central University, No. 300, Jung-da Rd., Chungli, Taoyuan 32054, Taiwan 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 eects 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 eects 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 eect; Tunnel stability; Centrifuge modeling; Numerical modeling

1. Introduction Tunneling in soft clayey soils has become very popular in recent years because it is one of the best construction methods for building mass rapid transit systems and sewage collection systems in densely populated cities. As the face of a tunnel is advanced, a means of supporting the ground close to the face may be needed; without such support, collapse might occur due to gross plastic deformation of the soil. Moreover, tunneling inevitably induces varying degrees of ground movement towards the tunnel opening and results in detrimental eects on nearby facilities, such as shallow foundations, piles, existing tunnels and other pipeline systems. Taking appropriate measures to protect nearby facilities before excavation is an important part of engineering practice. The predicCorresponding author. Tel.: +886 3 4227151x34135; fax: +886 3 4252960. E-mail address: cjleeciv@cc.ncu.edu.tw (C.J. Lee). 0886-7798/$ - see front matter 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.tust.2005.06.003
*

tion of tunneling-induced ground movements during excavation of soft ground tunnel has been carried out using various methods, including empirical methods derived from eld observations (Peck, 1969; Clough and Schmidt, 1981) and centrifuge modeling (Mair et al., 1981; Wu and Lee, 2003; Lee et al., 2004), or numerical and analytical methods (Lee and Rowe, 1991). Terzaghi (1943) explained how stress transfer from yielding parts of a soil mass to adjacent non-yielding parts leads to the formation of an arching zone. This problem has two modes of displacement, depending on whether the trap door is translated into the soil (passive mode) or away from it (active mode). The passive mode can be used for the evaluation of the uplift force of anchors, or of any buried structure that can be idealized as an anchor. The active mode can be used to study the gravitational ow of granular material between vertical walls (the silo problem) or the ground pressure on tunnel liners. Ladanyi and Hoyaux (1969) performed a series of model trap-door tests

120

C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119132

under 1 g conditions in order to check the validity of the classic bin theory. Handy (1983) analyzed soil arching action behind retaining walls, and Wang and Yen (1973) carried out this analysis for slopes. Nakai et al. (1997) performed a series of physical model tests under 1 g conditions and carried out numerical analysis of these tests to investigate the arching eect. They found that the results obtained from the model tests were in good agreement with those obtained from the numerical analysis. Park and Adachi (2002) performed model tests under 1 g conditions to simulate tunneling events in unconsolidated ground with various levels of inclined layers. They found that remarkable non-symmetrical distributions of the earth pressure arose when a tunneling event took place in inclined layers with 60 of inclination. Stone and Newson (2002) presented the results of a series of centrifuge tests designed to investigate the eects of arching on soilstructure interaction. Koutsabeloulis and Griths (1989) implemented a nite element method to investigate the trap-door problem. The concept of soil arching was recently adopted in the analysis of the mobilization of resistance from passive pile groups subjected to lateral soil movement (Chen and Martin, 2002). When tunneling is conducted in the vicinity of existing pile foundations, the axial load transfer mechanism and failure mode on existing piles vary depending on the distance between the existing piles and the new driving tunnel and relative elevation of the piles with respect to the centerline of the tunnel (Lee and Chiang, 2004). These behaviors result from the complicated redistributions of stress around tunnels during tunneling. Hence, the stress distribution in the vicinity of a tunnel or of several tunnels stacked closely in an underground station needs to be established before appropriate protection measures for nearby existing piles can be implemented. By deepening our understanding of the arching eect in various geotechnical problems, we can improve the design of the protection measures required for existing underground structures nearby new tunneling. 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 a tunnel, and the boundaries of the arching zone during tunneling in clayey soils are investigated and discussed. Firstly, a series of centrifuge model tunnel tests was conducted. A nite dierence program (FLAC) was then chosen for numerical analysis of the system described by the centrifuge model to provide insight into the arching mechanism and the boundaries of the arching zone during tunneling. Finally, the results from the numerical modeling and the measurements from the centrifuge modeling were compared in order to assess their predictions.

2. Centrifuge and numerical modeling 2.1. Centrifuge modeling The basic principle of centrifuge modeling is to recreate the stress conditions that are present in full-scale constructions in models of greatly reduced scale. The full-scale system modeled with a centrifuge model (with dimensions N times larger than those of the model if it is tested in an acceleration that is N times earth gravity) is referred to as the prototype. It is intended that the prototype should include all the important characteristics of the eld situation of interest. Centrifuge modeling provides an opportunity to study for example the ground responses due to tunneling before and after collapse; collapse is of course not permitted to occur in the eld. This experimental study was undertaken in the geotechnical centrifuge at the National Central University. The NCU centrifuge has a nominal radius of 3 m and is capable of accelerating a 1 tonne model package to 100 g and 0.55 tonne to 200 g. In the single-tunnel model tests, one model tunnel, 60 mm in diameter, was embedded at various depths specied by the cover-to-diameter ratio (C/D). In the parallel-tunnel model tests, two model tunnels (60 mm in diameter) were separated by a specied center-to-center distance (d) (as shown in Fig. 1), and buried at various depths specied by the cover-todiameter ratio (C/D). All the model tests reported in this study were carried out under a centrifugal acceleration of 100 g in order to model a prototype tunnel with a diameter of 6 m embedded at depths with the tested C/D ratios. The soil used in the model tests had a plasticity index of 18 and was classied as CL in the Unied Soil Classication System. The soil slurry was remolded at about twice its liquid limit in a mixer and poured into the consolidometer. Consolidation pressure was applied in ve stages, with a nal pressure of 196 kPa. Further details of the soil bed preparation can be found in Wu and

LVDT

C D (60 mm) PPT Marked Tunnel spaghetti d Tunnel deformation gauges (Dimensions are in model scale) 820 mm

Fig. 1. Setup of test package for two parallel tunnels (model scale).

480 mm

C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119132

121

Lee (2003). The basic properties of the prepared soil bed are listed in Table 1. On completion of the consolidation, the soil bed, which had an undrained shear strength prole of 3040 kPa, was lifted and placed in a strong box. The set-up of the test package for the parallel-tunnel model is shown in Fig. 1. Five PPTs were inserted at selected positions to monitor changes in the pore water pressure. The pore pressure transducer (PPT) was inserted into a pre-drilled hole and then the hole was fully back-lled with thick slurry. Eight LVDTs were xed on top of the strong box to record the transverse surface settlements. The test package was rst spun at an acceleration of 100 g for 5 min so that any voids generated during the installation of the PPTs and the assembly of the test package might be lled. After the centrifuge was stopped, one or two 60 mm diameter model tunnels were cut manually, depending on the test conditions, and then rubber bags were inserted into the tunnels. Tunnel deformation gauges consisting of four thin cantilevers made from stainless steel were installed inside the rubber bags to measure the deformations at the crowns, inverts, and side-walls of the tunnels. The test package shown in Fig. 1 was prepared for further tunnel collapse tests by connecting air pressure lines to the rubber bags. The air pressure in the rubber bags was carefully regulated to balance the overburden pressure at the tunnel center during the reacceleration of the model up to a centrifuge acceleration of 100 g. The tunneling event was simulated by simultaneously reducing the air pressure inside the tunnels and eventually down to zero. This air-pressure method of simulating tunnel excavation was adapted from Mair (1979). The air-pressure method was chosen to support the tunnel during the accelerating stages because measuring the tunnel deformation was needed and the waterproong of this measuring device was dicult if an incompressible heavy uid-pressure method was used. The method used in the study may cause smaller surface settlements and larger settlement trough widths at the corresponding supporting pressures but no dierence in the supporting pressure at collapse compared to the incompressible heavy uid-pressure method. No more than 15 min were spent under 100 g prior to collapse (including the accelerating stage). The dissipaTable 1 Basic properties of the prepared soil bed Specic gravity, Gs Liquid limit, LL Plastic limit, PL Plasticity index, PI Unit weight, c (kN/m3) Compression index, Cc Swell index, Cr Coecient of consolidation, Cv (cm2/s) Permeability, k (m/s) 2.67 40 22 18 18.1 0.28 0.0275 0.010524 4.5 109

tion of pore water pressure (or changes of eective stress) during these two stages may be less than 1% in average. A torvane apparatus was used for determining the undrained shear strength at various depths on the side of the soil bed before and after the test. The test results show that no obvious changes in the undrained shear strengths (less than 2 kPa) at the same depths were found. Therefore, we assumed that there was minimal migration of pore water pressure and hence that the soil was subjected to undrained shearing. The changes in the pore water pressures, the tunnel deformations, and the surface settlements induced by tunneling were measured continuously. After each model test, the soil bed was cautiously excavated to expose the implanted spaghetti. Further, six undisturbed samples were taken from the soil bed at selected depths for use in unconned compression tests. The undrained shear strength (su) and the secant Youngs modulus (E50), for each soil bed was determined from the average of the results for the six samples, as listed in Table 2. The surrounding soil squeezes into the tunnel as the supporting air pressure is gradually reduced, which nally causes tunnel collapse. The overload factor (OF), as dened below, is a useful index for describing tunnel stability throughout the entire test process. rvo pi OF ; 1 su where rvo is the overburden pressure at the tunnel center, and pi is the supporting pressure. Fig. 2 shows plots of the surface settlements at the tunnel axis versus OF for the single-tunnel test, Test7 (C/D = 3), and of those at the symmetrical axis of the settlement trough for the parallel-tunnel test, Twin3 (C/D = 3, d/D = 1.5). This

Table 2 Test congurations Test no.a Test11 Test12 Test5 Test8 Twin4 Twin5 Test3 Test9 Twin1 Twin2 Twin9 Twin12 Test6 Test7 Twin3 Twin6 Test10 Twin10 Twin11
a

C/D 0.5 0.5 1 1 1 1 2 2 2 2 2 2 3 3 3 3 4 4 4

d/D 1.5 1.5 3 1.5 1.5 2 1.5 1.5 1.5 1.5

su (kPa) 31.00 35.12 36.90 37.90 33.00 39.10 30.25 35.79 48.70 41.00 39.52 35.10 33.30 34.00 36.10 32.90 32.17 34.25 32.83

Test3Test12: single tunnel; Twin1Twin11: two parallel tunnels.

122

C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119132

Overload factor, (OF) Settlementin prototype scale, (m)

0 0.0
(OF )c =3.2

10

11

12

(OF ) c =3.96

0.5 1.0 1.5 2.0 2.5

Twin3 Test7

Fig. 2. Denition of overload factor at collapse, (OF)c.

gure shows that the surface settlements, S, increase dramatically once OF exceeds a critical value. Extending the straight-line portions of the rst and second parts of the S vs OF curves to intersect at the points shown in Fig. 2, the horizontal ordinate of each of these critical points is dened as the overload factor at collapse, (OF)c. In engineering practice, the tunnel must be supported against collapse during tunneling. Thus, the load factor (LF) is regarded as the reciprocal of the safety factor LF rvo pi OF ; rvo pi c OF c 2

FLAC2D (Cundall et al., 1993). A plane-strain model with large-strain formulation was used to simulate the deformation behavior of the ground surrounding unlined tunneling. In order to compare the results of the numerical and centrifuge modeling, the boundary conditions and soil properties used in the numerical model were chosen to be the same as those studied in the centrifuge model tests. The numerical analysis considered a 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 soil bed used in the centrifuge model tests. The left and right boundaries were xed in the x-direction, and the bottom boundary was xed in the x- and ydirections. 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. The soil bed was treated as an isotropic and elastic perfectly plastic continuum following the MohrCoulomb failure criterion (/ = 0). A total of six numerical models were analyzed. The model conditions and the mechanical properties of soil bed measured from the unconned compression tests used in the numerical analysis are shown in Table 3. The numerical modeling was commenced at a state of geostatic equilibrium, and allowed to come to numerical equilibrium under the force of gravity. This step provides an estimate of the in situ stress in the soil prior

where (pi)c is the measured supporting pressure at collapse in the centrifuge model tests. The value of LF varies 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 parallel-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 nite dierence 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. Ntest8 Ntwin4 Ntest9 Ntwin2 Ntest7 Ntwin3 C/D 1 1 2 2 3 3 d/D 1.5 1.5 1.5 su (kPa) 37.90 33.00 35.79 41.00 34.00 36.10 E50 (kPa) 3500 3500 3500 4000 3500 3500 c (kN/m3) 18.1 18.1 18.1 18.1 18.1 18.1 m 0.49 0.49 0.49 0.49 0.49 0.49

C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119132

123

to tunneling. Secondly, the elements representing the excavated soil in the tunnel were nulled and a uniform supporting pressure equal to the overburden pressure at the center of tunnel was applied to the interior boundary 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 supporting pressure that was applied in the centrifuge model tests. For FLAC, the value of maximum nodal unbalance force is used to determine if a simulation having reached equilibrium. In the current study, an equilibrium state was regarded as having converged when the maximum unbalance force of every node in the mesh was less than 10 N (the ratio of the maximum unbalance force to the overburden pressure on the element was about 0.005%) in the simulation of per step. The simulation was then moved to the next step (reducing the supporting pressure by a decrement of 10 kPa in the study). The supporting pressure was reduced further until an error message indicating bad geometry of mesh appeared. The message of the bad geometry of mesh implies a dramatic increase in the displacement within the mesh, therefore, the supporting pressure at this step is dened as the collapse supporting pressure, (pnum)c, which is determined from the numerical modeling. By substituting (pnum)c into Eqs. (1) and (2), the overload factor, (OFnum)c, at collapse and the load factor (LF)num, respectively, can be determined. The relationship between the overload factor and the maximum surface settlement obtained from the centrifuge modeling and that computed with the numerical modeling for the singletunnel and parallel-tunnel models are in good agreement before tunnel collapse, as shown in Fig. 4. Numerical modeling can be used to examine deformation more precisely at small strain levels and at more locations than can be achieved with centrifuge modeling, but is not as precise as the measurements of the failure and post-failure behavior that the centrifuge modeling

can provide. Therefore, integrating and comparing the results derived from the numerical and centrifuge models provides improved understanding of the deformation behavior and the arching mechanism during tunneling.

3. Comparison of the results from centrifuge and numerical modeling 3.1. Tunnel stability The tunneling event was simulated by reducing the supporting pressure inside the tunnels in both the centrifuge modeling and the numerical experiments. Fig. 5 shows the relations between the C/D ratio and (OF)c measured in the single-tunnel and the parallel-tunnel model tests (solid symbols), and the values of (OFnum)c calculated from the numerical modeling (hollow symbols). The relationships between the C/D ratio and the lower bound of the overload factor, (OF)L, for singletunnel (Lee et al., 1999), and the upper bound of overload factor, (OF)L for single-tunnel and parallel tunnels (Wu and Lee, 2003), are also displayed in Fig. 5. The test results from Mair (1979) are also included in this gure. The increase in (OF)c with the C/D ratio illustrates that the stability of the tunnel improves if the tunnel is embedded more deeply. In addition, Fig. 5 also shows that the stability of parallel-tunnel is worse than that of single tunnel; lower overload factors at collapse were found in the parallel-tunnel model. The overload factors at collapse (solid symbols) obtained from the centrifuge model for both the single-tunnel and parallel-tunnel models have nearly the same values as those derived from the corresponding numerical model (hollow symbols), as shown in Fig. 5. The values of (OF)c and (OFnum)c for the single-tunnel model are all conned by upper and lower bounds. The values of (OF)c and (OFnum)c for the parallel-tunnel model are

Overload factor at collapse, (OF)c

6 5 4 3 2 1 0 0 1 2 3 4
Lower bound (single - tunnel) (Lee et al. , 1999) Upper bound (parallel- tunnel) (Wu and Lee, 2003) Single - tunnel Parallel -tunnel centrifuge model of single - tunnel (Mair, 1979) Numerical solution (single - tunnel) Numerical solution (parallel- tunnel) Upper bound (single - tunnel) (Wu and Lee, 2003)

Overload factor, (OF)

0 1 2 3 4 5 6 7 8 9 10 11 12

Settlement in prototype, (m)

0.0 0.5 1.0 1.5 2.0 2.5

Test7 (C/D=3) NTest7 (C/D=3) Twin3 (C/D=3, d/D=1.5) NTwin3 (C/D=3, d/D=1.5)

Cover-to-diameter ratio, C/D

Fig. 5. Comparison of the (OF)c values obtained from numerical and centrifuge modeling.

Fig. 4. Relations between the overload factor and maximum surface settlement derived from the numerical and centrifuge modeling.

124

C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119132

also conned by the upper bound, but no lower bound solution has yet been derived. The centrifuge and numerical modeling provide consistent evaluations of the tunnel stability. 3.2. Surface settlement troughs Figs. 6 and 7 compare the computed (represented by lines) and measured surface settlement troughs (represented by symbols) at selected load factors for the single-tunnel model (C/D = 2) and for the parallel-tunnel model (C/D = 1, d/D = 1.5), respectively. The distances, X, oset from the tunnel center-line (or from the center-line of the two tunnels in the parallel-tunnel tests) and the surface settlements, S, are both normalized with respect to the tunnel diameter, D. The computed settlement troughs compare reasonably well both in shape and magnitude with those measured in the centrifuge models at the correspond-

ing load factors. The shape of the settlement trough for the single tunnel approximates closely to that of the error function. An empirical approach derived from centrifuge modeling has been proposed for calculating the surface and subsurface settlement troughs for various ground losses due to tunneling in soft clay and in sandy soils (Wu and Lee, 2003; Lee et al., 2004). In addition, these researchers proposed a superimposition method for estimating the surface settlement troughs caused by parallel tunneling from the parameters obtained for single tunneling. Their methodology for predicting the settlement troughs based on the ground loss was veried by comparison with 12 sets of monitored eld data (Wu and Lee, 2003). 3.3. Comparison of the excess pore water pressures obtained from centrifuge and numerical modeling 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 approach suggested by Henkel, the changes in the pore water pressure, Du, can be determined with the equation 1 2 Du Dr1 Dr2 Dr3 aDr1 Dr2 3 Dr2 Dr3 2 Dr3 Dr1 2 1=2 ;

Normalized distance from centre-line, (X/D)

-8 0.00 -6 -4 -2 0 2 4 6 8

Settlement ratio, (S/D)

0.02 0.04 0.06 0.08 0.10

NTest9 (C/D=2)
LF=0.567 LF=0.646 LF=0.730 LF=0.814 LF=0.893 LF=1.0

Test9 (C/D=2)
LF=0.570 LF=0.651 LF=0.733 LF=0.814 LF=0.896 LF=1.0

where a is Henkels pore water pressure parameter, and Dr1, Dr2, Dr3 are the changes in the major, intermediate, and minor principle stresses, respectively. A comparison of a and Skemptons parameter, A, derived from triaxial compression tests gives   1 1 a p A . 4 3 2 This denition is useful because it enables the prediction of the Du associated with loading conditions under plane strain conditions if we assume that the soil bed is an isotropic and elastic-plastic material. By using a value of a of 0.3 (determined from the triaxial compression tests) and Dr2 = 1/2(Dr1 + Dr3) for the plane strain and undrained conditions, the developed excess pore water pressure can be estimated after the changes in total stress are calculated in the numerical simulations. In the centrifuge modeling, the stress changes on a soil element during tunneling cannot be measured, whereas the changes of pore water pressure can be directly measured. The numerical analysis in terms of total stress approach can give the changes of total stress but it cannot give the changes of pore water pressure. However, the changes of pore water pressure can be estimated using Eq. (3) after the total stress changes were obtained from the numerical analysis. In this study, the excess pore water pressure ratio, (Du/rvo), dened

Fig. 6. Comparison of surface settlement troughs at various load factors obtained from numerical and centrifuge modeling for a single tunnel.

Normalized distance from centre-line of parallel tunnels, (X/D)

-8 0.00

-6

-4

-2

Settlement ratio, (S/D)

0.02 0.04 0.06 0.08 0.10

NTwin4(C/D=1, d/D=1.5) LF=0.369 LF=0.493 LF=0.616 LF=0.739 LF=0.862 LF=0.985

Twin4(C/D=1, d/D=1.5) LF=0.374 LF=0.503 LF=0.615 LF=0.738 LF=0.880 LF=1.0

Fig. 7. Comparison of surface settlement troughs at various load factors obtained from numerical and centrifuge modeling for two parallel tunnels.

C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119132

125

as the excess pore water pressure normalized with the total overburden pressure of the measured point, is used to follow the variations of the excess pore water pressure during the tunneling simulations. The method for comparison of the measured and calculated (Du/rvo) is described as follows. The eects of tunneling on the total stress within the soil mass around a tunnel can be estimated from the computed responses at the grids of points shown in Fig. 8(a) (Test7, C/D = 3), Fig. 8(b) (Test9, C/D = 2), and Fig. 8(c) (Test8, C/D = 1) for the single-tunnel tests and from those shown in Figs. 9(a) (Twin3; C/D = 3, d/D = 1.5) and 9(b) (Twin2; C/D = 2, d/D = 1.5) for the parallel-tunnel tests. In the centrifuge model, the

Unit: m
Line A Line B Line C Line D Line E

A5

x 1.78 5.89

18

A4

B5 Negative arching zone B4

C5

D5

C4

D4 Positive arching zone D3 D2 D1

E4

6.04
E3

pore water pressures were measured at the points represented by solid circle symbols in Figs. 8 and 9 (i.e., A0, A3, B1, C1, and C3 in Fig. 8(a); A0, B2, C1, D1, and D2 in Fig. 8(a)). Figs. 10 and 11 compare the variations in the measured (represented by lines) and computed (represented by hollow circle symbols) Du/rvo with the overload factor at the corresponding locations for Test7 (C/D = 3) and for Twin3 (C/D = 3. d/D = 1.5), respectively. There is very reasonable agreement between the measured and computed excess pore water pressures before tunnel collapse. The small discrepancy between them may result from pore water pressure 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 found in the tests with dierent C/D and d/D ratios. Thus the stress states computed from the numerical experiments can be and are used to investigate the arching behavior of the soil mass around the tunnel in the next section, although the stress measurements in the centrifuge model tests were impossible.

A3

B3

C3 C2 C1

A2 Plastic zone B2

7.29
E2 E1

C.L.

Spring line

B1

7.29

Unit: m
A4

Line A

Line B

Line C

Line D

Line E

Line F

x 1.78

A0

z
(a)

B0

C0

D0

B4

C4

D4

5.89 6.04
F2

7.5
Unit: m
Line A

3
Line B

5.5
Line C Line D

12
Line E

A3

x 3.44
A2
C.L.

B3 Negative Arching zone B2

C3

D3 E3 Positive Arching zone D2 E2

C2

A3

12
A2

Negative B3 arching zone B2 Plastic zone

7.29
Plastic zone C1 D1 E1 F1

C3

D3 E2

4.23
A1 Spring line

C2 D2 Positive arching zone C1 D1

7.42

7.33
A0

6
(b)

C.L. Spring line

B1

E1

(a)

4.5
Line A

7.5

6
Line C Line D

7
Line E

8
x 1.24

z
Unit: m

7.5

4.5

4.5

6.5

Unit: m
Line B

Line A

Line B

Line C

Line D

A2

x 1.71 7.29

A3

B3
Negative arching zone

C3

D3

Negative B2 arching zone

C2
Positive arching zone

D2

12
A2

6.43
C2 D2
Positive arching zone

B2

E2

6 z
(c)

Plastic zone Spring line C.L.

7.33
E1

B1

C1

D1

A1

Spring line

C1

D1

C.L.

7.5

7.5

5.5
(b)

A0

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 arching zone for Test9 (C/D = 2). (c) Locations of reference soil elements and the arching zone for Test8 (C/D = 1).

4.5

7.5

Fig. 9. (a) Locations of reference soil elements and the arching zone for Twin3 (C/D = 3, d/D = 1.5). (b) Locations of reference soil elements and the arching zone for Twin2 (C/D = 2, d/D = 1.5).

126
0.4 0.3 (u / vo ) 0.2 0.1 0.0 -0.1 -0.2 0 2

C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119132
0.4 Test7 (C/D = 3) B1 (u / vo ) Measured Calculated 0.3 0.2 0.1 0.0 -0.1 C1 Measured Calculated

(OF)c
4 6 8 10 12 Overload factor, (OF )

-0.2 0 2 4

(OF)c
6 8 10 12 Overload factor, (OF )

(a)
0.4 0.3 0.2 (u / vo ) 0.1 0.0 -0.1 -0.2 0 2

(b)
0.4

C3 Measured Calculated (u / vo )

0.3 0.2 0.1 0.0 -0.1 -0.2

A3 Measured Calculated

(OF)c
4 6 8 10 12 Overload factor, (OF ) 0.4

(OF)c
0 2 4 6 8 10 12 Overload factor, (OF)

(c)

(d)

A0 0.3 0.2 (u / vo ) 0.1 0.0 -0.1 -0.2 0 2 4 6 8 10 12 Measured Calculated

(OF)c
Overload factor, (OF)

(e)

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

4. Mechanism of arching around tunnels 4.1. Evolution of arching eect The eects of tunneling on the total stress within the soil mass around a tunnel can be estimated from the computed responses at the grids of points shown in Fig. 8(a) (Test7, C/D = 3), Fig. 8(b) (Test9, C/D = 2), and Fig. 8(c) (Test8, C/D = 1) for the single-tunnel tests and from those shown in Figs. 9(a) (Twin3; C/D = 3, d/D = 1.5) and 9(b) (Twin2; C/D = 2, d/D = 1.5) for the parallel-tunnel tests. A more detailed understanding of the stress transfer in a tunneling problem from moving parts of the soil (settle more) to adjacent parts (settle less) can be achieved by considering the vertical stress redistributions in the soil mass above the spring line. The arching ratio is dened as

AR%

Drv 100; rvo

where Drv is the change in the vertical stress during tunneling and rvo is the total overburden pressure. In this study, we used the arching ratio to describe the arching behavior quantitatively at various locations. An element that receives higher load transfers from adjoining yielding or exible elements will generate a larger positive arching ratio. Conversely, a negative arching ratio will arise if an element shifts load to non-yielding parts or to more rigid elements. Fig. 12 shows a plot of the arching ratio versus overload factor for NTest7 (C/D = 3) at the grid of points shown in Fig. 8(a). As the overload factor increases, all of the elements on the vertical center-line (Line A) experience a decrease in vertical stress (negative AR) but the elements

C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119132
0.4 0.3 (u /vo ) 0.2 0.1 0.0 -0.1 -0.2 0 2 Twin3 (C/D=3, d/D=1.5) C1 Measured calculated (u / vo) 0.4 0.3 0.2 0.1 0.0 -0.1 D1 Measured calculated

127

(OF)c
4 6 8 10 12

-0.2 0 2

(OF)c
4 6 8 10 12

(a)
0.4 0.3 (u /vo ) 0.2 0.1 0.0 -0.1 -0.2 0 2

Overload factor, (OF)

(b)
0.4 D2 Measured calculated 0.3 (u /vo ) 0.2 0.1 0.0 -0.1

Overload factor, (OF)

B2 Measured calculated

(OF)c
4 6 8 Overload factor, (OF) 0.4 0.3 0.2 (u /vo ) 0.1 0.0 -0.1 -0.2 0 2 10 12

-0.2 0 2

(OF)c
4 6 8 10 12

(c)

(d)

Overload factor, (OF)

A0 Measured calculated

(OF)c
4 6 8 Overload factor, (OF) 10 12

(e)

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

on the spring line (B1E1) experience an increase in vertical stress (positive AR), as shown in Fig. 12(a). Similarly, as can be seen in Fig. 12(b) and (c), the magnitude of the arching ratio is also related to the distance oset from the tunnel center and to the overload factor (comparing the arching ratios on Lines A, B, C, D, and E). Fig. 13 summarizes the changes in the arching ratio for three overload factors (OF = 1, 3, and (OFnum)c) for the elements on Lines A to E for NTest7 (C/D = 3). The magnitude of the positive arching ratio increases with increases in the overload factor for the elements along Line B but declines rapidly once the element is yielding (Fig. 13(b)). The element at C1 experiences the largest positive arching ratio (about a 9% rise). The elements located at a greater distance from the tunnel center experience smaller positive arching ratios. In contrast, the arching ratio becomes more neg-

ative as the overload factor increases for the elements on Line A (Fig. 13(a)). Similarly, Fig. 14 summarizes the arching ratio as a function of the overload factor for the elements on the spring line (B1E1) and on the vertical center-line (A2A3) shown in Fig. 8(b) for NTest9 (C/D = 2). Fig. 15 summarizes the arching ratio as a function of the overload factor for the elements on the spring line (B1D1) and on the vertical center-line (A2) shown in Fig. 8(c) for NTest8 (C/D = 1). The evolution of the arching ratio in the shallower tunneling shown in these two gures is similar to those presented in Figs. 12 and 13 but smaller positive and negative arching ratios are obtained at the corresponding points for deeper tunneling (NTest7, C/D = 3). Shallower tunneling imposes a larger arching eect on the surrounding soil mass.

128
10

C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119132

NTest7 (C/D=3)
5

-5

-10

-15

A0 A3 A4 A5 B1 C1 D1 E1 0 1 2 3

(OF)c
4 5

-20

(a)
10

Overload factor, (OF)

B0 B1 B3 B4 B5 C0 C1 C3 C4 C5

NTest7 (C/D=3)

-2 0 1 2 3

(OF)c
4 5

(b)
10

Overload factor, (OF)

NTest7 (C/D=3)
8 D0 D1 D3 D4 D5 E1 E3 E4

(C/D = 3, d/D = 1.5). Fig. 17 shows the changes in the arching ratio for the elements on Lines AF (Fig. 9(b)) for NTwin2 (C/D = 2, d/D = 1.5). A comparison of the changes in the arching ratios shown in Figs. 13 and 16 indicates that the arching ratios in parallel tunneling evolve in nearly the same way as in single tunneling, except at Element A1, which is located on Line A and on the spring line. At this point, a larger positive arching ratio develops initially, but its value rapidly declines to near zero once the stress state becomes yielding, in the same manner as the other elements (Fig. 16(a)). This part of the soil mass takes the load that is transferred from the two compressive arches above the tunnels due to parallel tunneling, and behaves like a pillar. After examining in detail the changes in the arching ratio shown in Figs. 1315 for the single-tunnel tests and in Figs. 16 and 17 for the parallel-tunnel tests, we reached the same conclusion for parallel tunneling as obtained for single tunneling, namely that shallower tunneling imposes a larger arching eect on the surrounding soil mass because of the higher arching ratio. In the centrifuge modeling, the excess pore water pressure ratio will increase on an element in the soil bed that receives load transfers from adjoining yielding or exible elements. Conversely, the excess pore water pressure ratio will decrease if an element shifts load to non-yielding parts or to more rigid elements. Therefore, the changes in the measured excess pore water pressure ratio can also be used to track the load transfers among the elements during tunneling simulation in the centrifuge modeling. For example, the excess pore water pressure ratios shown in Figs. 14 and 15 initially rise with increases in the overload factor, but later fall as the elements in non-yielding states progress to yielding states during increases in the overload factor. 4.2. Boundaries of the arching and plastic zones in the tunnel collapse stage As discussed in the previous section, the outer boundaries of the arching and plastic zones expand outward from the excavated area as the overload factor increases. Knowledge of the boundaries of the arching and plastic zones in the tunnel collapse stage is thus crucial for engineering practice, and is now discussed. Fig. 18 presents the variations in the shear stress ratio, q/su, at (OFnum)c for the soil elements on the selected lines at the elevations of 3.5, 7.29, 11, and 16 m above the spring line (NTest7, C/D = 3). Here q = 1/2(r1 r3). The arrows shown in Fig. 18 indicate the positions of the outermost boundary of the plastic zone at the elevations where the elements have stress states of su/q = 1 and are regarded as yielding. As can be seen in the gure, the plastic boundaries extend from the tunnel center-line as far as 12 m along the spring line (about twice the tunnel diameter) and but only to 10 m at an

Arching ratio, (%)

Arching ratio, (%)

Arching ratio, (%)

-2 0 1 2 3

(OF)c
4 5

(c)

Overload factor, (OF )

Fig. 12. Arching ratio versus overload factor at various locations (NTest7 C/D = 3).

Fig. 16 summarizes the changes in the arching ratio for the elements on Lines AF (Fig. 9(a)) for three overload factors (OF = 1.5, 2.5, and (OFnum)c) for NTwin3

C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119132
25 Distance from spring line, (m) 20 15 10 5 0 -5 -10 -20 -15 Spring line 25 25 NTest7 (C/D=3) Line B OF=1 OF=3 (OFnum)c 20 15 10 5 0 -5 -10 -20 NTest7 (C/D=7) Line C OF=1 OF=3 (OFnum)c

129

OF=1 NTest7 (C/D=3) Line A OF=3 (OFnum)c

20 15 10 5 0 -5 -10 -20

(a)

-10 -5 0 Arching ratio, (%) 25 Distance from spring line, (m) 20 15 10 5 0 -5 -10 -20 -15

10

-15

(b)

-10 -5 0 Arching ratio, (%) 25

10

-15

(c)

-10 -5 0 Arching ratio, (%)

10

OF=1 NTest7 (C/D=3) Line D OF=3 (OFnum)c

20 15 10 5

NTest7 (C/D=3) Line E OF=1 OF=3 (OFnum)c

Spring line

0 -5 -10

-10

-5

10

-20

-15

(d)

Arching ratio, (%)

(e)

-10 -5 0 Arching ratio, (%)

10

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 15 10

35

NTest9 (C/D=2)

30 25

Arching ratio, (%)

Arching ratio,(%)

5 0 -5 -10 -15 -20 -25 -30 0.0 B1 C1 D1 E1 A2 A3

20 15 10 5 0 -5 B1 C1 D1 A2

(OF)c
0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5

-10 (OF)c -15 0 1 2 3 4

Overload factor, (OF)

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

Overload factor, (OF)

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

elevation 3.5 m above the spring line. Therefore, the outer boundary of the plastic zone in the tunnel collapse stage can be inferred by examining all the elements regarded as yielding (not just those considered in Fig. 18), and is depicted in Fig. 8(a) as a thick line of

dashes. Fig. 8(b) and (c) also display the outer boundaries of the plastic zone for the single-tunnel models (C/D = 1, 2). Similarly, Fig. 19 presents the variations of the shear stress ratio, q/su, at (OFnum)c for the soil elements on the

130
20 Distance from spring line, (m) 15 10 5 spring line 0 -5 -10 -30 -20 NTwin3 Line A (C/D=3, d/D=1.5)

C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119132
20 20 NTwin3 Line B (C/D=3, d/D=1.5)

OF=1.5 OF=2.5 (OFnum)c

15 10 5

OF=1.5 OF=2.5 (OFnum)c

15 10 5

NTwin3 Line C (C/D=3, d/D=1.5)

spring line 0 -5 -10 -30 0 -5 -10 -30

spring line

OF=1.5 OF=2.5 (OFnum)c

-20 -10 0 10 20

(a)
20 Distance from spring line, (m) 15 10 5 0 -5 -10 -30

-10 0 Arching ratio, (%)

10

20

-20

-10

10

20

(b)
20

Arching ratio, (%)

(c)
20

Arching ratio, (%)

NTwin3 Line D (C/D=3, d/D=1.5)

15 10 5

NTwin3 Line E (C/D=3, d/D=1.5)

15 10 5

NTwin3 Line F (C/D=3, d/D=1.5)

spring line

spring line 0 -5 -10 -30 0

spring line

OF=1.5 OF=2.5 (OFnum)c

-20 -10 0 10 20

OF=1.5 OF=2.5 (OFnum)c

-20 -10 0 10 20

-5 -10 -30

OF=1.5 OF=2.5 (OFnum)c

-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 above the spring line (NTwin3, C/D = 3, d/D = 1.5). As can be seen in the gure, the boundaries of the plastic zone extend from the center-line of the two tunnels as far as 15 m along the spring line and 8 m at an elevation of 3.5 m above the spring line. The outer boundary can also be determined with the procedure described in the previous paragraph, as shown in Figs. 9(a) (Twin3) and 9(b) (Twin2) (thick lines of dashes). The non-yielding elements in the arching zone receive a load transfer from the elements in the plastic zone. The area lying between the outer boundary of the plastic zone and the boundary at which the elements have an arching ratio larger than 1% is regarded as the arching zone. Thus the outer boundaries of the single-tunneling arching zones can easily be determined, and are shown as thick lines in Fig. 8(a)(c) for various C/D ratios. The arching mechanism in the case of two paralleltunnels is similar to that for a single-tunnel, so the same procedure can be used to infer the boundaries of the arching zone. They are depicted in Fig. 9(a) and (b) for Twin3 and Twin2. Tunneling at dierent burial depths can result in arching and plastic zones of dierent extents. Deep single tunneling will result in a wider arching zone, but two

parallel tunnels only generate a slightly wider arching zone for the same burial depth. The extents of the arching zones for a single-tunnel and parallel-tunnels embedded at various depths (C/D = 1, 2, 3) are shown in Fig. 20(a) and (b), respectively. The curves representing the outer bounds of the arching zones were obtained as follows:     C z 0.1 exp 0.305x 0.045 x D for single tunnel;     C z 0.105 exp 0.27x 0.0375 x D for two parallel tunnels d =D 1.5; 6a

6b

in which x is the distance oset from the tunnel center for the single-tunnel and the distance oset from the center-line of the two parallel-tunnels, and z is the depth. 4.3. Positive and negative arching zones After examining the variations in the arching ratios of the elements surrounding the tunnel (or tunnels), as shown in Figs. 13, 16, and 17, it was found that each arching zone can be divided into two zones, the positive

C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119132
15 Distance from spring line, (m) NTwin2 Line A (C/D=2, d/D=1.5) 10 15 NTwin2 Line B (C/D=2, d/D=1.5) 15 NTwin2 Line C (C/D=2, d/D=1.5) 10

131

5 spring line 0

OF=0.97 OF=1.95 (OFnum)c

10

5 spring line 0 (OF)=0.97 OF=1.95 (OFnum)c

5 spring line 0

-5

-5

-5

OF=0.97 OF=1.95 (OFnum)c

-20 -10 0 10 20 30

-10 -30

-20

-10

10

20

30

-10 -30

-20

-10

10

20

30

-10 -30

(a)

Arching ratio, (%) 15 Distance from spring line, (m) NTwin2 Line D (C/D=2, d/D=1.5)

(b)

Arching ratio, (%) 15

(c)

Arching ratio, (%)

10

10

NTwin2 Line E (C/D=2, d/D=1.5)

5 spring line 0

5 spring line 0

-5

OF=0.97 OF=1.95 (OFnum)c

-20 -10 0 10 20 30

-5

OF=0.97 OF=1.95 (OFnum)c

-20 -10 0 10 20 30

-10 -30

-10 -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).

Ratio of shear stress, (q/s u )

1.2 1.0 0.8 0.6 0.4 0.2 0.0 0

Ratio of Shear stress,q/s u

NTest7 (C/D=3) spring line 3.5m above spring line 7.29m above spring line 11m above spring line 16m above spring line

1.2 NTwin3 (C/D=3, d/D=1.5) 1.0 0.8 0.6 0.4 0.2 0.0 0 10 20 30 40 Distance from center line of two parallel tunnels, (m)
Fig. 19. Variations in the shear stress ratio at (OFnum)c for soil elements along lines at various elevations above the spring line for two parallel tunnels (NTwin3 C/D = 3 d/D = 1.5).

spring line 3.5m above spring line 7.29m above spring line 11m above spring line 16m above spring line

10 20 30 Distance from centerline of tunnel, (m)

40

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 single tunnel (NTest7 C/D = 3).

arching zone (with positive AR) and the negative arching zone (with negative AR). The boundaries (dash dot lines) of the two zones derived from the numerical analysis are shown in Fig. 8(a)(c) for the single-tunnel with various C/D ratios and in Fig. 9(a) and (b) for two parallel-tunnels with C/D = 2, 3 and d/D = 1.5. The shaded rectangular regions in these gures correspond to the half-width of the sliding wedge, which

can be determined from the upper bound solution using the collapse mechanism proposed by Wu and Lee (2003). The sliding wedges are reasonably consistent with the boundaries of the negative arching zones. A pile embedded in the negative arching zone would partially lose both end bearing capacity and skin friction on the pile body due to the reduction of the vertical stresses

132

C.J. Lee et al. / Tunnelling and Underground Space Technology 21 (2006) 119132
Distance from tunnel center, (m) 0 0 5 10 20 30 0 5 Distance from centerline of two parallel tunnels, (m) 0 10 20 30

Depth, (m)

10 15 20 25

Depth, (m)

Single-tunnel

Paralle-tunnel 10 15 20 25

C/D=1 C/D=2 C/D=3 Fitting curve z=0.1exp{0.305x-0.045(C/D)x]

C/D=1, d/D=1.5 C/D=2, d/D=1.5 C/D=3, d/D=1.5 Fitting curve z=0.105exp[0.27x-0.0375(C/D)x]

(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 during new nearby tunneling. 5. Summaries and conclusions A series of centrifuge model tests were carried out to investigate the surface settlement trough, excess pore water pressure generation, and tunnel stability of the tunnels with various C/D ratios (single-tunnel) and d/ D ratios (two parallel-tunnels). Numerical analysis was also conducted to evaluate the tunnel stability and arching eects that develop during tunneling in soft clayey soil. The centrifuge and numerical modeling produced consistent results in their predictions of the surface settlement trough, excess pore water generation, and the overload factors at collapse for both single tunneling and parallel tunneling. An arching ratio derived from the numerical analysis was dened to describe the evolution of the arching eect in the soil mass surrounding the tunnels. 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. Construction engineers can easily locate the boundaries of the positive and negative arching zones and then take appropriate measures to mitigate possible damages to underground structures due to new nearby tunneling. Acknowledgments The nancial support provided by the National Science Council, Taiwan, under Grants NSC 87-2211-E-008-024 and NSC 91-2211-E-008-026 is gratefully acknowledged.

References
Clough, G.W., Schmidt, B., 1981. Design and performance of excavation and tunnels in soft clay. In: Soft Clay Engineering. Elsevier, Amsterdam, pp. 600634 (Chapter 8).

Chen, C.Y., Martin, G.R., 2002. Soilstructure interaction for landslide stabilizing pile. Computer and Geotechnics 29 (5), 363386. Cundall, P.A., Coetzee, M.J., Hart, R.D., Varona, P.M., 1993. FLAC Users Manual. Itasca Consulting Group, USA. Handy, R.L., 1983. The arch in soil arching. Journal of Geotechnical Engineering, ASCE 111 (3), 302318. Koutsabeloulis, N.C., Griths, D.V., 1989. Numerical modeling of the trap door problem. Geotechnique 39 (1), 7789. Ladanyi, B., Hoyaux, B., 1969. A study of the trap-door problem in a granular mass. Canadian Geotechnical Journal 6 (1), 111. Lee, C.J., Wu, B.R., Chiou, S.Y., 1999. Soil movements around a tunnel in soft soils. Proceedings of the National Science Council, Part A: Physical Science and Engineering 23 (2), 235247. Lee, C.J., Chiang, K.H., 2004. Load transfer on single pile near new tunneling in sandy ground. In: Matsui, Tanaka, Mimura (Eds), Proceedings of the International Symposium on Engineering Practice and Performance of Soft Deposits (IS-OSAKA 2004), pp. 495506. Lee, C.J., Chiang, K.H., Kou, C.M., 2004. Ground movement and tunnel stability when tunneling in sandy ground. Journal of the Chinese Institute of Engineers 27 (7), 10211032. Lee, K.M., Rowe, R.K., 1991. An analysis of three-dimensional ground movements: the thunder bay tunnel. Canadian Geotechnical Journal 28 (1), 2541. Mair, R.J., 1979. Centrifugal modeling of tunnel construction in soft clay. Ph.D Thesis, University of Cambridge, UK. Mair, R.J., Gunn, M.J., OReilly, M.P., 1981. Ground movements around shallow tunnels in soft clay. In: Proceedings of the 10th International Conference on Soil Mechanics and Foundation Engineering, pp. 323328. Nakai, T., Xu, L., Yamazaki, H., 1997. 3D and 2D model tests and numerical analyses of settlements and earth pressures due to tunnel excavation. Soils and Foundations 37 (3), 3141. 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, 181193. Peck, R.B., 1969. Deep excavation and tunneling in soft ground. In: Proceedings of 7th International Conference on Soil Mechanics and Foundation Engineering, Mexico, State of the Art Volume, pp. 225290. Stone, K.J.L., Newson, T.A., 2002. Arching eects in soilstructure interaction. In: Phillips, Guo, Popescu (Eds.), Physical Modeling in Geotechnics: ICPMG 02, pp. 935939. Terzaghi, K., 1943. Theoretical Soil Mechanics. Wiley, New York. Wang, W.L., Yen, B.C., 1973. Soil arching in slopes. Journal of the Geotechnical Engineering, ASCE 100 (1), 6178. Wu, B.R., Lee, C.J., 2003. Ground movements and collapse mechanisms induced by tunneling in clayey soil. International Journal of Physical Modelling in Geotechnics 3 (4), 1327.