Lea, E. & Domieux, L. (1990). Giotechnique 40, No. 4, 58146 Upper and lower bound solutions for the face stability of shallow circular tunnels in frictional material E. LECA* and L. DORMIEUX* With the recent increase in underground urban development as well as for transportation, tunnels need to he driven in increasingly dif&ult soil conditions. In most cases the ground itself is not stable and face stability is achieved by applying fluid pressure to the tunnel front. The question of determining the retaining pressure to he used has received special consideration in the past hecause of the concern for safety during construction, and also hecause of the damage that could he caused to surface structures by the failure of a shallow tunnel. The problem is three-dimensional and can he studied by using the limit state design method. Solutions are available for the case of a circular tunnel in purely cohesive ground, but very little is known of the face stability of a tunnel driven in sandy soils. This Paper approaches this latter problem from the point of view of limit analysis. Both safety against face collapse and blow-out are considered. Three upper hound solutions are derived from the consideration of mechanisms based on the motion of rigid conical blocks. The results are compared with lower hound solutions published in a previous article. A failure criterion is proposed for the tunnel face in the general case of a cohesive and frictional soil, and charts are provided to allow a bracketed estimate of the required retaining pressure. A comparison with centrifuge laboratory tests shows close agreement between the theoretical upper hound solutions and the face pressures at collapse measured experimentally. KEYWORDS: limit state design/analysis; city; research; sands; stability; tunnels. plasti- INTRODUCTION The stability of the front of a tunnel driven in cohesive material has been studied by several authors since the paper by Broms & Bennermark (1967) on a stability criterion based on laboratory extrusion tests and field observations. Similar Discussion on this Paper closes 5 April 1991; for further details see p. ii. * Ecole Nationale des Ponts et Chaussees, 93167, Noisy-laGrand Cedex, Paris, France. L’utilisation croissante du sous-sol en site urbain ainsi que le dtveloppement des riseaux de transports enter& conduit i construire des tunnels dans des conditions toujours plus difficiles. Dans la plupart des cas le terrain est instable et ii est n& cessaire d’appliquer une pression de souttinement au niveau du front de taille. Le cboix de la pression I utiliser a fait l’objet de qombreuses 6tudes au tours des derniLres an&es en raison des problemes po&s pour de tels projets du point de vue de la &curiti en tours de construction et des cons& quences d’une rupture sur les structures situ&s en surface. 11 s’agit d’un problbme tridimensionnel qui peut 6tre notamment btudii! par la mbtbode de I’analyse limite. Des solutions ont dbjji &ti! propos&s pour le cas d’un tunnel circulaire en terrain cob&rent, mais les connaissances sont encore limit&es quant ai la stabilite du front de taille I’un tunnel creu& dans des sols sableux. Ce dernier problkme est examin dans le pr&sent article du point de vue de I’analyse limite. On s’intkresse h la fois aux risques de rupture par effondrement et par explosion. L’examen de trois m6canismes ha.6 sur le mouvement rigide de blocs coniques permet d’ahoutir 1 une approche par l’extbrieur des conditions de stabilitb. Les rbultats sont comparb i I’approcbe par I’int&ieur p&sent&e dans un article p&&dent. L’6tude ahoutit P la mise en C?vidence d’un critere de rupture pour le front de taille dans le cas g&n&al d’un sol frottant-cob&ent. On propose igalement une drie d’abaques permettant d’obtenir un encadrement de la pression de soutb nement P utiliser. L’application I des essais en centrifuge montre que les homes sup&ieures obtenues analytiquement sont t&s proches des P la pressions mesuri?es exp&imentalement rupture. conclusions were presented by Peck (1969) and more recently a more theoretical approach was proposed by Davis et al. (1980), which is derived from the limit state design concept. Their results are supported by centrifuge model tests (Schofield. 1980) and nrovide a more general expression of the stability criterion, -which accounts for the effects of tunnel depth. Yet very little is known about the face stability of tunnels when the ground is characterized by a Mohr-Coulomb yield criterion. This question 581 Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. LECA AND DORMIEUX 582 has, however, received more attention recently because of the construction of tunnels in increasingly difficult ground conditions and even in sandy materials. Some lower bound solutions were proposed by Muelhaus (1984); they essentially refer to the stability of the tunnel lining, but an estimate of the critical unsupported length behind the face is also provided. A more general solution for the tunnel front is given by Leca & Panet (1988). This Paper addresses the problem of the face stability of a tunnel driven in a Mohr-Coulomb material from the point of view of limit analysis. The major difficulty of such a problem comes from the fact that it is fully three-dimensional. An upper bound solution is derived from the analyses of three failure mechanisms. The results obtained for the lower bound (Leca & Panet, 1988) are re-examined and compared with the upper bound solutions. This allows a bracketed estimate to be made of stability conditions for the tunnel front. The critical pressures obtained from this method are compared with experimental results from centrifuge tests. The construction procedure for shallow tunnels in soils is described in the Paper by Davis et al. (1980). The problem can be idealized, as shown in Fig. 1, by considering a circular rigid tunnel of diameter D driven under a depth of cover C, i.e. at a depth FM+; The unit weight of the soil is y and a surcharge us is applied at ground surface. The unsupported length behind the face P is taken as zero (which is a reasonable assumption for shield driven tunnels) and a retaining pressure rrr is applied to the front. Such support can be achieved by using com- pressed air, bentonite slurry or earth pressure (EPB shield). In the case of this study err is assumed to be constant, which best represents the case of compressed air, but also gives some information for slurry or EPB shields. The ground conditions are also assumed to be uniform around the tunnel. The soil is modelled as a Mohr-Coulomb material, characterized by its cohesion c’ and its friction angle 4’. For such materials it is helpful to introduce the unconfined compression strength Uc = 2 c’ cos #J’ 1 - sin 4’ and the Rankine K, = earth pressure coefficients 1 - sin 4’ 1 + sin f#~’ for active failure and K = 1 + sin 4 P 1 - sin 4 for passive failure. As an alternative set of parameters uc and K, (or KP) can be used to characterize the soil’s resistance. For a cohesionless material K, or K, can be used alternatively with 4’. Dimensional analysis shows that this problem can be analysed in terms of five dimensionless parameters : C/D, a&,, 4~, Y%, and K, (or KA). LIMIT ANALYSIS The purpose of limit analysis is to provide an estimate of stability conditions for a mechanical system regardless of the behaviour of the material it is made of (Salencon, 1983, 1990). Stability con- Sectm A-A Fig. 1. Simplified geometry for tbe front stability of a shallow tunnel Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. FACE STABILITY OF TUNNELS IN FRICTIONAL SOILS 583 Fig. 2. Bracketed estimate of the domain of supported loads ditions for this system are derived in terms of the loads that can be applied to the system without causing its failure. An upper bound estimate of such loads is found by considering a kinematically admissible failure mechanism for which the power 8, of the loads applied to the system is larger than the power P, that can be dissipated inside the system during its movement (the upper bound theorem). On the other hand, any set of loads for which a stress field can be found that satisfies equilibrium and the material yield criterion, is a lower bound solution (lower bound theorem). Stability can be characterized by the domain .X of supported loads; .X is bracketed by a lower bound domain Xx- and an upper bound domain X’. This is shown in Fig. 2 for the case of a system subjected to a set of two loads Qi and Q,.Any set of loads located inside X- is potentially supported by the system while any set of loads outside X’ is unstable. The domain X is convex, which means that if the two sets of loads Q’ and Q” are supportable, any linear combination of these loads @’ + (1 - A@” where 0 < 1 < 1 is also supportable. In Fig. 2 any point M located on the straight line between A and A’ is inside &‘- as long as A and A’ both belong to X‘-. This convexity can be used to obtain an improved lower bound solution. For the tunnel problem analysed in this Paper three loading parameters need to be considered: %lo, t %/UC7 and yDJa,.Since soil yield is controlled by the Mohr-Coulomb criterion, the failure mechanisms must be chosen such that along any failure surface X,, the angle ed between the discontinuity velocity V,, and & (Fig. 3) satisfies the condition &<8*<rC-4 (5) Otherwise, the dissipated power P, is equal to infinity and no upper bound solution can be derived from the mechanisms analysed. Fig. 3. Discontinuity velocity along the failure surface UPPER BOUND SOLUTIONS Three failure mechanisms have been considered. They all involve the movement of solid conical blocks with circular cross-sections. The opening of each cone is equal to 24’ and its velocity Y is parallel to its axis (Fig. 4). Therefore condition (5) is satisfied along the failure surfaces between the moving blocks and the rest of the ground. The three mechanisms MI, MI1 and MI11 are shown in Fig. 5. MI and MI1 are collapse mechanisms, whereas MI11 refers to blowout failure. Even though safety against collapse is a major concern during tunnelling, case MI11 may be of interest for very shallow tunnels bored in weak soils, when the pressure (TVcan become so great that soil is heaved in front of the shield. \ --_ \ \\ //fi\ TR \\F--- \ \\ V Fig. 4. Conical blocks and kinematic conditions used in MI, MI1 and MID Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. LECA AND DORMIEUX / A’ (4 / I (?I’) / Fig. 5. Mechanisms (a) MI, (b) MI1 and (c) MI11 Such a phenomenon has been observed during tunnelling projects (Clough et al., 1983). Failure is due to the collapse of one conical block in MI and two blocks in MII. The geometry is a little more complex in the latter case: the first cone (block 1) is truncated by a plane K perpendicular to the plane of symmetry of the tunnel, which projects as A on Fig. 5(b). The second cone (block 2) is a mirror image of the first with respect to plane rr’ (perpendicular to rr and going through the centre of the intersection Xi2 between n and block 1; plane ri projects as A on Fig. 5(b)). This ensures that plane x intersects both blocks along the same elliptical surface Zr2. Plane x is chosen such that the axis of the second cone is vertical. Therefore, both MI and MI1 are characterized by only one parameter, the angle a between the axis of the cone adjacent to the tunnel and the horizontal. MI11 is also characterized by a (Fig. 5(c)); the geometry is similar to that of MI except that the cone is inverted and the velocity reversed. It should be noted that for Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. FACE STABILITY OF TUNNELS Fig. 6. Area of failure at the tunnel face all three mechanisms the intersection of the tunnel with its adjacent cone is an ellipse &,i the major semi-axis length of which is equal to D/2 (Fig. 6). This implies that only part of the tunnel face is failing. However, limit analysis theory remains valid for such a geometry and upper bound solutions can still be derived from these three mechanisms. The derivations related to MI, MI1 and MI11 are given in Appendices I, II and III respectively. In all three cases the power 8, of external loads Wffc > a&,, yD/u,}and the dissipation power P, are first calculated separately. An upper bound solution is then found, given that in order for the set of loads {as/u,, a&,-, yD/u,}to be stable, 8, and Pv must satisfy PC < Pv (6) To interpret the results obtained in Appendices I, II and III it is convenient to rewrite the three loading parameters Qs = (K, - 1) z + 1 QT = (KP - 1) 2 +1 Q, = WP- 1) $ (7a) (7c) Then, relation (6) leads to upper bound that can be put in the form (8) NsQs+N,Q,GQ, for collapse mechanisms solutions MI and MI1 and NsQs+N,Q,aQ, for blow-out mechanism MIII. and (9) Ns and N, are weighting depend on the angle a between cone adjacent to the tunnel and (9) In relations (8) coefficients that the axis of the the horizontal. IN FRICTIONAL SOILS 585 The two collapse mechanisms are optimized when u is chosen such that N, and N, are at maximum. MI11 is optimized when Ns and N, are at minimum. For all three mechanisms the coefficients N, and N, of the best upper bound results will be respectively referred to as Ns’+, N,‘+ for collapse and Nsb+, NYb+ for blow-out. The results of these optimizations are summarized on Fig. 7(a) for collapse and Fig. 7(b) for blow-out failure. In Fig. 7, optimal values Ns’+, N,” and Nsb+, N b+ are plotted as functions of the depth ratio CID for common values of the friction angle 4’ (20”45”). It was found that MIX provides the best upper bound for collapse in most cases except for very shallow tunnels (C/D< 0.25) or friction angle values smaller than 30”. MI and MI1 lead to similar results when C/D is greater than 1.0. Fig. 7(a) shows that N,‘+ is almost always smaller than N,‘+ and is equal to zero for any value of 4’ when C/D 2 0.6. This suggests that if the actual failure conditions are similar to those predicted by MI and MII, the surcharge us will have very little influence on face collapse, unless the tunnel is very shallow. For most conditions, failure would not reach the ground surface and MI as well as MI1 could be considered as local collapse mechanisms. However, such mechanisms could lead to some larger scale failure with the formation of sink holes, since a large amount of soil would be left unsupported once initial collapse of the tunnel face had occurred. Another conclusion from the analysis of collapse mechanisms was that the optimal values N Et NC+ are always obtained for essentially the sa”,e’vaLe a’+ of parameter a a’+ N 49” - $ This means that in the plane of symmetry of the tunnel, the angle of the critical failure surface to the horizontal P+ = 49” + r&/2 is larger than the angle of active failure in plane strain conditions 6, = 45” + #/2 (Fig. 8(a)). Therefore the area in front of the tunnel which is influenced by the collapse is more limited than in the case of a long open cut. This could be seen as a stabilizing effect due to three-dimensional equilibrium conditions around the tunnel face and would again need to be compared to actual failures observed in the field. The values of NSb+ and NYb+ computed for the blow-out case, however, are large and increase sharply with the depth ratio CID,which is consistent with the fact that such a failure would only occur with very shallow tunnels. The critical geometry is obtained when ab+ N 49” Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. (11) 586 LECA AND e I N”,’ ’ DORMIEUX ___ Iv,+ - --- NC,+ 022 / \ g’ = 20 0.16- \ \ \ 0.16- \ \ \ 0.14- I$’ = 25 \ \ \ 0.12-\ \ \ O.lO- \ ‘\ 0.06 - lp’ = 30 \ \ \ \ \ \ \ \ $’ = 35’ \ fp’ = 40” f$’ = 45” _ I’ = 20” -L__ 0.2 0.4 0.6 (a1 Fig. 7. Upper bound values of Ns and N, (a) for collapse and (b) for blow-out Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. CID FACE STABILITY OF TUNNELS IN FRICTIONAL SOILS 587 (W Fig. 8. Critical geometries for (a) collapse and (b) blow-out This means that the angle of the failure surface to the horizontal Sb+ = 49” - 4’ is smaller than the angle of passive failure in plane strain conditions 6, = 45” - @/2 for common values of 4’ (Fig. 8(b)). LOWER BOUND SOLUTIONS Some lower bound solutions for the case of a Mohr-Coulomb material have been published previously (Leca & Panet, 1988). They are based on three stress fields similar to those used by Davis et al. (1980). These three stress fields SI, SII, 311 are shown on Fig. 9. SI is a geostatic stress field and it can actually be used in the general case of a soil with weight (y > 0). In the geometry shown on Fig. 9(a), the ground is divided in three layers: above and below the tunnel (y > D/2 or y < -D/2) the stress field is also isotropic; in the layer located at tunnel depth (-D/2 < y < D/2) the horizontal component of stress along Oz is equal to a,. SII and SIII both apply to a weightless soil (y = 0). Even though this assumption may not be realistic in the case of a shallow tunnel the lower bound derived from SII and SIII will be used to improve the general solution obtained from the consideration of stress field SI. SII is axisymmetric around the tunnel axis. Within the cylinder C, which extends the tunnel in the Oz direction, the axial stress is equal to (or and the radial and tangential stresses to a constant value go, which is chosen such that the ground is at yield everywhere in C,. Outside the cylinder C,, which is tangential to the ground surface, the stress field is isotropic and equal to 0s. Between cylinders C, and C, a solution for radial stress or and tangential stress bg is derived from solving the equations of equilibrium with the assumption that the soil is at yield. SIII is spherically symmetrical around the centre 0 of the tunnel face. The stresses are isotropic inside sphere S, and outside sphere S,, where they are equal to rrT and us respectively. Between S, and S, the radial stress ur and both tangential stresses bg, B, are determined as for SII by solving the equations of equilibrium with the assumption that the soil is at yield. SI, SII and SIII all satisfy the equations of equilibrium and boundary conditions of the present problem. Therefore a lower bound solution can be found from these three stress fields by assuming that the yield criterion is not exceeded in the soil mass. The derivations of these lower bounds can be found in the paper by Leca & Panet (1988). The result can be put in each case in the form of a double inequality. For stress field SI and Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. 588 LECA AND DORMIEUX os+yv-Y) ~s+Y(+Y) +!I a__________ I -+ Section A A-A (b) Fig. 9. Stress fields (a) SI, (b) SII and (c) SIII Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. FACE STABILITY OF TUNNELS IN FRICTIONAL Table 1. Values of (C/D)* for collapse and blow-out. For stress field SII l'y +2;+1 P ( > (Kp- l)a&, + 1 WP - lb,/% + *(i’Kp- l) ~ (Kr - l)a$a, ( > zw- < ( > 2;+1 + 1 + 1 1) (14) (15) and for safety against blow-out N,Q,+N,Q,~QT (16) In relations (15) and (16) Qs, Q= and Q, are defined as in equations (7). The values of N, and N, related to the best lower bound will be referred to respectively as Nsc-, NYC- for collapse and Nab-, NYb- for blow-out. A closed form solution can be obtained for these coefficients; for stress field SI, we can find for collapse = K, (174 NYC- = K, and for blow-out Nsbm = K, NbY (17c) =K$ D (174 For both SII and SIII, Q, is not present in the solution and NYC-, NYb- are thus equal to zero. The values of NSC-, Nsb- associated to SII are, for collapse N,‘- = K, The coefficients Nsb- and Nsb- associated are, for collapse to SIII 2(1-KP) It can easily be shown that the three sets of inequalities can be written with the same general form as for upper bound solutions. i.e. for safety against collapse &Qs+N,Q,GQ, (C/D)* blow-out 20” 25” 30” 35” 40” 45” 2;+1 (KP - l)or/ck (C/D)* collapse 1 1 For stress field SIII Ns’- 589 SOlLS (lga) N,‘-= 2;+1 ( (19a) > and for blow-out Z(1-W Nsb- = 2; ( + 1 > Relations (17) to (19) show that the estimated lower bounds all depend on the depth ratio C/D and the friction angle 4’ of the soil. The values of Nsc-, N,‘- and Nsb-, NYb- have been computed for 4’ = 20”, 25”, 30”, 35”, 40” and 45” and plotted as a function of C/D (for C/D < 3.0). These plots are shown in Fig. 10 for the general case (y > 0) and Fig. 11 for the case of a weightless soil. Figs 10(a) and 11(a) refer to collapse and Figs 10(b) and 11(b) to blow-out. The results obtained from stress fields SII and SIII can be compared on Fig. 11. For these conditions, NYCand NYb- are equal to zero, and the best lower bound is obtained when N,‘- is at minimum for collapse and when Nsb- is at maximum for blowout. It is found, as in the case of cohesive material (Davis et al., 1980), that this best estimate of the lower bound is provided by SII for shallow tunnels and SIII for deeper tunnels. The values (C/D)*of the depth ratio, for which both stress fields lead to the same value of Nsc- or Nsb- are given in Table 1. DISCUSSION From an engineering point of view, the parameters Qs and Q, are imposed by geometric and loading conditions, and the supporting pressure oT (i.e. the parameter QT) should be chosen such that failure of the tunnel during construction is prevented. The upper and lower bound solutions derived can be written N,Qs+N,Q,GQ, and for blow-out (19b) (20) for collapse and (lgb) NsQs + N,Q, 2 QT Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. (21) LECA AND DORMIEUX 590 N”,-, N”,- i - w, ---- N”,- 2 -,N:“Nb, i - Iv,- --__ e- 20 (b) Fig. 10. Lower bouud values of N, and N, (7 > 0) (a) for collapse and (b) for blow-out Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. FACE STABILITY OF TUNNELS IN FRICTIONAL N,C~/:’ 0, ---30 - SII ----- Sill 591 SOILS Sill , 01 0 I , , , , , 1 ’ I I I I 2 I I I I I I 3 (b) Fig. 11. Lower bound values of N, in a weightless soil (a) for collapse and (b) for blow-out Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. C/D LECA AND DORMIEUX 592 for blow-out. This means that the value of Qr at failure can be written (2-a QT=NsQs+N~Q~ where Qr, Qs and Q, are given by equations (7) and Ns, N, are weighting coefficients for loads Qs and Q, that can be bracketed from the values of NsC- , Nsc+ , N,C- and N ‘+, or Nsb-, Nsb+, NYband N, bt derived previoYusly. In other words the problem of the stability of the tunnel front can be analysed by the same methods used for determining the bearing capacity of a foundation. The analogy applies better to blow-out since the ground is then failing when Qr becomes too large. Equation (22) actually provides a lower estimate Qr- and an upper estimate Qr+ of the ultimate load Qr*. Qr- is found by using the lower bound values NsC- and N,,for collapse and N b- and N b- for blow-out; Qr+ is found by using the up& bound values Ns’+ and N,,‘+ for collapse and Nsb+ and Nrb+ for blow-out. In order to reduce the uncertainty on this estimate of Qr*, the general lower bound solution (y > 0) can then be improved by making use of the solutions obtained in the case of a weightless soil. Relations (20) and (21) can also be written N,G+N Qr N,s+N QT &l ’ QT ’ &l (24) QT. A Q,./ QT 5- I Upper I Therefore only two loading parameters Qs/Qr and Q,/Qr need to be considered and the stability of the tunnel front can be investigated in the loading plane (Qs/Qr. Q,/Qr) by braketing the domain Xx’ of the load combinations (QJQr, Q,/Qr) that are stable. The procedure is shown in Fig. 12 for the case 4 = 20”, C/D = 0.5. Lines A,E, and A,& in Fig. 12 represent the lower bound solution provided by the general stress field SI (y > 0). Since Xx’ is convex and contains A,, B, and B,’ (best lower bound solution for y = 0, i.e. Q,/Q, = 0) all points between lines A,B, and A,B,’ belong to X’. For the same reason, all points between lines A, B, and A, B,’ also belong to X’. The lower bound estimate Xx’- of X’ can in this way be extended from the domain AlBIB A, to the domain A,B,‘B2’A,. From a practical point of view, this improved lower bound solution can be obtained for collapse, by rewriting relation (20) with NYC- taken equal to the value found in case SI and with Nscequal to the best solution found in the case of a weightless soil (i.e. in case SII or in case SIII). For blow-out, it can be obtained by rewriting relation (21) with NYb- equal to the solution found for SI, and N,“- equal to the solution for a weightless soil. This improved lower bound solution is compared with the upper bound solution on Figs 13(a) and 14(a) for Ns values, and Figs 13(b) and 14(b) for N, values. For any value of the friction angle 4’ between 20” and 45” and a depth ratio C/D between 0.0 and 3.0, the failure load Qr* can be bracketed by I bound I I I for collapse I I 4--- Lower . 3. bound Best lower and Sill from bound SI from Fig. 12. Improved lower bound solution for the case I$’ = 20”, C/D = 05 Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. SII FACE STABILITY OF TUNNELS IN FRICTIONAL 4 1 1 / /’ I’ /’ f’ / /’ / / /’ /’ ,,’ ,I’ / I’ ,-’ I /’ / /’ / / /’ _,’ ,/’ I’ 0 +“- ./’ .’ /’ .’ /- I- ,’ 0,’ ,.’ rp’ = 20” Y- OI ,I’ r’ ,,_. /’ ./’ / ,Q.- //’ ’ /’ ,- /’ ,’ ,/’ TX’ ho”,’ ,G Q, x’ ,’ ,,/ /’ / /’ ,,I’ ,/’ /’ / / I Q/’ ,I’ ,/’ ,/’ /’ /’ -,~‘,,,~‘//rl ,,’ ,/ / /’ ,,’ ,’ ,’ 0.5~,/’/’ // / & ,/’ ,/I 6!/ / , /’ Y 4, /’ ,/’ 1’ ’ /’ I’ I /’ NC,+ ,/‘, if I’ /’ / - r$ I’ 1/,/ 0, I w- I o / 8, i --__ 593 SOILS fp’ = 25” 1 (b) 2 , $5’ = _ @’ = -rp’ = ‘4)’ = 30” 35” 40 45 CID Fig. 13. Bracketing values of (a) N, and (b) N, for collapse Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. 594 LECA AND DORMIEUX NS _____ t - N, Nb+ 5 -_-- &it- - Nb' Y 40 30 20 io_ q’ = 25” _ @’ = 20” O( 1 (W 2 Fig. 14. Bracketing values of (a) Ns and (b) N, for blow-out Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. FACE STABILITY OF TUNNELS using equation (22) together with Figs 13 and 14. Using the values of NSc- , NYC- from Fig. 13 provides a lower estimate QT of QT* for collapse. Using NSc+, NYC+ provides an upper estimate QT+ . The weighting coefficients for blow-out failure are given in Fig. 14. QT- is obtained by substituting Nsb- and NYb- for N, and N, in equation (21), and QT+ by substituting Nsb+, N Yb+ for N,, N,in this equation. The case of a tunnel driven in a cohesionless material is of special interest. In this situation ~c is equal to zero and it is not possible to consider the dimensionless coefficients a Ja,, U&T,- and yD/a,. As a result, coefficients Q,, & and G, are not defined. However. it can easilv be seen that the pressure at failure CJ~ can be estimated by writing oT= N,a,+N,yD (25) and by choosing N, and N, from Figs 13 and 14 as described previously. COMPARISON WITH EXPERIMENTAL RESULTS Having provided a simple method to estimate the value of the face pressure bT at failure, we may now apply it to typical conditions for the experimental study of face stability of tunnels in sands. In this way one can quantify the differences between the predicted upper and lower bound values and establish the validity of the method. In this application, it must be remembered that oT acts as a retaining load for collapse. Therefore upper bound estimates of the pressure at failure (iT* should be expected to be smaller than both oT* and the lower bounds. Centrifuge tests have been carried out in Nantes, France, to study the face stability of tunnels in sands (Chambon & Corte, 1989). In these tests, the tunnel was modelled as a rigid cylinder, a soft membrane covered the front part of the cylinder and allowed a supporting pressure + to be applied to the face. The centrifuge was operated at 50 g; at this acceleration level, the 80 mm cylinder modelled a tunnel with a 4 m Table 2. CID 1.0 1.0 2.0 2.0 IN FRICTIONAL SOILS internal diameter. Failure at the face was induced by decreasing the face pressure Go. The soil used for the experiments was a dry fine sand (Fontainebleau sand). The pressure gT. was obtained by filling the cylinder with air (uniform pressure) or with water (hydrostatic pressure). A surcharge os could also be applied on top of the model. Test results reported by Chambon & Corte (1989) show that (1) failure is sudden; (2) it occurs when the face pressure is decreased to a small value r+/ (a few kPa); (3) for the range of values C/D has little influence on the limit considered, pressure; (4) the failed area is bulb-shaped with its largest dimensions at face level; (5) this geometry is not affected much by C/D or soil density; and (6) failure does not reach the ground surface for C/D > 1.0. The tests with compressed air support were run at two depth ratios, C/D = I.0 and C/D = 2.0. Two soil conditions were examined. A loose sand (7 = 15.3 kN/m3, D, = 62%), and a dense sand (y = 16.1 kN/m3, D, = 86%). Shear strength tests on these soils had shown that c‘ = 2.3 kPa, 4’ = 35.2” for the loose sand, and c’ = 1.1 kPa, 4’ = 38.3” for the dense sand. The results obtained from the four tests are summarized in Table 2. The limit analysis estimates of the critical face pressures for these four tests are also recorded in Table 2. It is apparent that lower bound solutions are significantly higher than the upper bound values, as well as measured pressures at failure. On the other hand, the upper bound estimates gT+ are in close agreement with test results, with uTt values slightly lower and almost identical to the pressures c+/ measured at failure in the centrifuge. Other similarities between upper bound solutions and experimental results are shown in Fig. 15, in which the failure zone observed in the centrifuge along the tunnel centreplane is represented for the case of a loose sand for C/D = 1.0. The critical geometry associated with the best upper bound solution is shown with dashed lines. Even though it does not extend in the vertical direction as much as the actual failure area, it coincides almost perfectly with the observed surface in front Comparison between predicted and measured pressures at fake y: kN/m3 15.3 16.1 15.3 16.1 595 Critical pressures predicted from limit analysis: kPa QT- + UT 29 29 46 44 2 3 2 3 Measured pressures at failure in the centrifuge: kPa OI 6 3 4 4 Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. LECA AND DORMIEUX 596 1 \\\\\\\\\\\\\\\< ---- Failure atea observed Crltlcal failure surface in the centrifuge from llmlt analysrs Fig. 15. Comparison between theoretical critical surface and observed failure area of the tunnel. In particular the extent of failure ahead of the tunnel face is the same as that observed. The large amount of failed material in the centrifuge over the tunnel crown, however, could have resulted from the progression of failure in unsupported ground once face collapse had occurred. CONCLUSION The limit analysis concept has been used to examine the stability conditions of the face of a shallow tunnel driven in a frictional material. Safety against both collapse and blow-out have been analysed. Upper bound solutions have been derived from consideration of three failure mechanisms based on the movement of rigid blocks with conical shapes. The amount of material involved in these mechanisms is limited, but such geometries could be representative of initial ground movements that could lead to larger scale failures. In particular the results suggest that if the predicted failure conditions are close to the actual ones, the surcharge us has little effect on face stability (except for very shallow tunnels), and the extent of the failure zone in front of the tunnel is smaller than in the case of a long open cut. The upper bounds have been compared with lower bounds published previously (Leca & Panet, 1988). In both cases the problem reduces to one of two loading parameters Qs/Q= and QYIQT,with Qs, Q, and Qr defined above. It is found that it can be treated similar to the method used for determining the bearing capacity of foundations, i.e. the critical supporting load is equal to QT=bQ,+N,Q, (26) with N, and N, estimated from Fig. 13 (collapse) or Fig. 14 (blow-out). These figures actually allow one to bracket N, and N, between lower bound values N,‘- (or N,“-) and N,,- (or N,b-) and upper bound values NQ+ (or Nsb+) and N,‘+ (or NYb+). This way a lower estimate and an upper estimate can be found for the collapse load QT. All conclusions remain valid in the particular case of a cohesionless soil as long as Qs, Qr and Q, are assumed to be equal to es, eT and yD respectively. The method has been applied to centrifuge tests for the face stability of shallow circular tunnels in sands. Reasonable agreement has been found between the theoretical upper bound estimates and the face pressures measured at failure in the tests. Other similarities are evident between the critical failure mechanisms derived from limit analysis, and observed failure areas in the centrifuge. These conclusions support the idea that the upper bound solutions are closer to the actual pressures at failure than the lower bound values, and can provide reasonable estimates of critical face pressures. APPENDIX I. DERIVATION OF THE UPPER BOUND SOLUTIONS ASSOCIATED WITH MECHANISM MI Geometric properties First, some geometric quantities need to be determined, prior to the derivation of the external power B, and the dissipation power P, associated with mechanism MI (see Fig. 16). For this purpose it is more convenient to consider two cones W and ‘8’. Both cones have the same apex R and the same axis A, but the base C of V is in the Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. FACE STABILITY Fig. 16. OF TUNNELS SOILS 597 Geometry of mechanism MI plane of the tunnel face (xOy), whereas the base r of W is at the ground surface. The moving block ~8 corresponds to cone V minus cone r. Of course only one cone, V, needs to be considered when the ground surface is not reached. Two axis systems are used: (0, x, y, z), associated with the tunnel and (a, X, Y, Z), associated with the cones. The equations below allow transformation between both co-ordinate systems. x=x (27a) Y = 4 sin a tg@ - y cos a + 2 sin o! (27b) D cos a - y sin a - 2 cos a (27~) Z=- IN FRICTIONAL 2 tsdf Both cones are characterized by the same equation with respect to co-ordinate system (Cl, X, Y, Z) x2 + Y* = tg%#J’z* Therefore, the area J& of the cone base is ~ _ ~0' J[cos (a - 4’) cos (a + &)] 4 The same parameters v’. Its height h’ is sin 2ci - F can be determined for cone sin 2f#/ h’+ sin 2& c’ is an ellipse, the semi-axis D sin 2a - y lengths of which are sin 24’ (33a) ” = Z 2 sin (a - 4’) sin (a + 4’) D (28) sin 2~ - g ” = Z 2 cos #&sin The height h of cone V is (31) cos l#J’ sin 24’ (a - 4’) sin (a + #)I (34b) h = D cos (a - 4’) cos (a + 4’) sin ~C#J (29) Its area ~4’ is 2 Its intersection with the tunnel face (Fig. 16) is an ellipse for which semi-axis lenths a and b are given by KD= d’ = 4 4 sin 2a - $! ( cos #[sin sin 24’ > (a - 9’) sin (a + @)1312 D (304 a=- 2 b = D JCcos (a - 4’) ~0s (a + 4’11 2 cos 4’ (35) The volume Y, of the block is equal to (30b) Vs=Y-V-’ Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. (36) LECA AND DORMIEUX 598 where $‘” is the volume of cone W, and Y’ is the volume of cone v’. y _ nD3[cos(a - 4’) 12 qs’)]3’2 cos (u + (37) cos 4 sin 24’ sin 24’ > ’ cos 4’ sin 2&[sin (a - 4’) sin (a + #)13” The lateral area 9, (38) of the block is equal to Y,=Y-Y (39) where Y is the lateral area of cone %‘, and Y’ is the lateral area of cone v’. An expression for Y as well as Y’ is found by using complex number integration x-cos a [cos(a - 4’) cos (a + $‘)I”’ (40) ml u. 4 2 The expressions for Yb can be simplified the following parameters are introduced JCcos(c(- 4’) cos(a + 401 cos fp’ R = cos (a - &) cos (a + #I B dpv ~=C’Y-ncotc$’ (46) where n is the unit vector normal to the discontinuity surface at the point where dP,/dC is computed (Fig. 3). Since the angle Od between the velocity Y and the discontinuity surface is chosen equal to $‘, equation (46) can be written dP --I = C’V cos Cp’ dZ 9, = 8, loads 9. has three com- + 8, + Yr (48) sin 2a - E sin 24’ > ( ’ [sin (u - 4’) sin (a + #‘)13” R, = (45) where 8, is the power of the retaining pressure trr , Ps is the power of the surcharge rrs, and S’r is the power of the soil weight y. sin #J’ cos 4’ f V= Ve, Power of external loads The power of external ponents sin 4’ cos 4’ y’ = c all points Plastic energy can only be dissipated along discontinuities and the dissipation energy per unit area dP,JdZ is 3 sin 2a - i f Velocity field The moving block is rigid. Therefore of the block have the same velocity sin 24’ (41) once B, = (are3 ss I: WW (49) = -o,vcosad After substituting (424 . (Ve,) dC pT= -$cos expression (31) for & we find: aR,a,V The same method is used together with equations (35) and (43) for determining 8, and qY. cos ci R, = - (42~) w R, = sin 2a - F R, = &sin Equations (51) sin 24’ (42d) t--q) 2(a - 4’) sin 2(a + &)] (Vez)dV (42e) =yVsina*Y, (36) and (39) then become (52) that is to say (43) $/,=-- R,= R,’ 1 - 2 co3 f#hga cos Cp’ RE3 nD= R,Rc 4 1 8, = $ 2 sin a sin 24’ Rn2 R., R, 0, V (53) Rn3 %=4 nD2 sin a - 3 R,R.[I - ($+V Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. (54) FACE STABILITY Finally equation OF TUNNELS 2 sin a sin 24’ -%?-J Dissipation power The dissipation nism MI is N, = os D,V(55) RB(, -; power y associated with mecha- P, = (56) where dP,/dx is given by equation (47). Therefore P, = C’V cos 4’9, in equation (44). If we for 9, in the above Ro2 R,* 1 - 2 cos* #tgci ~ RE3 1 c’V (58) Upper bound theorem The upper bound solutions associated with mechanism MI are found by writing relation (6) with 8, and P, computed as in equations (55) and (58) respectively. This substitution leads to the inequation RD*R, ~ RE3 + +tgzR, 599 thus 1 Rn2 cos (24’) - cos (2a) R, tgcl N, = $R, (61) tga The best upper bound solution for MI is obtained by choosing n such that N, and N, are at maximum. These expressions only apply to cases where failure reaches the ground surface; that is to say for depth ratios C/D such that 2$+147 sin 2a sin 24’ If C/D is larger than [(sin Zu/sin 24’) - 1]/2 relation (60) remains valid provided R, is set equal to zero in equations (61) and (62). (57) Y, has been determined substitute this expression equation we find 2tga sin 24’ - SOILS where Ns and N, are obtained (48) can be rewritten + +x IN FRICTIONAL [ 0, c’ cot gr$’ 1 - R,3 RE3 cr - ~ c’ cot g# 1 YD x,=x (5% If we introduce the soil unconfined compression strength uf and the Rankine earth pressure coefficient K, (equations (2) and (4)) relation (59) can be rearranged in the form E 1 +N,(K,-l)$ < (&, - 1) aT + 1 (60) n, (64a) Y, = i sin atg# - y cos a + z sin a c’ cot g@ R,* R,* < 1 - 2tga cos2 I$’ ~ RE3 APPENDIX II. DERIVATION OF THE UPPER BOUND SOLUTION ASSOCIATED WITH MECHANISM MI1 Geometric properties Two cones are considered, vi (apex a,, axis A,) of which the base x1 is in the same plane as the tunnel face; and V, (apex Q2, axis A2) of which the base El2 is in plane 7~(U, and %?*have the same geometric properties) (see Fig. 17). The first moving block .%3icorresponds to the portion of %?i located below plane x and g2 is the portion of %?zlocated below the ground surface. Four axis systems will be referred to: (0, x, y, z) associated with the tunnel front; (Q,, Xi, Yi, Z,) associated with vi; (a,, X2, Y,, Z,) associated with %:,; and (B, X’, Y’, Z’) associated with planes II and R’. The following co-ordinate transformations are used D cos a - y sin a - z cos a 2 tg4’ z,=X,=x’ (64b) 0% (654 y _ -Dcos(a+@)sin&cos/? 1 2 sin (B + 4’) cos 4’ - Y’ sin p - z’ cos jl -Dcos(a+#~‘) ” = 2 sin (@+ 4’) Wb) cos 4 sin p 7 sin 4’ + Y’ cos /I - z’ sin /3 Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. (65~) LECA AND DORMIEUX 600 Fig. 17. Geometry of mechanism MI1 Since plane rc is chosen such that A2 is vertical, we have between a and p the relation The area &,, d 28 - a = n/2 Wb) 12 - of X:,, is 0 cos’ (a + 4’) &sin (B - $‘)I cos fp’ [sin (B + @)1312 4 The heights h, and h, of cones V, and V, respectively are h = D cos (a - 4’) cos (a + 4’) 1 (67) sin 24’ (72) The intersection C, of Wz with ground surface is a circle of which the radius r2 is equal to D h = D sin (/I - 4’) cos (a + 4’) 2 sin 24’ (68) ‘* = 2 W, intersects with the tunnel face along an ellipse C, (Fig. 17) of which the semi-axis lengths are D a, = - (@a) 2 b 1 = g JCcos (a - 4’) cos (a + 441 cos Cp’ 2 sin /? cos a sin 4’ sin @ + 4’) - 2;+1 tg@ ( >I Its area CZI, is sin b cos a 4 sin (j? + 4’) (69b) The area zZI of X1 is ~ 1 _ aDZ J[COS (a - 4’) cos (a + @)] cos 4’ 4 (74) (70) The intersection Z,, of V, with plane n is also an ellipse (Fig. 17) with semi-axial lengths a ,2=D2 (73) cos (a + 4’) sin (.jI + 4’) P’lb) Since .G’#ris obtained by removing a cone identical to V$ to cone V,, its volume Vt,, and its lateral area Y,,, can be put in the form “Vbr = W”, - Vz (75) Y,, (76) = 9, - ,4v, where V, and V2, and 9, and 9, are, respectively, the volumes and lateral areas of cones %‘r and Wz. Y,, ^v,, 9, and 9, are determined by Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. FACE STABILITY using the same methods = c as in Appendix OF TUNNELS I, i.e. IN FRICTIONAL SOILS 601 These coefllcients allow one to write Yb,, Y,, and Y,, in a simplified form [cos (cz- 4’) cos (a + @)]3’2 Y,, = $ (R/, R, - s R,‘) (90) R, - cos (a + qY)Rc R, 2 Y, [cos (a - 4’) cos (a + #)]“2 (79) sin 4’ cos 4’ Y,=-_- nD2 sin p cos’(a + 4’) 4 sin 4’ (80) The volume Vb2 and the lateral block .??I2are determined thus area Y,, of sin 4’ cos (a + 4’) *ybz=V2-ly3 (81) .4pb,=c4p2-Ys (82) where Y2, 9, are given by equations (78) and (80), and V”, and Y, are, respectively, the volume and the lateral area of the portion of V2 located above the ground sin 4’ sin (B + 4’) (83) nD2 sin 4’ y,=-- 4 cos2 4’ sin /? cos a x 2 B = JCcos (a - 4’) cos (a + cos l$’ 4’11 = cos (a - 4’) cos (a + 4’) sin (24’) R c = cos;o:+/) [;z ;; VelocityJeld Both BI and a2 are respective velocities are rigid blocks ;;]I” co2 4’ cos (a + 4’) R, - $ the VI = be,, (94) V2 = bez2 (95) Since W, and g2 are not moved at the same speed, a discontinuity in velocities is created along their intersection YZ12,and a relative velocity V,, needs to be considered between gI and B2 (Fig. 18). In order for relation (5) to be satisfied, the surface angle between V,, and the discontinuity XI2 will again be chosen equal to 4’. This implies the following relations between VI, V, and VI, the fol- v,= sin (D + 4’) v sin (j3 - 4’) (96) ’ (85) (86) (87) sin j3 (88) RD = sin 4’ sin (j3 + 4’) R, = and (84) sin 4’ sin (B + 4’) It is convenient at this point to introduce lowing coefficients A (92) sin 4’ cos (a + 4’) cos f#l’ &sin (B - &‘)I x [sin @ + @)13/* R 1 +cos. x R Yob,, sin 4’ (89) Fig. 18. Relative velocities of a, and 9, Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. 602 LECA AND DORMIEUX cos &:2= CL sin @ - I$‘) (97) V2 p,=-- aDZ 1 cos’ (CL+ 4’) 4 3 cos2 4’ R,R, + cos 0: X sin a ~,2 Plastic energy can be dissipated along the lateral area of Y?r and Wz and along Z,,. In all three cases the dissipation energy per unit area is (cf. Appendix I) cos f#J’cos (B + 4’) R x 2 sin 4 sin (B + 4’) dpv E C’V cos (fi = ’ cos2 4’ (98) cos (a + 4’) withV=Vion%?,,V=I/,onV,,andV=I/,, on Zi2. 3 Rc2 - 2 g sin 4’ 2 sin f$’ cos2 (cr + 4’) rDv, 'I (105) Equation Power of external loads The power of external 9, = (99) can then be written nD2 cos2 (a + 4’) 4 loads 8, is 9, = Yr + 8s + Yy co? f#l’ RE2 x cos2(a + 4’) (99) where Yr, 9, and S’r are defined as in Appendix I. + cos a 0 s-R,~RA~T RA4, sin a - Rc2 8,= (Vlez,) dC (-~TeJ + cos a cos 4’ cos @ + 4’) R 2 sin C#J’ sin (J + 4’) ufw = -crTVl cos ad, ’ 22 1v, Rn3 _ 2 sin 4’ cos’ (a + c#J’) > 3 9s = (V2ez,) dC (-u,e,J = a,V*d* 9, = (101) (-re,Wd Dissipation power The dissipation energy associated nism MI1 can be written d-lr ( - re,XV,ez,) dv + = yV, sin czYb, + yV2’zy,, P” = p,, + p,, + P,,, (102) After substituting equations (70), (74), (90), (92) and (96) for &‘,, d,, Yb,, Y,, and VI we find = 9 _ T e COS c( CO2 RA UT v, RC2 zz co? x I$’ sin # P2” Rc2 cos (a + c#/) P 12v (108) = = CT, cos 4’ ,40b, sin’ 4 D2 cos2 4’ ; = C’V, cos f$’ P,, (103) (107) contri- PI” = (a + +) x - 8, where PI,, P,, and P,,, are the respective butions of gi, SJ2 and X,2 cos2cp 4 with mecha- (109) = = C’V12cos I#/ d,, Equations (91), (93) and (72) together equations (96) and (97) allow one to Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. (110) with write FACE STABILITY equations p cosz (a + 4’) 4 IN FRICTIONAL cos2 Cp’ xcr cos c#l’V, 1 (111) (a +4’) p=gcos2 cos a R, X 2” - 1- cos sin 6’ R, 4 cos2 x (a: + 2 N, (K, - 1) ; C RcRD (a + 4’) sin f#~’~0s’ (a + 4’) 1 N, = x c’ cos r#J’V, -- 9 12” - (112) 4 cos2 sin /3 x cl cos 4’ RC Y v, X cos24’ RE2 cos u R, --_ sin 4’ RC2 sin 4’ cos2 (a + @) x cl cos l#l’V, 1 RE2 + sin tl- a -= R,aT RARB RC2 2 sin 4’ sin @I+ 4’) < RE3 ’ YD 2 sin 4’ cos’ (0: + 4’) > 3 R, RE2 cos lx RC2 cos2 (a + 4’) > x c’ cot g& cos 4’ cos (B + 4’) Rc3 2 sin 4’ sin (B + #) R, 1 sin (j3 - 4’) RE3 ’ sin (B + #) R, (118) Relation (116) provides the best upper bound associated with MI1 when a is chosen such that N, and N, are at maximum. As in Appendix I, the above results only apply when the ground surface is reached by the failure mechanism, that is to say when 5 < cos (a + #) sin (B - 4’) D’ 2 sin c$’ sin (B + 4’) (119) For deeper tunnels (relation (119) reversed) expression (116) is still valid provided R, is set equal to zero in relations (117) and (118). APPENDIX III. DERIVATION OF THE UPPER BOUND SOLUTION ASSOCIATED WITH MECHANISM MI11 Geometric properties Two cones, $? and v’ are considered (see Fig. 19). Both have the same appex n and the same axis A. The moving block SI corresponds to cone W minus cone v’. %’ intersects the tunnel face along an ellipse z’ (base of cone W) and the ground surface along another ellipse x (base of cone U). Two axis systems are used: (0, x, y, z) associated with the tunnel, and (a, X, Y, Z) associated with the cones. The co-ordinate transformation between both systems is given by the following equations. s RC2 + cos a cos 4’ cos (B + I$‘) R - (117) (114) Upper bound theorem The upper bound solutions associated with mechanism MI1 are obtained by substituting equations (106) and (114) for B, and P, in relation (6) cos2(a + 4’) [ (113) nD2cos’ (a + 4’) 4 N, and N, are in the cos a 60s’ 4’ sin (B + 4’) R, tgaR, + (116) - 2 sin 4’ cos a cos’ 4 1 Substituting equations (1 ll), (112) and (113) for pi,, p2, and pi,, in equation (107) leads to p _ 1 1 f#J’ 2 cos /I sin 24 cos (a + 4’) Ro2 X l)%+ ec c sin @ - 4’) RE2 1 N, = f cos’ (a + 4’) nD2 <(K,- The weighting coeflients present case equal to cos2 4’ cos 1 + 1 + N,(K, - 1) F 4 RE2 - 4’) 603 SOILS Relation (115) can be rearranged in the same form as in Appendix I once the unconfined compression strength cr, and the passive earth pressure coefficient K, are introduced. (108) (109) and (110) in the form = e I” OF TUNNELS (115) x=x Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. (120) 604 LECA AND DORMIEUX Fig. 19. Geometry of Mechanism MI11 Y = t sin atgf#/ + y cos a - 2 sin a D cos w. z=-- 2 44’ The semi-axis lengths of ellipse c’ are D a’ = - + y sin u + 2 cos a The height h of cone V and the height h’ of cone v’ are sin 2a + g D e JCcos(a - = 2 44 ~0s (a + &‘)I (129) cos 4’ of (123) sin 24’ = b’ Equations (126H129) allow determination the area S’ of x and the area &’ of r sin 24’ h=; h’ (128) 2 cm (a + 4’) cm (a - 4’) sin 2q5’ (124) &cjf=E!c 4 The equation of cones V and v’ with respect the co-ordinate system (a, X, Y, Z) is x2 + Y2 = rgSj5’22 2 ( sin 2cr + T sin 24’ > 4 cos @[sin (a - 4’) sin (a + &)13j2 to (125) (130) &, = 9 JCCOS (a + 4’) cm 4 (a - f$‘)] cos 4’ (131) The semi-axis lengths of ellipse I: are D sin 2a + $ a = Z 2 sin (a - 4’) sin (a + 4’) sin 2a + F (a - 4’) sin (a + #)I of block (126) sin 24’ b=D 2 2 cos $‘J[sin The volume Y, and the lateral area 9, g are found by writing sin 24’ (127) V,=Y-v-’ (132) 9, (133) = Y - 9’ where V” and y, and V’ and Y’ are the volume Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. FACE STABILITY and lateral areas respectively W >I 3 [( y-=e of cone V and cone sin 2a + g f OF TUNNELS sin 2 4’ 12 sin 24’ cos 4’ (134) sin a (a + 4’) sin (a - C/J’)] X l ^I” = e [cos(a - (135) I dP -=c’vcos~’ dZ (144) with 8,, 8, and 8, defined as in Appendix I. After substitution of equations (139) to (141) for &, &’ and Y, we find (136) (146) 2 .V’=+cosa g?,= --- nD2 R, Rc2 4 x YY = - G As in Appendices I and II, Y and Y’ are determined by using complex integration. Equation (130) to (133) can be put in a simplified form by introducing three coefficients R, = cos aJ[cos (a + 4’) cos (a - c#J’)] (138a) R, = sin a&sin (a + 4’) sin (a - @)] R = sin 2a + (2CID + 1) sin 24’ E cos 24’ - cos 2a The expressions (138b) (EF)‘-(SJDv sin a sin 24’ cos $J’ (148) and equation (145) can be written I RB Rc2 -------a,+-0 cos 4’ R.4 cos 4’ T _isin-a(E)3-(2Y% sin 24’ cos I#J’ I V (149) 4 sm a cos f$’ R.4 (139) 1 x[(zz>‘-(ZJ] that can be dissipated is given by equation (144) P, = C’V cos qb’Yb (141) (142) in (150) where dP,/dX ~0~ R, Rc2 - R, 12 sin 4’ cos 4 Dissipation power The plastic energy mechanism MI11 is (140) 12 sin 24’ cos 4’ 9,=- 2 (138~) RBRc2 sin a cos 4’ y,x, (147) 05 v for &, zZ’, Vb and Y, are then 4 .d’=aD2 cos qs JCcos(a + 4’) cos(a - 4’11 sin 4’ cos f$’ &I!!?, (145) 4’) cos (a + &)13” sin 24’ cos f#i 12 (143) Y’, = YT + Ys + YY sin 24’ cos 24’ - cos 2a Velocityjeld The velocity of moving block _6gis Power of external loads The power of external loads Y’, can be written sin f$’ cos 4’ sin 2a + T 605 Since this is a rigid body motion, plastic energy can only be dissipated along the lateral surface of &?; the dissipation energy per unit area dP,/dX is 2 x &sin SOILS V= Ve, x [sin (a - 4’) sin (a + @)I”” Y = $ IN FRICTIONAL (151) After substitution of equation equation (151) we find p, = q [R, Rc2 - (142) for 9, C’V &I 7 sm 4’ Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved. in (152) 606 LECA AND DORMIEUX Upper bound theorem The upper bound solutions associated with MI11 are found by writing relation (6) with 8, and P, as computed in equations (149) and (152) K, K, y 0, or D -R,Rc2s-+RR,~ cos I$’ cos 4’ H C P Ye P, (RBRc2 - RA) -T---- (153) can also be rearranged into the < This inequality form N, (K,, - 1) ; sin 4’ 1 + 1 + N,(K, >(K,- - 1) F l)z+ 1 (154) where IS=and K, refer to the unconfined compression strength and the passive earth pressure coefficient of the soil respectively, and N, and N, are N,=- 4, Rc2 RA N,= 3R;;J2,‘[(g-($J] (156) (155) The best upper bound associated with MI11 is found by choosing a such that N, and N, are at minimum. Since failure always reaches the ground surface, expression (154) is valid for all values of C/D. NOTATION ’ soil cohesion soil friction angle ;I oc unconfined compression strength Rankine earth pressure coefficient (active) Rankine earth pressure coefficient (passive) soil unit weight surcharge pressure tunnel pressure tunnel diameter tunnel depth tunnel depth of cover unsupported length behind tunnel front power of external loads dissipation power REFERENCES H. (1967). Stability of clay at vertical openings. J. Soi/ Me& Fndn Div. Am. Sot. Civ. Engrs 93, No. SMl, 71-94. Chambon, P. & Corte, J. F. (1989). Stabiliti: du front de taille d’un tunnel faiblement enterrt: modtlisation en centrifugeuse. Proc. Int. Co@ Tunneling and Broms, B. B. & Bennermark, Microtunneling in Soft Ground: From Field to Theory, Paris, pp. 307-315. Clough, G. W., Sweeney, S. P. & Finno, R. J. (1983). Measured soil response to EPB shield tunnelling. J. Geotech. Engng Div., Am. Sot. Ciu. Engrs 109, No. 2, 131-149. Davis, E. H., Gunn, M. J., Mair, R. J. & Seneviratne, H. N. (1980). The stability of shallow tunnels and underground openings in cohesive material. GCotechnique 30, No. 4,397-416. Leca, E. & Panet, M. (1988). Application du Calcul a la Rupture a la stabilitb du front de taille dun tunnel. Revue FranGaise de Giotechnique, No. 43, 5-19. Muelhaus, H. B. (1985). Lower bound solutions for circular tunnels in two or three dimensions. Rock Mech. & Rock Engng l&37-52. Peck, R. B. (1969). Deep excavations and tunneling in soft ground. Proc. 7th Int. Con& Soil Mech. and Fndn Engng, Mexico, Balkema 3,225290. Salenqon, J. (1983). Calcul b la Rupture et Analyse Limite. Salenqon, J. (1990). An introduction to the yield design theory and its application to soil mechanics. European J. of Mech., A/solids 9, No. 5. 477-500. Schofield, A. N. (1980). Cambridge geotechnical centrifuge operations. Gbotechnique 30, No. 3, 227-288. Downloaded by [ Middle East Technical University] on [19/06/17]. Copyright © ICE Publishing, all rights reserved.