Rock mass classification for choosing between TBM and drill- and-blast or a hybrid solution N. Barton Barcon& Associates, Ola, So Paulo ABSTRACT: The speeds of TBM tunnelling and drilland-blast tunnelling are compared, using the new Qrox model for TBM performance estimation, andthe conventional Q-value for dl nd-blast prognoses. By using these two methods it ean be estimated whether a hybrid solution might be the most economic and timely. For instance one would dill and-bast the most prob Jematic ground, if early access was feasible, while waiting for TBM delivery. hybrid solution was used at the 18 km long Qinling Tunnel in China, and is also planned in Braz, where sbea- sive, massive rocks occur at both ends ofthe tunnel. Logging methods that ean conveniently be used to describe the ground, including the use of seismic, ae described inthis paper, together ‘with some of the details ofthe Qra method, nluding a worked example | INTRODUCTION ‘The pressing need for fst tunnelling solutions for infastucture development has naturally fo- ‘cussed attention on TBM tunneling. In hydropower development an even more obvious need for TBM tunnelling is apparent, due to the potentially favourable smooth profile obtained i the rock mass has favourable properties. ‘Wester countries noted with interest the recent introduction of two large TBM into China for «planned 27-month completion ofthe 185 km long Qinling rail tunnel. Te hard granites and ‘very hard gneiss reportedly gave best penetration rates of about 4 m/hr but slowed to only 03 ruin the hardest gneiss. Besides the reportedly massive rock, an overburden as high a3 1600 1, and averaging 1000 m, probably played its par in slowing the machines. Utilisation was less (an 30% in ¢ 24-hour day on average, and cutter wear was significant (Wallis, 2000), ‘A political decision to dril-and-biast the central section ofthe tunnel to bring forward com- pletion deadlines, while the two TBM completed 5.3 and 5.6 km from the N and S portals cor venienly focuses attention on our subject "Choosing between TBM and érilland-bias", A hy- ‘brid solution combining the benefs of both methods of tunneling should always be carefully assessed beforehand, and compared tothe single solutions of one (or two) TEM, or dall-anc- biasing alone. How best to make this assessment? Recently, feasibility studies fora 16 km water tunnel in Brazil have focussed on deliberately uilising the hybrid solution, due fo the presence of massive, abrasive granites and sandstones st Fespective ends ofthe tunnel and "idea! TBM rock" (phylies and schists) inthe central half of| the tunnel, The method used for this assessment was the newly develeped Qray, concept (Bar ton, 1998; 2000), which gives detailed prognoses based on rock and rock mise characterisation slong planaed tunnels.© [Gee reon | erroon [verrroon | ronan frocovecod] one E =| F 2 — 5 7 Ped 8 s el 3 3 peel 3 os GA)—221_ sus. B 5 po pratt | é oot ce} ai aaa value Fig 1. Conceptual relation between TBM performance and Q-value. 2 QANDQray InFigure tthe relative magnitudes of TBM penetration rates and atual advance rates are draw in relation tothe Q-value obtained ftom rock mass clasification (Berton eal, 1974). Ina later figure we will aso see how speeds for drill-and-blast tunnelling are also related withthe Qe value but in significantly different way. Inthe case of TBM, massive rock is unfevourable for fast penetration, while for diland-blat itis obviously favourable due to the lack of tune! support needs, and canbe drilled at reasonable speed despite the lack of jointing ‘The new Qray method is built onto the six Q-system parameters that ace now widely used around the world. However, there are differences, including the use of RQD, — an oriented QD relevant to the tunnelling direction. Other Qrny parameters shown in Figure 2 rete tothe ratio ofrock mass strength SIGMA and cuter force F, the cuter life index, the quarts conten! of the rock and te estimated stress level atthe tunnel face, Each of theve addtional parameters are normalised by atypical value, therefore giving higher or lower values of Qranw The most impor ‘ant single parameter by far is cutter free. More details ae given later 3 THE LAW OF DECELERATION ‘A very important factofife for TBM is that there are diferent advance rate curves for eich time period (i. 24 hours, 1 week, | month, et), These are shown in Figute 2, ‘An extensive review of 145 TBM tunnels, totalling more than 1000 km in length (Barton, 2000) has shown that there isa consistent deceleration or decline in the average advance rate ‘with time. Ths is of course known, but hardly quantified or discussed inthe hterature, where ‘est weeks and best months make more impressive reading. ‘The general trends ofthese numerous cases are drawn in Figure 3, where WR (world resord), good", “us, "poor" and “excremely poor” lines of performance are draw. Te actual caver: cords af plotied in Figure 4 ‘The declining advance with log time can be quantified bythe gradient (m), which has units of deceleration (LT). The importance of this quantification is that the utilisation (U) that links Penetration rate (PR) and advance rate (AR): ‘AR= UxPR @ rust be quantified as a time dependent variable, which is a very necessary step for corect prognoses, We therefore wate: ust @ ‘where T isin hours, and m is always negative. In many typical eases, when neglecting major fault zones: ust ®RELATIVE DIFFICULTY OF GROUND FOR TEM USE aes jrsesssue| endttine | ran [coco] e002 leoeo] vam | rove a | e,, ‘ SE unstaste: 47 Beoe : ZT estan eovte “2” sf So i $1 ovecoroaks orp _ a 2] problems, cutter head ra 7 fen tore ot 3] tesa 2 xh y dey SMA 20 a 3 * see * F%a0* * cu * 20 Fig. 2. The Qraw method that builds directly onto the O-system. Barton, 1999) oye aoueapy Penetration rate Fig. 3. Thedecelerarng rae of advance in TAM tunneling. (Barton, 1999)ee — eee) © syqut 9724 eoueapy Fig. ¢. Case records; 145 tunnel totalling more than 1000 km (Barton, 2000) ‘When we read from the Qinling Tunnel (Wallis, 2000) that “utilisation in a 24 hour day ran at less than 30% on average this implies an average gradient (m) given by ma 280 ® iogT 7 AtQinling we therefore have: (90523 ns OF (9038 ‘This immediately signifies an unfavourably seep gradient of deceleration in Figure 3, presums- bly due to the hard, abrasive and massive nature of the rock reported at Qinling, and areas ef high stres; infact probably far rom ideal for TBM tunnelling ‘The great majority of average TBM performances lie between lies (1) and (3) in Figure 3, ‘which makes alot of TBM tunnelling very suovessfil. However, there are periods with “unex. pected events" (i. low Q-values in fault zones) that may seriously delay average performance, For entirely different reasons, hard, abrasive, massive conditions and fault zones with clay and ‘maybe water, give unfavourably steep gradients of deceleration, Each are quantified in specifi, but different ways, inthe Qran approach, 4 COMPARING SPEEDS OF TBM AND DRILL-AND-BLAST Favourable geology and hydrogeology for the TBM tunnelling option produce dazzling results, \which are plysically difficult o visualise. or example, 150 m in day, $00 m ina week, 2k in a month and 12 kam in a year (or even better results) have been achieved in some well Dublicised projects. However, these are not typical, nor are they average performances (except Dethaps the 12 km ina year) ‘Before committing to a significant investment, anda significant delivery and assembly time, itis wise to carefully compare TBM and crill-and-blat options, and perhaps find that a hybrid Solution is more idea. If lot of faulted ground is present, the TEM option should probably be avoided, to help maintain the excelent record of THM successes, Altemately one may be able to drilland-blast the most faulted (or most massive) part ofthe tunnel and use the TBM forthe more appropriate conditions, when it arrives tthe ste sometime le.RE Gi Rock mass quaity = (SR * Sie) Fig. 5. Q-statisties for a planned TBM tunnel, (Laset, 1999) ‘The Q-value statistic along a planned tunnel route (e-. Figure 5), which may also be used for Nick Berton and Asocictes eck Engineering, : = LH%n LOL60—co1.4s Fee ereoe EE Seg) Eo POTD Q (typical range) = @ (mean) = ¥{———] («#3 f Bsos DSS 752s wEHDDSTE TTS Fig.9 — Qulogging of a section of horizontal drill care, ahead of a slowing TBM. (NBA, 2000)QQ. —> Rockmase quay Seismic velocty Deformation modus eI VrkgOrsstinte —=100% (Gra _er030"F!_(o>0) creme | Ext. | Exc. egy | wy [row [re | one [to] fh [Bt eo sent « Jeo 6 S40 40 i feo ao g2o 20 19 serstyn oot ar + 2 10) 4 100 «400 1000 tox 3 SH) Fig. 10 A method to convert V0 a Q-value, with due regard forthe compression strength “and porosity ofthe rock material. andthe depth ofthe seismic measurement. (Bar- ton, 1995) {USING SEISMIC REFRACTION DATA IN PROGNOSES, ‘Unfortunately the practice of long horizontal eoce drilling isnot yt widespread, either before projet start or during the tunnelling, There will therefore seldom be such detail a forthe 200 m ‘bf ground logged in Figure 9. (Thi was part ofa much longer pilot hole.) Much mare frequent willbe vertial or inclined boreholes, often limited to both ends ofthe tunnel, due to overburden Constraints or perhaps due to a sub-sc location. Tt then becomes more important to utilise and interpret shallow reftaction seismic measure ments, and eventual cros-hole seismic velocities (mean values or tomographic) that may (ot Should) be evallable across major fault zones, Conversion to approximate Q-vaues, using the Q. hormalisation method shown in Figure 10, bas proved to give useful estimates of quality along the tunnel of interest. (TBM or dril-and-blast) The continuous seismie velocity log is form of extrapolation (actualy intrapoation) berween boreholes, and can be calibrated by the intermit tent coreslogging. The intertlationships shown in Figure 10 have been obtained by just such calibrations (Baron, 1998). none recent TBM tunnelling problem, with difficult sub-ea conditions and extensive delays die to faulted ground and pre-injetion needs, the use of the Vj-Qe-Q conversion, and estima- tion of Qroe (Figure 2) proved far more accurate than “conventional” methods of TBM progno- Sh which vastly overpredicted the actual progress. The tunne! concemed has significant Jenaths of low Q-value which seriously undermine the expectation of faster TBM tunnelling than drilland-biast tunnelling. In Figure 6 nomenclature, we are too far to the let, and the tunnelling is now taking too long.9 PRESTIMATED FROM Qa, ‘The methods used to predict penetration rate (PR) with uninterrupted boring, and average ad- vance rate (AR) fr different periods of tunnelling will now be briefly summarised. More details ae given by Barton, 2000. The basic empirical equation for estimating Qrew is given in Figure 2 and iss follows: rey a{ Bex Je, te), SIGMA, 20. 4 oe ry SRF ate F*/20" "cui" 20" 5 © In the parentheses is an oriented Q-value, Q based on oriented ROD (ROD, in the tunneling dicecton) and 1/1 relevant to the ease (or dificult) of boring with the most optimally oriented Joint set, Rough, discontinuous joints hinder progress (PR) while smooth continuous joints help PR but hinder AR if support needs increase. (We take care ofthe later by suitable choice of gradient), SIGMA ~a measure of the rock mass strength, allows anisotropic behaviour or large ratios of ‘ily (due to cleavage oF foliation) to influence the penewation rate. SIMA in is simplest form is given| I SIGMA, =57Q.? MPa © where 1 = rock density (gmies), Q.= Qu x Fe, (Cutter force F (tnt) is ealeulated from the net thrustcutter, ie. minus the cylinder pressure needed to pull the back-up equipment. It has an approximately quadratic effect on PR (Le. changing F from 20 to 30 tnffeuter may increase PR by a factor of about 9/4), However, be- ‘cause Fis compared to SIGMA, the Qrox model also allows PR to reduce with increased cutter force (as it may in very hard rocks, e.g. Grandori et al, 1995). The ease shown in Figure 11 ‘from Nelson et al, 1983, is graphie example of this possibilty, especially when cuter loading is limited, ‘The term CLI (cutter life index) in equation is an empirical term developed by the Univer- sity of Trondheim (NTH, 1994). It is obtained from the combined use of # miniature drilling test and a cutter steel abrasion (weightloss) tet. Typical values of CLI might be $ for quartzite, 15 for gneiss, 30 for phyllites, 80 for shale and 100 for limestone. However, quite wide ranges ate seen, due partly to quartz content (q) which isa separate term in the equation for Qreu (qin ‘%). Finally, we have a rough estimate of the maximum biaxial stress component on the tunnel face, normalised by 5 MPa, an assumed value for 100 m depth ‘AS will be noted in Figure 2, Qrau allows one to estimate the average penetration rate PR, or ‘to back-caleulate it from the average performance ofthis parameter. PReSQmc? oO Ff Qray =(5/PR)? o Where PR is in muh In Figure 2, ony the “central” range of Qrew is shown, With very low cutter fore in very hard rock, Oran becomes a very large number, wale in fault zones, twill become too low and the ‘operitor must reduce thrust to allow time fr sablisation. The table below gives feel forthe sort of magnitudes involved. ‘Table 1. Estimates of Qray for lines WR, 1, 2,3. and 4in Fi Penh 7 WR (world reo 10 005 Line 1 good") 5 1 Tine 2 (air) 3 3 Line 3 poor”) 2 98 Line 4 (ext. poor” 1 312s Exceptional case o1 312,500,000mamirev. LEGEND, = 01630% 2 ato som, a2 3 eo cos * e040 100% PENETRATION RATE Limestone 125 AVERAGE THRUST PER CUTTER Fig. 11 Peneration rate versus thrust per eutter for specified percentages of limestone. (Nelson etal, 1983) 10 ARESTIMATED FROM PR ‘As we have seen earlier when discussing gradients of deceleration, the simple relation between PRand AR: AR=PR«U isto simple, ests theme prod (ds pek, month ete) neds to be speci since ‘A constant gradient (ie, m= -0.20) a3 seen in central areas of Figure 3 (midway between lines 2 and 3) would give uilisation percentages tat fall, naturally, with inereasing time periods asin Table 2. ‘Table 2. Typical PR, U and AR dats for one year of TBM tunnelling (when m = 0.20) PR Ishi Téay week [month 3 months Tyear Ihe Shs 246 168s 720/hrs_2160 hrs_8760 brs U= lows 6% 53% 36% 27% 22% 16% AR. 30206 0806 See ‘This isthe reason why specifying utilisation as “532” must be qualified bythe relevant time pe- ‘od. fn the sbove ease the mean utlization per 24-hour day. ‘The table below shows how gradient (m) is estimated, As seen in both the table and in Figure 12, there isa strong dependence onthe conventional Q-value when stability is poor (Q = 0.1, (0.61 and 01001), but most dependence on abrasion factors when the rock is stable and perhaps “Table 3, Gradient m and uiisaton a function of Q-valus for average I-month periods. Qvale O00r OOO 10 10 100 1000 m 99 87 85 92201719021 Ung 003 610s 0s 038029028410 a8 as 2s 3a 23 Geant ectig vane fn) Ooimesooe Domo Of aa4 t aa4 1 wow M00 meswer 1000 SiGe) Fig. 12 Preliminary eximate of declining advance rato gradient (-Jm, as a function of O- value, Rock mass quality @ Usilisation percentages are enormously dependent on the relevant range of Q-values, especially ‘when longer time periods are involved. Table 3 tells us that when penetrating a fauit zone that takes | month (720 brs) only 1% of the time would be spendin advancing the tunnel ifthe Q- value was as low as 0.01, ‘Because abrasion terms ate so important when the rock is of “good” quality for stability (Le Q~ 100 or ever 1000) empirically derived “fine tuning” has been necessary in estimating the final gradient m. The following equation is used: ieee 5) (ea) (3) 3) ® hc D = et ann es psy er i et). nada eles ya mle en boas eka scene Seagate Fc eet enema fo Te Fg 1 myo ont tans na trans tnt aed ea ae uso ia pon w sina ne se (A) fen nd emia Tae AR=PRxU UH PR=5Qru > which give: AR®5Q rm"? xT™ ° 1 WORKED EXAMPLE (On the final pages of this keynote lecture, a worked example is show. Iisa hypothetical ex- ample but is based on areal ease) and consists of equal lengths (4 km) of massive abrasive sand- stone, then phyllites, then mica schists and finally massive, high strength granite, Readers wil ‘ote from the estimation of time (T) 1 advance through each geological domain, where Lyre (4) « tee fr nsh ne aden a i 0 T= 90 hes Schechterend ofthe tunnel while waiting for TBM delivery for the central 8 km of “easy to bore” phyl- lites and mica schist, Thus the best solution i the hybrid solution, as discussed in reference 12 the Qinling Tunnel atthe beginning ofthis article ‘Without such a solution, using only one TEM, the accumulated time for each of the four goo- logical zones would be $3,393 hours (an unacceptable 6 years), but actualy longer than this ‘sum (27) dve to the accumulated deceleration. The final AR = 0.28 mir ie seen in section J of the worked example, ‘In practical application of this Qraus method, after hand-calculated examples such as the “coarse” lam zoning given in Table 4, it is belpfl to use computerized calculation. A plot of PR, AR(end) and T for each domain inthe log-log format shown in Figure 3 i included inthe Excel program Qrau developed by my colleague Ricardo Abrabio in Sto Paulo. This can be ‘obisined through warw.Qibm.com. ‘Table . Worked example, Input datasheet & Calculation sheet AA) STABILITY (and gradient m,) (ee ETO] o-(*°), Teonssones [oir | woos" | aos a BPaaies [eT sa esr tar te reese bar SOe [oT ae (eesttnveraie tr sa) FF Grarips [0.18 | Wooe | Bt Oss Lar 3B) ORIENTED Qo (n tunneling direction) Zone | & [ROD | at SAF] Ga] = (2224). (4) (Segoe are “eps apa CR SG) Se) peehies ae se | z Hi (most affecting utters) | [2Mlea senSie|o0_[ a5 [40] oer ’ neusem) [eGranies —[70@0| 00s [2 | oer (© ROCK MASS STRENGTH (SIGMA) SICMAgy = 570? Zone [1 | & [Oe] Sar] fo | O [S| Se Se ee oon) -0-8) SE ee ee P) Or SIGMA 20 Oram = Qo Sage * CLI" 20". (20 w0.00s4 winastr cute) SHAT 2 zoe & [Ge] oo | | [ita | om | Sanam sos oe zPyaies | ~g-[ 14 |-25-[-20 aos] Steen ene [3] Wo [ as eae Grantee 20-2} 30] ze. 2) GRADIENT M (ae Ta a os 7 sanananas or} oss boar Pryiies [A ],2, 8 Tso Shea sents [0.207197 298] Beanies aa] FoePENETRATION RATE aoe : = ce PR= 5(Qren)™ [2-Phylites 08 O77 en ee Pose | — ae os Gye 10 apvanceeNotmt i TAR] Peapens: ~~ 7 ‘S736 nee . Samaras 00 oor “2snye (ebay [sue a0 a te Sua scuas| aioe [ae Tae er sot ane bane fase Tease are ELT) ET 5280 fs) SE.00 7) Frans AR arc foro oe our) 1 OVERALL PERFORMANCE FReegneonen. srs [| or (Pera o}_|_ sy | vie | mite Cohee | PERERA ET 0 | ra FR (ABLeARaLee) [Presence see} -sea- 6000 | i000 (Usietde) gene toon feos] 350 | ose Tee fe) Eamon a tn amatEND oF Prostct me | me meme | Pee oa [zrrats | cae | a0 | om | 02s The ideal taneling predicted inthe pylites and schists cleat indicates the great bene- Sof TBM taneling. In this example the massive hrdo-bore sandstones aa gases occur at either end ofthe tunel nd coud be dril-and- based while waking for TBM driven, 12 CONCLUSIONS ‘This brief comparison ofthe two principal tunnelling methods used in rock, has drawn attention to some potential eros in our profession, and the need for careful planning forthe bes altea- In rock tunneling its clear that costly and time-consuming errors can be made by the wrong choice of method (TBM instead of crill-and-bast or vice versa). We have seen that sometimes the hybrid solution ean be the best choice ffom the start. While waiting for TBM delivery, ene cr both porals can be driven, ora central section opened in one or both diretions, as recentiy done a the Oinling Tunnel in China, when TBM progress was less than expected Ina worked example we have seen that estimation of TBM progress rates through different geo logical zones may highlight the need for a hybrid solution. Both deilland-blast and TBM tun- nelling wil benefit from such decisions; the excellent reputation of both can be preserved, and ‘there wil be less claims and conflict‘The key to such decisions is a good pre-investigation (geological mapping, seismic refraction, core drilling, rock mass characterisations, some lab testing) Longitudinal logs of Q, Q, and an estimate of Qraw (depends on assumed eatter force are requited o make the estimates of rla- tive tunneling rates (TEM or drill-and-blast as reliable a possible, The consequences of erors of judgement are extremely costly to the tunelling industy, as evidenced by too many wele publicised cases. 13 REFERENCES Barton, N, Lien, R.& Lunde, J, 1974. Engineering classification of rock masses forthe design of tunnel support. Rock Mechanics. Vo. 6:4: 189-236, Springer Verlag Barton, N. & Grimstad, E. 1994. The Q-System following twenty years of application in NMT suppor: selection. 43rd Geomechanie Colloquy, Salzburg, Felsbau, 194; 428-436. Barta, N. 1995. The influence of joint properties in modelling jointed rock masses, Keynote lecture, 8* SRM Congress, Tokyo. Val. 3 Barton, N, 1999. TBM performance estimation in rock using Qray- Tunnels ané Tunnelling In- temational. September: 30-34. London. Barton, N. 1998. Rock mass charectrization from seismic messurements. Keynote lecture. Proc. SAROCKS "98, 2" Brazilian Symposium, 5* South American Rock Mechanics Con- agrees, Santos, Brazil. L. Da Silva et al (eds). Vol. 2 Barton, N. 2000, TBM tunnelling in jointed and faulted rock. 173 p. Rotterdam: Balkema Grimstaé, E, & Barton, N. 1993. Updating the Q-system for NMT. In Kormpen Opsah & Berg (ds) Proc. ofthe Intemational Symposium on Sprayed Concrete - Modern use of wet mix prayed concrete for underground support. Fageres, Oslo: Norwegian Concrete Asso~ ation, Grimstad, E. 1999. Private communication, Lose, F. 1999, Private communication. NBA, 200. Contact report. Nelson, P,, O'Rourke, T.D. & Kulhawy, F-H. 1983. Factors affecting TBM penetration rats in sedimentary rocks. In Mathewson (ed), 24" US symposium on rock mechanics, Texas ‘A&M Univ. 227-237. Colloge Station, TX: Assocation of Enginecring Geologists NGL, 1998. Contact report ‘Wallis, S. 2000. A Chinese experience. Tunnel 1/2000, 18-25,