36 2. <strong>Stability</strong> Analysis of the <strong>Tunnel</strong> <strong>Face</strong>σ ′ h231zFigure 2.32: Possible horizontal stress distributions along wedge sidesMeas. Calculated s ′ divided by Measured s ′ a BestTest b s ′ (kPa) a ,0 a ,a b 1,0 b 1,a b 2,0 b 2,a c 1,0 c 1,a c 2,0 c 2,a ResultCS0.5/5 3.6 3.65 4.45 3.26 4.20 3.76 4.42 2.54 3.70 3.73 4.31 c 1,0CS0.5/5 3.3 4.00 4.88 3.58 4.61 4.12 4.85 2.79 4.06 4.09 4.73 c 1,0CS1/5 3.5 5.52 6.85 4.08 5.81 5.14 6.33 1.97 4.19 4.48 5.52 c 1,0CS1/5 3.0 6.41 7.95 4.73 6.74 5.97 7.35 2.28 4.87 5.20 6.41 c 1,0CS1/5 3.3 5.79 7.18 4.27 6.09 5.39 6.64 2.06 4.39 4.70 5.79 c 1,0CS1/10 7.4 5.13 6.35 3.77 5.40 4.78 5.89 1.83 3.89 4.15 5.13 c 1,0CS2/5 4.0 7.64 9.60 3.52 6.48 5.78 7.67 0.00 2.88 4.12 5.53 c 1,0 ,c 1,aCS2/10 8.0 7.66 9.63 3.53 6.48 5.79 7.68 0.00 2.88 4.12 5.54 c 1,0 ,c 1,aCS4/10 8.2 13.2 16.8 1.81 6.91 6.82 10.1 0.00 0.12 4.09 5.81 b 1,0CS4/13 13.4 10.9 13.8 1.40 5.58 5.53 8.17 0.00 0.02 3.31 4.70 b 1,0BS0.8/10 6.6 1.76 3.24 1.76 2.91 2.16 2.67 0.98 2.41 1.96 2.41 c 1,0BC0.8/10 17.0 1.30 1.50 1.04 1.29 1.22 1.39 0.67 0.99 1.09 1.25 b 1,0BC0.6/10 11.0 1.69 1.69 1.45 1.51 1.65 1.87 1.07 1.16 1.55 1.75 c 1,0a Format is a n,m : a = arching model (a = none, b = 2D, c = 3D), n = stress on wedge sides (1 = linear, 2 = incl.arching), m = K-value used (0 = K 0 , a = K a )b Format is ATn/m: A = Author, T = Type of Test, n = Overburden-Diameter Ratio, m = Diameter in m. Authors:B=Bezuijen [37], C=Chambon [60]. Type: S=Sand, C=Clay.Table 2.4: Comparison of centrifuge test results with models with different arching implementations
2.2. Wedge <strong>Stability</strong> Model 37with an overburden 2 times the tunnel diameter, the c 1,0 model again yields the closest match,but this time it gives an unsafe prediction of the required support pressure and the best safeprediction is obtained with the c 1,a model. For the deeper tunnels, with C/D = 4, the bestprediction is obtained with the b 1,0 model. In this case both the c 1,0 and the c 1,a model yield anunsafe prediction.The tests by Bezuijen on the other hand have been done in completely saturated sand (ϕ ′ =35.6 ◦ ) and soft clay samples (ϕ ′ = 23 ◦ , c ′ = 1kPa) with low overburdens, C/D = 0.6 or 0.8.The results of two tests are best predicted by the c 1,0 arching model. In one case this a slightlyunsafe prediction and the best safe prediction is obtained with the b 1,0 model. For the third testthe b 1,0 model yields the best safe prediction, although the c 1,a and the c 2,0 models are also veryclose.Based on this small number of tests it must be concluded that overall the c 1,0 model yieldsthe most accurate prediction of the minimal support pressure at which complete failure occursand the c 1,0 model will be used in further calculations. This model does however yield an unsafeprediction for several cases, especially the deeper tunnels with C/D = 4. In those cases the b 1,0model yields a better prediction. That the c 1,0 model yields the unsafe prediction that the faceis stable even without support may be due to an overestimation of the stress reduction causedby soil arching. Given the fact that the b 1,0 yields a safe prediction in all cases, this model maybe preferable from an engineering point of view. This choice may be further strengthened if itis desired to limit the deformations caused by the excavation process.The various arching implementations can be roughly associated with the amount of deformationthat occurs at the face. The support pressure found from the c models is the pressure atwhich total failure of the face occurs, at which point large deformations have already occurred.The b models on the other hand seem to yield the support pressure at which only limited deformationof the face occurs. Such limited deformations are needed to initiate the friction forcesand are inevitable in the boring process. The a models are equated to a situation where nodeformation has occurred and as a result the a model strongly overpredicts the minimal requiredsupport pressure. In combination with the effect of slurry infiltration and excess pore pressuresas described next, these models would yield extremely high required support pressures, resultingin extremely high overburden requirements, which are not backed by any field experience.The influence of the choice of the arching model and K y will be briefly illustrated in section2.3 for the reference case described there. Based on the limited evidence available, however,all further calculations will be made with the c 1,0 model in order to best predict the minimalrequired support pressure. The fact that it may yield unsafe support pressures, possibly due tovariations in the soil properties, should be overcome by the use of safety factors. And as is truein all cases where limit equilibrium models are used, the possible occurrence of failure by adifferent mechanism, in particular micro-stability, must be checked separately.2.2.3 Penetration of Support MediumAnother effect that has to be taken into account, in particular when boring in permeable soils, isthe penetration of the support medium into the soil. This infiltration may in turn lead to excesspore pressures in front of the TBM as well as a reduction of the effective support force.If a bentonite slurry is used, the bentonite should ideally form a thin impermeable filter cakeon top of the face. The entire support force from the slurry is then transported onto the soilskeleton at the face. In reality the slurry will always infiltrate the pores to a certain extent beforeclogging of the pores occurs and a filter cake is formed. The support pressure is then transportedonto the soil skeleton over this infiltration length. This situation is sketched in figure 2.33. In