ROCK MASS CHARACTERIZATION AND EVALUATION OF SUPPORTS FOR TUNNELS IN HIMALAYA

Abstract

Engineers in the recent times look towards empirical methods for selection and design of tunnel supports due to the complexities and uncertainties inherent in the analytical design methods. Engineering rock mass classification (RMC) is the best-known empirical approach for assessing the stability of underground openings. This approach has got enormous potential and forms the backbone of present day rock engineering. As a matter of fact, almost all the modern underground constructions are utilizing RMC approach due to its simplicity and practical approach. Out of several systems, Rock Structure Rating (RSR), Rock Mass Rating (RMR), and Rock Mass Quality (Q) systems present quantitative methods for describing the quality of a rock mass for selecting the appropriate ground support. Rock Mass index (RMi) system, recently proposed by Palmstrom (1995a) is also a quantitative method; applicability of which is yet to be established in Himalaya. The head race tunnel (HRT) of Nathpa Jhakri Project (NJP) is 27.4 km long with a diameter of 10.15 m. It is the longest power tunnel in the world being constructed in Higher Himalaya. M ~ -Excavation of this long tunnel has imposed many challenges to the field engineers. These include problems related to geothermic, heavy inflow of groundwater, excessive rock covers, flowing, slabbing and squeezing ground conditions. Rock types in the area encompassed by the project comprise a variety of metamorphic rocks like gneisses, schistose gneiss, schist, quartzite and basic intrusive (amphibolites), granite and pegmatite belonging to Jeori Wangtu Gneissic Complex of Precambrian age. In the present study, different RMC systems such as RSR, RMR, Q and RMi have been applied besides use of Rock Mass Number (N) and Rock Condition Rating (RCR). In all 685 tunnel sections covering a total length of 22159 m have been considered. A new interactive 11 computer program ROMAC has been developed in C++ language for classification using RSR, RMR, Q and RMi systems. Correlations have also been developed between different classification systems. The RSR concept may be viewed as an improvement of Terzaghi's method rather than an independent system. Wickham et al. (1972) described rock mass quantitatively in the form of RSR-value whereas Terzaghi (1946) described it qualitatively. The RSR uses parameter `A', which is estimated from the rock hardness and geological structure in addition to rock type origin. Wickham et al. (1972) have given qualitative terms for the hardness of rock such as hard, medium, soft and decomposed. In the present work, hardness has been related with uniaxial compressive strength (UCS) of intact rock. Deere and Miller (1966) classification has been utilized for the purpose. Wickham et al. (1974) have given recommendations for estimation of spacing of steel ribs as per FPS system of units for different tunnel diameters for datum condition. Moreover, these recommendations have been made for tunnel diameters up to 9 m only. Therefore, to facilitate use of steel ribs manufactured in India as per Indian Standards, a chart has been developed to get the spacing of various steel ribs for datum condition, beyond 9 m also. On the basis of recommendations given by Bieniawski (1989), equations have been developed to interpolate support details for a particular RMR-value. The support recommendations given by Bieniawski (1989) simply state light, medium and heavy ribs. It does not give formulation for estimation of pressure. It has been found from the present study that on transition from `Fair' to `Poor' rock class, there is a huge jump in terms of support capacity due to the introduction of steel ribs for `Poor' rock class. The recommendations given by Barton et al. (1974) do not suggest stress reduction factor (SRF) values for `competent rock, rock stress problems' category for moderately jointed rocks in 111 categories L, M and N. In the present work, SRF-values selected are 9, 15 and 20 respectively for these. But in all these sections rocks are moderately jointed (2 joint sets or more) not massive. Consequently, Q-values might be erroneous. Therefore, new SRF-values have been proposed in moderately jointed rocks and a correlation has also been developed for the same. RMi uses a number of parameters and computations involved in its estimation are complex. In estimation of RMi, volume of block (Vb) plays a very significant role. According to Palmstrom (1995), RMi is a volumetric parameter indicating the approximate UCS of a rock mass. A comparison between UCS of rock mass (qcmass) and RMi indicates that RMi-values are far less than gcmass (Singh et al., 1992) and no correlation could be established between the two. Reason for this may be that Vb is very low for the rock mass considered in the present study. RMi consists of UCS of intact rock (q~) and jointing parameter (JP), which is dependent on Vb and joint condition factor (jC). In jC roughness and alteration are almost similar to those considered by Barton (1974). Therefore, Vb is the only parameter, which might have influenced estimation of RMi. It means that RMi is very sensitive to Vb in jointed rock masses or in other words its proper estimation is heavily dependent on the correct evaluation of Vb. Estimation as per Unal (1983) gives higher pressures in majority of sections compared to RCR, Q and N but lesser pressures compared to RSR in non-squeezing ground conditions. For squeezing ground conditions, these are far lower than the capacity of installed support. Earlier also, Goel and Jethwa (1991) reported that the estimated support pressures as per Unal (1983) were unsafe under squeezing ground conditions. Pressures estimated from RMR as per Goel and Jethwa (1991) and from RCR as per Goel (1994) for non-squeezing ground conditions are slightly on lower side as compared to N. Therefore modified correlations have been proposed. Further, it has been found that pressure estimated from iv Q has been less than N in squeezing ground conditions. An excellent correlation is obtained between Q and N, once all the corrections are applied in Q as suggested by Singh et al. (1992). Palmstrom (2000a) proposed charts for the design of support in blocky and continuous grounds separately, without any, correlation for estimation of roof pressure. In the present work, a correlation has been developed for estimation of roof pressure for blocky ground. A general trend with regard to pressure estimation may be given as follows: Support Pressure (Non-Squeezing): PRCR <PQSPN <PRMR<PRSR Support Pressure (Squeezing]: PRMR < PRCR < PQ < PN It is difficult to compare four classification systems from support point of view since different types of supports have been recommended by them. Singh et al. (1995b) have proposed a semi - empirical method for the design of supports for tunnels and caverns. From the present study, it has been found that semi - empirical method is not applicable beyond 0.25 MPa pressure. A new correlation has been developed which may render original method applicable for higher pressures as well. Support charts for estimating capacity of shotcrete/SFRS and rock bolt systems have also been developed from the database contained in this thesis. Capacity of support recommended by RSR and that installed by the Project Authorities match only in about 10 percent sections whereas in about 77 percent sections, capacity of support installed by the Project Authorities is less than that recommended by RSR. This indicates that in non-squeezing ground conditions RSR overestimates support requirements. Further it has been found that capacity of support recommended by RMR-system has been the least in both non-squeezing and squeezing ground conditions. v It has also been found that capacity of support recommended by the Q-system is less compared to that installed by the Project Authorities and interestingly capacity of support installed by the Project Authorities is less compared to pressure estimated from N. Also it has been found that capacity of support recommended by the RMi-system has been highest of all. The capacity of support is so high that in about 82 percent sections, support installed by the Project Authorities are of lower capacity than that recommended by RMi in non-squeezing ground conditions. On the other hand in squeezing ground conditions, in about 57 percent sections, the Project Authorities have installed lesser supports and in the rest about 43 percent sections greater supports than those recommended by RMi. A general trend with regard to support capacity may be given as follows: Support Capacity (Non-Squeezing): CRMR < CQ < CRSR < < CRMi Support Capacity S ueezin : CRMR < CQ < CRMi The Project Authorities have classified and designated the rock masses using the Q-system. Later, following the same designation they tried to choose supports using the RMR-system. From the study of tunnel sections in squeezing ground conditions, it has been found that majority of sections are unstable, since pressures after corrections as per Singh et al. (1992) are more than the support capacity. At one of the locations bending/twisting of ribs has been observed due to the fact that pressure after corrections is more than the support capacity notwithstanding the fact that pressure before corrections is less than the support capacity. It is feared that other sections might also show signs of instability in future where pressures exceeded capacities of supports installed. A general trend with regard to position of supports installed by the Project Authorities may be given as follows: Support Capacity (Non-Squeezing): CRMR < PRCR < CPROJECT < PN = CQ < CRSR < CRMi Support Capacity (Squeezing): CRMR < PRCR < CQ < CPROJECT < PN < CRMi vi Rock mass — support interaction analysis is a well-known method of design of tunnel support system. It is considered as one of the most promising methods for understanding the mechanics of tunnel deformation and development of rock loads. The interactive computer programs have been developed in C++ language for determination of: e Ground Response Curve (GRC) as per the formulations of Ladanyi (1974) and Brown et al. (1983). Support Reaction Curve (SRC) for shotcrete/rock bolt and steel rib/concrete supports. From the study it has been found that: (i) GRC is very sensitive towards RMR or Geological Strength Index (GSI) values, (ii) In `Good rock', all the classification systems recommended supports whereas through the analysis it has been found that hardly any support is needed, as deformations are very small, .(iii) In `Fair rock', supports recommended by RMR are usually inadequate whereas those by RMi and Q should be installed after some initial deformation, (iv) In `Poor rock' only steel ribs, as recommended by RSR and RMR systems are adequate for stability whereas supports recommended by Q are inadequate, (v) In all the sections considered, rock mass behaviour has been found to be elasto-plastic in non-squeezing ground conditions also, where it should have been elastic, (vi) GRC using N lies far above GRC using RCR in all the sections indicating that RCR underestimates pressures in squeezing ground conditions and (vii) Supports recommended by RMR are adequate whereas those recommended by Q prove inadequate in squeezing ground conditions. In the present work, it has been found from the rock mass - support interaction analysis that for `Good' and `Fair' rocks (using RMR-system), observed and predicted deformations are in close agreement. Deformation modulii have been computed from the correlation of Verman et al. (1993). vii Therefore it may be concluded that deformation modulii estimated from RMR for `Good' and `Fair' rocks appear to be accurate. According to Singh et al. (1998), Mohr's theory is not applicable for anisotropic and jointed rock masses on account of significant strength enhancement due to in-situ stress along tunnels. In the present work, an attempt has been made to evaluate the effect of strength enhancement in rock mass and then compare it with Mohr's criterion. The tunnel sections chosen for this purpose lie under a rock cover of more than 1000 m with a maximum value of 1430 m. Rock mass contains two or more joint sets implying that this study pertains to jointed rock mass only. The study shows that heavy rock burst predicted by Mohr's criterion is moderated to a great extent if Singh's criterion is adopted. During the excavation of these tunnel reaches also, it has been observed that behaviour of rocks actually follows Singh's criterion. At none of the sections, there were heavy rock burst since rock mass is jointed and also due to the enhancement of strength in rock masses, there have been no stress related problems, except slabbing with cracking noise. Prediction of ground condition prior to the actual excavation can be extremely useful. In the present work therefore, correlations have been developed for prediction of ground conditions based on: (i) tangential stress (ae)/q'cmass, (ii) Q and N and (iii) joint roughness number (Jr) and joint alteration number (Ja). From the overall analysis of the classification performed in the present study, it has been found that revised Q-system is the best.