13.07.2015 Views

Modelling Swelling Rock Behaviour in Tunnelling.pdf - Plaxis

Modelling Swelling Rock Behaviour in Tunnelling.pdf - Plaxis

Modelling Swelling Rock Behaviour in Tunnelling.pdf - Plaxis

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

TitleFoto by Christian Ammer<strong>in</strong>g<strong>Modell<strong>in</strong>g</strong> <strong>Swell<strong>in</strong>g</strong> <strong>Rock</strong> <strong>Behaviour</strong> <strong>in</strong> Tunnell<strong>in</strong>gBert Schädlich & Helmut F. Schweiger, Institute for Soil Mechanics and Foundation Eng<strong>in</strong>eer<strong>in</strong>g, Graz University of Technology, Graz, AustriaThomas Marcher, ILF Consult<strong>in</strong>g Eng<strong>in</strong>eers, Innsbruck, AustriaAlthough a great amount of practical experience has been ga<strong>in</strong>ed <strong>in</strong> the last decades, tunnel design <strong>in</strong> swell<strong>in</strong>grock is still a very challeng<strong>in</strong>g task, as the recent examples of the Engelbergtunnel <strong>in</strong> southern Germany and theChienbergtunnel <strong>in</strong> Switzerland demonstrate. Reliable prediction of swell<strong>in</strong>g pressures and swell<strong>in</strong>g deformationsespecially <strong>in</strong> anhydritic rock is extremely difficult due to the heterogeneity of the material and the complexity ofthe <strong>in</strong>volved transport mechanisms. However, modern design codes and eng<strong>in</strong>eer<strong>in</strong>g practice demand capacitychecks for tunnel l<strong>in</strong><strong>in</strong>gs, which usually can only be provided by numerical analysis with an appropriate constitutivemodel. Such a constitutive swell<strong>in</strong>g model, which adds swell<strong>in</strong>g stra<strong>in</strong>s <strong>in</strong> dependence on the stress level andaccounts for the time dependent evolution of swell<strong>in</strong>g, has been implemented for <strong>Plaxis</strong>. This article compares theresults of a numerical back analysis with this model to <strong>in</strong>-situ measurements, us<strong>in</strong>g swell<strong>in</strong>g parameters derivedfrom laboratory swell<strong>in</strong>g tests.rocks” are geomaterials which»“<strong>Swell<strong>in</strong>g</strong><strong>in</strong>crease <strong>in</strong> volume if water is allowedto <strong>in</strong>filtrate. The most prom<strong>in</strong>ent rock typesexhibit<strong>in</strong>g swell<strong>in</strong>g behaviour are certa<strong>in</strong> typesof claystone and anhydrite-bear<strong>in</strong>g rocks, whichcan be commonly found <strong>in</strong> northern Switzerlandand southern Germany. Tunnell<strong>in</strong>g <strong>in</strong> suchmaterials is notoriously difficult: If a flexible<strong>in</strong>vert l<strong>in</strong><strong>in</strong>g is <strong>in</strong>stalled, large <strong>in</strong>vert heaveevolves after tunnel excavation. In case thesedeformations are prevented by a rigid supportconcept, large swell<strong>in</strong>g pressures may developat the tunnel l<strong>in</strong><strong>in</strong>g. It is well known that swell<strong>in</strong>gdeformations – at least <strong>in</strong> claystone – reducewith the logarithm of stress, and that swell<strong>in</strong>gdeformations can be completely suppressed bysufficiently high pressure. The chemical processes<strong>in</strong> anhydrite swell<strong>in</strong>g, on the other hand, arecompletely different, and the semi-logarithmicrelationship between swell<strong>in</strong>g stra<strong>in</strong>s and stresslevel (Grob 1972) is not universally accepted forthese materials. Evolution of swell<strong>in</strong>g with time<strong>in</strong> both claystone and anhydrite depends onthe availability of water, which is governed bythe permeability of the material, layer<strong>in</strong>g of thesubsoil and the amount of water recharge. Assome of these factors relate to characteristics ofthe specific boundary value problem rather thanthe material itself, parameters determ<strong>in</strong><strong>in</strong>g thetime-swell behaviour cannot be transferred fromlaboratory tests to large-scale problems.Constitutive modelThe constitutive model used <strong>in</strong> this paper hasbeen implemented by T. Benz (NTNU Norway) asa user-def<strong>in</strong>ed soil model for PLAXIS. The modelemploys four parameters for strength and stiffnessand three parameters for swell<strong>in</strong>g. f’ and c’ arethe well-known Mohr-Coulomb friction angle andcohesion, E and n are the isotropic elastic Young’smodulus and Poisson’s ratio, respectively. Crossanisotropicelasticity can also be considered but isnot used <strong>in</strong> this study. The mean<strong>in</strong>g of the swell<strong>in</strong>gparameters is shown <strong>in</strong> Figure 1 and Figure 2. Themaximum swell<strong>in</strong>g pressure s q0is the axial stressbeyond which no swell<strong>in</strong>g occurs, the swell<strong>in</strong>gpotential k qgives the <strong>in</strong>cl<strong>in</strong>ation of the swell<strong>in</strong>gcurve <strong>in</strong> semi-logarithmic scale and the parameterh qis related to the time until the f<strong>in</strong>al swell<strong>in</strong>gstra<strong>in</strong> has developed (Wittke & Wittke 2005).Cross-anisotropic swell<strong>in</strong>g can be considered <strong>in</strong>the model, but aga<strong>in</strong> this feature is not used here.(1)If necessary the time-swell behaviour can be relatedto elastic and plastic volumetric stra<strong>in</strong>s, e velande vpl, by us<strong>in</strong>g parameters A eland A plto def<strong>in</strong>e thetime swell<strong>in</strong>g parameter h q:Positive volumetric stra<strong>in</strong>s (loosen<strong>in</strong>g of thematerial) result <strong>in</strong> faster approach of the f<strong>in</strong>alswell<strong>in</strong>g stra<strong>in</strong>, while negative volumetric stra<strong>in</strong>sdelay or may even stop the evolution of the swell<strong>in</strong>gstra<strong>in</strong>s. This approach accounts for the dependencyof the swell<strong>in</strong>g rate on the penetration rate of water.Construction of the PfändertunnelThe 6.7 km long first tube of the Pfaendertunnelnear Bregenz (Austria) was constructed <strong>in</strong> 1976-1980accord<strong>in</strong>g to the pr<strong>in</strong>ciples of the New AustrianTunnell<strong>in</strong>g method (NATM). While top head<strong>in</strong>gand bench excavation were carried out withoutmajor difficulties, significant <strong>in</strong>vert heave of upto 30 cm was observed after about 75% of thetunnel length was excavated. These observationslead to detailed laboratory <strong>in</strong>vestigations of theswell<strong>in</strong>g characteristics of the Pfaenderstockmaterial, an extensive monitor<strong>in</strong>g program and tothe <strong>in</strong>stallation of additional anchors <strong>in</strong> the tunnel<strong>in</strong>vert.(2)www.plaxis.nl l Spr<strong>in</strong>g issue 2013 l <strong>Plaxis</strong> Bullet<strong>in</strong> 5


Laboratory swell<strong>in</strong>g testsThe Pfaenderstock consists of various layers ofsandstone, conglomerate, claystone and marl,which are summarized as upper freshwatermolasse. The marl (claystone) layers wereidentified as the rock type caus<strong>in</strong>g the swell<strong>in</strong>gdue to their high content of Montmorillonite(Weiss et al. 1980). Czurda & G<strong>in</strong>ther (1983)dist<strong>in</strong>guished between undisturbed molasse marl(series A, Figure 4) and the fault zone material(series B, Figure 5). Series A samples showedhigher swell<strong>in</strong>g potential, but lower maximumswell<strong>in</strong>g pressures than the samples of seriesB. This notable difference was attributed torelaxation and swell<strong>in</strong>g of the series B samplesbefore the samples could be tested.For the back analysis two swell<strong>in</strong>g parametersets are considered, which represent the upperand lower boundary of the test results. The timeswell<strong>in</strong>g parameters A 0, A eland A plare calibratedto match the <strong>in</strong> situ time-swell<strong>in</strong>g curve.Numerical model and material parametersThe 2D f<strong>in</strong>ite element model used <strong>in</strong> this studyis shown <strong>in</strong> Figure 6. Tunnel geometry and basicmaterial parameters of the marl layer (E = 2.5 GPa,f’ = 34°, c’ = 1000 kPa) have been taken from Johnet al. (2009). Tunnel overburden is ~200 m abovethe tunnel crown, which is representative of thecross section at km 5+373. L<strong>in</strong>ear elastic plateelements are used for the shotcrete l<strong>in</strong><strong>in</strong>g, with E =7.5 GPa for the young and E = 15 GPa for the curedshotcrete. The f<strong>in</strong>al concrete l<strong>in</strong><strong>in</strong>g is modelledwith volume elements assum<strong>in</strong>g l<strong>in</strong>ear elasticbehaviour and a stiffness of E = 30 GPa. The f<strong>in</strong>all<strong>in</strong><strong>in</strong>g thickness varies between 50 cm at the <strong>in</strong>vertand 25 cm at the crown.<strong>Swell<strong>in</strong>g</strong> parameters are listed <strong>in</strong> Table 1. Sets 1a,1b and 2a only employ A 0for the time dependencyof swell<strong>in</strong>g, while <strong>in</strong> set 2b evolution of swell<strong>in</strong>gwith time is entirely governed by elastic volumetricstra<strong>in</strong>s.Fig. 1: Semi-logarithmic swell<strong>in</strong>g law (Grob 1972)Figure 2: Influence of h qon evolution of swell<strong>in</strong>g stra<strong>in</strong>sFigure 3: Pfaendertunnel cross section 1st tube (after John & Pilser 2011)Figure 4: <strong>Swell<strong>in</strong>g</strong> test results, series AFigure 5: <strong>Swell<strong>in</strong>g</strong> test results, series B6 <strong>Plaxis</strong> Bullet<strong>in</strong> l Spr<strong>in</strong>g issue 2013 l www.plaxis.nl


Figure 6: F<strong>in</strong>ite element model (dimensions <strong>in</strong> m)Figure 7: Development of <strong>in</strong>vert heave with timeFigure 8: Profile of vertical displacements, a) numerical analysis at t = 7180 d, b) measurements km5+820 (after John 1982)Figure 9: Development of pressure on the l<strong>in</strong><strong>in</strong>g (set 1b)Parameter Set 1a Set 1b Set 2a Set 2b<strong>Swell<strong>in</strong>g</strong> potential k q[%] 3.0 3.0 0.75 0.75Max. swell<strong>in</strong>g stress s q0[kPa] 1000 1500 4000 4000Table 1: <strong>Swell<strong>in</strong>g</strong> parametersA 05.0e -3 2.5e -3 3.0e -3 0.0A el0.0 0.0 0.0 9.0A pl0.0 0.0 0.0 0.0Calculation phasesAfter top head<strong>in</strong>g / <strong>in</strong>vert excavation (assum<strong>in</strong>gpre-relaxation factors of 75% and 37.5%,respectively), the concrete <strong>in</strong>vert arch is <strong>in</strong>stalled.<strong>Swell<strong>in</strong>g</strong> is conf<strong>in</strong>ed <strong>in</strong> the model to an area of 15m x 15 m below the tunnel <strong>in</strong>vert. After a swell<strong>in</strong>gphase of 65 days, the f<strong>in</strong>al l<strong>in</strong><strong>in</strong>g is activated,followed by another swell<strong>in</strong>g phase of 115 days.John (1982) reported that the decision on <strong>in</strong>vertanchor<strong>in</strong>g and pre-stress<strong>in</strong>g was based on theswell heave deformations observed up to thispo<strong>in</strong>t. In the cross section considered here thisresulted <strong>in</strong> an anchor pattern of 2.2 m spac<strong>in</strong>g.Evolution of <strong>in</strong>vert heave with timeFigure 9 compares the time-swell<strong>in</strong>g curvescalculated with the different parameter setswith the measured <strong>in</strong>vert heave <strong>in</strong> km 5+373.The measurements plot close to a straightl<strong>in</strong>e <strong>in</strong> logarithmic time scale, which cannot bereproduced exactly by the exponential approachemployed <strong>in</strong> the model. The match with themeasured <strong>in</strong>vert heave is, however, sufficient froma practical po<strong>in</strong>t of view.Set 1a delivers too little <strong>in</strong>vert heave (10mm), andthe development of deformations completelystops after activat<strong>in</strong>g the prestressed anchors.Increas<strong>in</strong>g the maximum swell<strong>in</strong>g stress by 50%(set 1b) yields ~50% more deformation and abetter match with the measurements. While such asignificant <strong>in</strong>fluence may be expected, it should benoted that experimental results for these two setsplot so close to each other that either of the twoparameter sets appears justified (Figure 4).Surpris<strong>in</strong>gly, sets 2a and 2b – which representmuch smaller free-swell deformations – delivermore <strong>in</strong>vert heave than sets 1a and 1b. This is aresult of the higher maximum swell<strong>in</strong>g stress <strong>in</strong>sets 2a and 2b, which activates swell<strong>in</strong>g <strong>in</strong> deeperrock layers, yet with a small swell<strong>in</strong>g potential.<strong>Swell<strong>in</strong>g</strong> deformations are thus more widelydistributed with set 2a and 2b.<strong>Modell<strong>in</strong>g</strong> the evolution of swell<strong>in</strong>g with timeentirely <strong>in</strong> dependence on elastic volumetricstra<strong>in</strong>s (set 2b) results <strong>in</strong> a slightly more prolongedtime-swell-curve than us<strong>in</strong>g a constant value of A 0(set 2a). In set 2b the rate of swell<strong>in</strong>g does not onlydecrease due to the convergence with the f<strong>in</strong>alswell<strong>in</strong>g stra<strong>in</strong>, but also due to negative elasticvolumetric stra<strong>in</strong>s. The large positive volumetricstra<strong>in</strong>s after tunnel excavation are graduallyreduced <strong>in</strong> the swell<strong>in</strong>g phases by the <strong>in</strong>creas<strong>in</strong>gswell<strong>in</strong>g pressure.www.plaxis.nl l Spr<strong>in</strong>g issue 2013 l <strong>Plaxis</strong> Bullet<strong>in</strong> 7


<strong>Modell<strong>in</strong>g</strong> <strong>Swell<strong>in</strong>g</strong> <strong>Rock</strong> <strong>Behaviour</strong> <strong>in</strong> Tunnell<strong>in</strong>gFigure 10: Variation of maximum swell<strong>in</strong>g stressFigure 11: Variation of time swell<strong>in</strong>g parameters (set 1b)Figure 12: Variation of rock stiffness (set 1b)Figure 13: Variation of stress pre-relaxationDistribution of swell<strong>in</strong>g stra<strong>in</strong>s over depthThe proportion of the rock mass which is affectedby swell<strong>in</strong>g depends primarily on the maximumswell<strong>in</strong>g stress. For set 1b (s q0= 1500 kPa) theswell<strong>in</strong>g zone is conf<strong>in</strong>ed to about 2 m below thetunnel <strong>in</strong>vert, which matches well with the slid<strong>in</strong>gmicrometer measurements <strong>in</strong> the neighbour<strong>in</strong>gcross section km 5+820 (Figure 8). The swell<strong>in</strong>gzone with set 2a (s q0= 4000 kPa) is much deeperdue to the higher maximum swell<strong>in</strong>g pressure,even though similar <strong>in</strong>vert heave is obta<strong>in</strong>ed withboth parameter sets. These results <strong>in</strong>dicate thatthe maximum <strong>in</strong> situ swell<strong>in</strong>g pressure is rather<strong>in</strong> the range of 1000-2000 kPa than close to the<strong>in</strong>-situ stresses.<strong>Swell<strong>in</strong>g</strong> pressureFigure 9 shows the distribution of swell<strong>in</strong>gpressure on the tunnel <strong>in</strong>vert l<strong>in</strong><strong>in</strong>g for differentstages <strong>in</strong> time for parameter set 1b. Thecircumferential distance L is measured from thetunnel <strong>in</strong>vert, such that L = 0 m is directly at the<strong>in</strong>vert and L = 5 m is the end of the swell<strong>in</strong>g area.No pressure measurements are available.Due to the stiffer support provided to the tunnell<strong>in</strong><strong>in</strong>g at the sides of the tunnel, the maximumswell<strong>in</strong>g pressure does not occur at the tunnel<strong>in</strong>vert but at a distance of ~3.8 m.Anchor prestress<strong>in</strong>g <strong>in</strong>creases the normal stresson the l<strong>in</strong><strong>in</strong>g by about 90 kPa. The difference tothe distributed prestress<strong>in</strong>g force of (0.8*640kN / 2.2 m / 2.2. m) = 106 kN/m 2 is a result of thealready closed f<strong>in</strong>al l<strong>in</strong><strong>in</strong>g, which distributes partof the applied load <strong>in</strong> circumferential direction.Compar<strong>in</strong>g the <strong>in</strong>crease <strong>in</strong> pressure to the swell<strong>in</strong>gl<strong>in</strong>e of set 1b at 200-300 kPa (Figure 4) expla<strong>in</strong>s thelimited <strong>in</strong>fluence of prestress<strong>in</strong>g <strong>in</strong> the numericalcalculations. Even though anchor prestress<strong>in</strong>g<strong>in</strong>creases the pressure by ~45%, reduction of f<strong>in</strong>alswell<strong>in</strong>g stra<strong>in</strong> is only about 18% due to the semilogarithmicswell<strong>in</strong>g law. Additionally, the effectof prestress<strong>in</strong>g dim<strong>in</strong>ishes rapidly with <strong>in</strong>creas<strong>in</strong>gdistance to the tunnel, and the deeper rock layersrema<strong>in</strong> virtually unaffected.Variation of maximum swell<strong>in</strong>g pressureThe calculated <strong>in</strong>vert heave is notably sensitiveto the maximum swell<strong>in</strong>g pressure s q0assumed<strong>in</strong> the numerical analysis. As the variation of thisparameter <strong>in</strong> the laboratory swell<strong>in</strong>g tests israther large– albeit concealed by the logarithmicstress scale – s q0has been varied from 500 kPato 2000 kPa (A 0= 2.5e -3 ). Results <strong>in</strong>dicate a l<strong>in</strong>ear<strong>in</strong>crease of <strong>in</strong>vert heave with s q0(Figure 10). Thisis primarily the result of the <strong>in</strong>creas<strong>in</strong>g depth ofthe swell<strong>in</strong>g zone below the tunnel <strong>in</strong>vert (Figure8), and not so much due to higher swell<strong>in</strong>g stra<strong>in</strong>sdirectly underneath the tunnel <strong>in</strong>vert.Influence of other material parametersThe <strong>in</strong>fluence of other material parameters onswell<strong>in</strong>g deformations is limited. As expected, thetime swell<strong>in</strong>g parameter A 0has a notable <strong>in</strong>fluenceon the evolution of swell<strong>in</strong>g deformations, butnot on f<strong>in</strong>al deformations (Figure 11). Vary<strong>in</strong>g theelastic rock stiffness had virtually no effect onswell<strong>in</strong>g deformations after tunnel excavation(Figure 12), but naturally changed deformationsdur<strong>in</strong>g tunnel excavation. Variation of the 2Dpre-relaxation factors – which <strong>in</strong> most practicalcases are an educated guess rather than athoroughly derived parameter – also had nonotably <strong>in</strong>fluence on swell<strong>in</strong>g deformation (Figure13). As no temporary <strong>in</strong>vert l<strong>in</strong><strong>in</strong>g was <strong>in</strong>stalledafter top head<strong>in</strong>g excavation, stresses <strong>in</strong> the rockmass at the tunnel <strong>in</strong>vert drop to ~0 dur<strong>in</strong>g tunnelexcavation, <strong>in</strong>dependent of the pre-relaxationfactors applied <strong>in</strong> the excavation phases.8 <strong>Plaxis</strong> Bullet<strong>in</strong> l Spr<strong>in</strong>g issue 2013 l www.plaxis.nl


Conclud<strong>in</strong>g remarksThis article presented the results of a backanalysis of measured swell<strong>in</strong>g deformations <strong>in</strong> thePfaendertunnel (Austria). A constitutive modelbased on Grob’s swell<strong>in</strong>g law and exponentialconvergence with f<strong>in</strong>al swell<strong>in</strong>g stra<strong>in</strong>s over timewas used for the numerical calculations. Inputswell<strong>in</strong>g parameters were derived from laboratoryswell<strong>in</strong>g tests. Due to the large variation oflaboratory test results, the sensitivity of modelpredictions on the <strong>in</strong>put swell<strong>in</strong>g parameters was<strong>in</strong>vestigated.Different sets of swell<strong>in</strong>g potential k qandmaximum swell<strong>in</strong>g stress s q0delivered verysimilar swell<strong>in</strong>g deformations at the tunnell<strong>in</strong><strong>in</strong>g, as <strong>in</strong>creas<strong>in</strong>g s q0is roughly equivalentto <strong>in</strong>creas<strong>in</strong>g k q. However, good match withthe measured displacement profile below thetunnel <strong>in</strong>vert was only obta<strong>in</strong>ed with s q0= 1500kPa, which represents the upper edge of theexperimental results on undisturbed molassemarl. Us<strong>in</strong>g higher values of s q0(and lower valuesof k q) delivers too large swell<strong>in</strong>g zones. The <strong>in</strong>vertheave measurements plot close to a straight l<strong>in</strong>e<strong>in</strong> logarithmic time scale, which cannot be exactlyreproduced by the exponential approach of theconstitutive model. The match with the measuredevolution of swell<strong>in</strong>g, however, is sufficient from apractical po<strong>in</strong>t of view.References• Czurda, K. A., and G<strong>in</strong>ther, G. (1983) “Quellverhaltender Molassemergel im Pfänderstock beiBregenz, Österreich“, Mitt. österr. geolog. Ges.,76, pp. 141-160.• Grob, H. (1972) “Schwelldruck im Belchentunnel“,Proc. Int. Symp. für Untertagebau, Luzern,pp. 99-119.• John, M. (1982) “Anwendung der neuen österreichischenTunnelbauweise bei quellendemGebirge im Pfändertunnel“ Proc. of the 31stGeomechanik Kolloquium, Salzburg, Austria.• John, M., Marcher, T., Pilser, G., and Alber, O.(2009) “Considerations of swell<strong>in</strong>g for the 2ndbore of the Pfändertunnel”, Proc. of the WorldTunnel Congress 2009, Budapest, Hungary, pp.50-61.• John, M., and Pilser, G. (2011) “Criteria forselect<strong>in</strong>g a tunnell<strong>in</strong>g method us<strong>in</strong>g the firstand the second tube of the Pfänder tunnel asexample”, Geomechanics and Tunnell<strong>in</strong>g, 4(11),pp. 527-533.• Weiss, E. H., Müller, H. M., Riedmüller, G.,and Schwaighofer, B. (1980) “Zum Problemquellfähiger Geste<strong>in</strong>e im Tunnelbau“, Geolog.Paläont. Mitt. Innsbruck, 10(5), pp. 207-210.• Wittke, W., and Wittke, M. (2005) “Design, constructionand supervision of tunnels <strong>in</strong> swell<strong>in</strong>grock”, Proc. 31st ITA World Tunnell<strong>in</strong>g Congress2005, pp. 1173-1178.www.plaxis.nl l Spr<strong>in</strong>g issue 2013 l <strong>Plaxis</strong> Bullet<strong>in</strong> 9

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!