TBM performance estimation using rock mass classifications (PDF ...

International Journal of Rock Mechanics & Mining Sciences 39 (2002) 771–788

TBM performance estimation using rock mass classifications M. Sapignia, M. Bertib,*, E. Bethazc, A. Busillod, G. Cardonee b

a Enelpower S.p.A., Via Torino 16, 30172 Venezia-Mestre, Italy Dipartimento di Scienze della Terra e Geologico-Ambientali, Universita" di Bologna, Via Zambonii 67, 40126 Bologna, Italy c Enelpower S.p.A., Ciso Regina Margherita 267, 10143 Torino, Italy d SELI S.p.A., Viale America 93, 00144, Roma, Italy e SOGIN S.p.A., Via Torino 6, 00184, Italy

Accepted 1 June 2002

Abstract Three tunnels for hydraulic purposes were excavated by tunnel-boring machines (TBM) in mostly hard metamorphic rocks in Northern Italy. A total of 14 km of tunnel was surveyed almost continually, yielding over 700 sets of data featuring rock mass characteristics and TBM performance. The empirical relations between rock mass rating and penetration rate clearly show that TBM performance reaches a maximum in the rock mass rating (RMR) range 40–70 while slower penetration is experienced in both too bad and too good rock masses. However, as different rocks gives different penetrations for the same RMR, the use of Bieniawski’s classification for predictive purpose is only possible provided one uses a normalized RMR index with reference to the basic factors affecting TBM tunneling. Comparison of actual penetrations with those predicted by the Innaurato and Barton models shows poor agreement, thus highlighting the difficulties involved in TBM performance prediction. r 2002 Elsevier Science Ltd. All rights reserved.

1. Introduction Since James S. Robbins built his tunnel-boring machine (TBM) in 1954, the TBM designs have improved greatly, in an effort to tackle ever-wider ranges of rock conditions at higher advance rates. Today’s TBMs can reach extremes of 1000 m/month [1] but advance rates of less than 50 m/month may be experienced in adverse geologic conditions or when support measures fail to maintain tunnel stability until the final lining [2]. A reliable estimation of excavation rates is needed for time planning, cost control and choice of excavation method in order to make tunnel boring economic in comparison with the classical drill and blasting method. As a consequence, great efforts have been made to correlate TBM performance with rock mass and machine parameters, either through empirical approach or physically based theories [3–7].

*Corresponding author. Tel.: +39-051-209-4546; fax: +39-051-20945-22. E-mail address: [email protected] (M. Berti).

Performance prediction of TBM drives requires the estimation of both penetration rate (PR) and advance rate (AR). Penetration rate is defined as the distance excavated divided by the operating time during a continuous excavation phase, while advance rate is the actual distance mined and supported divided by the total time and it includes downtimes for TBM maintenance, machine breakdown, and tunnel failure [8]. Even in stable rock, the rate of advance AR is considerably lower than the net rate of penetration PR; and utilization coefficients (U ¼ AR=PR) in the order of 30–50% have been reported by many authors mainly due to TBM daily maintenance [9–11]. In lowquality rock, the penetration rate can be potentially very high but the support needs, rock jams and gripper bearing failure result in slow advance rate, with utilization coefficients as low as 5–10% or less [2]. Simple performance correlations have been developed from data on conventional rock strength testing at the laboratory scale. These equations relate the penetration rate with intact rock parameters like the uniaxial compressive strength [12,13], the rock tensile strength [14] or the rock fracture toughness [15], showing good predictive ability in the case of homogenous

1365-1609/02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved. PII: S 1 3 6 5 - 1 6 0 9 ( 0 2 ) 0 0 0 6 9 - 2

772

M. Sapigni et al. / International Journal of Rock Mechanics & Mining Sciences 39 (2002) 771–788

low-fractured rocks. Belonging to these is the predictive model proposed by the Colorado School of Mines [16], in which TBM penetration and utilization are computed by means of a force equilibrium approach on the basis of cutter geometry and uniaxial and tensile strength of intact rock. In jointed rocks the presence of discontinuities reduces the rock mass strength increasing the rate of penetration for a given TBM thrust [17–19]. Predictive equations should be based on rock mass properties rather than intact rock strength, for example, relating TBM performance with rock mass strength derived by standard geomechanical classifications [20–24]. Barton [23,24] made the most progress in this direction. He proposed an expanded version of his well-known Q-system [25] in which additional rock– machine–rock mass interaction parameters were introduced in order to take into account both the rock conditions and the reaction of TBM to the conditions. QTBM allows one to estimate TBM penetration and advance rate in a wide range of rock conditions even if, as pointed out by the same author, improvements and corrections are possible by testing new case records. As far as we know, less attention has been paid to the correlation between TBM performance and Rock Mass Rating [26], despite the wide use of this geomechanical classification in daily practice [10,27–29]. The basic features of the correlation with rock mass rating (RMR) are presented in this paper, referring to three tunnels excavated in the Italian Alps in medium to hard metamorphic rocks. Fourteen kilometer of TBM tunnels were classified and analyzed, yielding over 700 sets of data featuring rock mass quality, TBM penetration, thrust and utilization coefficient.

2. Case studies 2.1. Sites characteristics Data for TBM-performance analysis have been obtained from three tunnels excavated in metamorphic rocks for hydraulic purposes. The three tunnels (Fig. 1) are located in the northwestern Alps (Italy) and consist of one inclined tunnel for the installation of a penstock (Maen) and two horizontal diversion tunnels (Pieve and Varzo). Descriptive information on the tunnel projects and tunneling equipment are summarized in Table 1 while Table 2 reports the main strength and drillability parameters determined through laboratory tests on intact rock samples. 2.1.1. Maen The area rock units consist of meta-ophiolites (serpentinite, metagabbro, metabasite, chlorite schist, talc schist) and meta-sediments (calc schist and silicate

marble) belonging to the Zermatt-Saas Zone of the Pennidic Domain [30,31]. The parent rocks were carbonate pelagic sequences and mafic crystalline rocks that underwent high-pressure low-temperature metamorphism during the early phases of the alpine orogenesis. Schists and serpentinite show a foliated texture while metagabbro and metabasite are generally weakly foliated. The attitude of rock units is more or less uniform throughout the tunnel, at N220–2701E/ 35–451 (dip direction/dip), so that the longitudinal axis of the inclined tunnel (plunging direction N1281E) is almost normal to the schistosity. A major shear zone, 20 m in thickness, is encountered within the tunnel. It is composed of massive blocks of serpentinite and metagabbro (0.5–1.5 m3) embedded in a sheared matrix of talc and chlorite schists associated with cataclastic bands. Even if the fault zone was clearly recognized by the geological investigations, as soon as the excavation reached the adverse stretch, massive blocks jammed the TBM cutterhead. In the attempt to move back the TBM, a large face and roof collapse occurred involving an estimated volume of 150–200 m3 of loosened rocks. The accident caused 4 months stoppage over the 14 months total construction time and it required an extensive grouting of the failed mass to be undertaken [32,33]. Dataset for performance analysis consists of 330 records featuring TBM parameters (head thrust, net boring time, total boring time) and rock mass classification indexes (RMR and Q). The open-type TBM allowed continuous surveying of the rock mass all over the tunnel length: RMR and Q were independently logged by surveying adjacent tunnel sections 5 m in length; penetration rate and advance rate were computed dividing the length of the surveyed section (5 m) by the net boring time and the total boring time, respectively. 2.1.2. Pieve vergonte Most of the Pieve Vergonte tunnel is located in the Sesia-Lanzo Zone of the Austroalpine Domain [34–36]. Excavated rocks consist of two metamorphic complexes made up of gneiss and micaschists separated by a metadiorite intrusive body with minor masses of metaquartzdiorite and metagabbro. The first upstream reach (1.5 km) crosses the metagranite belonging to the Pennidic Domain (M. Rosa tectonic unit) and, for a short reach approximately 100 m in length, chlorite and amphibole schists which separate the Austroalpine from the Pennidic Domain. Micaschists, chlorite schists and amphibole schists are characterized by a foliated texture, gneiss and metamorphic rocks of the intrusive complex are non-foliated or weakly foliated. The geological structure is complicated by multiple folding associated with shear zones and brittle fault zones, but the general attitude of rock units forms a


Fig. 1. Geological sections along the three tunnels.

773


774

monocline dipping at N140–1801E/30–601 (dip direction/ dip), so that the longitudinal axis of the tunnel (direction N070–050E) is mainly parallel to the schistosity. Due to the continuous segmental lining, rock mass survey was possible only during the daily maintenance of the boring machine, accessing the excavation face beyond the TBM cutterhead. We then had to assume that the rock mass surveyed in the short reach

Table 1 Summary description of tunnel projects and tunneling equipment

Total tunnel length (m) Total excavation time (days) Surveyed section length (m) Excavated diameter (m) Tunnel slope (1) TBM model TBM type Number of cutters Cutter spacing (mm) Cutter diameter (in) Maximum thrust (kN) Boring stroke (m) Cutterhead curvature Cutterhead rotation rate (rpm)

Maen

Pieve

Varzo

1750 413

9600 809

6600 468

1750

6400

5800

4.20 24–35 Wirth 340/ 420 E Open

4.05 D0 Robbins 1111–234–3 Double shield 27 75 1700 4602 0.63 Flat 11.3

4.05 D0 Robbins 1214–240/1 Double shield 27 75 1700 8827 0.63 Flat 4.5–8.9

36 66 1700 7920 1.5 Domed 5.5–11

between the rock face and the cutterhead (1 m) was representative of the whole section bored over a working day (17 m on average); a rather hard assumption that it was finally accepted, given the homogeneity of the rock mass and the high surveying frequency. The dataset consists of 301 daily records describing rock mass quality, mean head thrust, net boring time, and excavated length for the first 6.4 km of the tunnel. RMR was logged in all the surveyed sections, Q in only 44 sections regularly spaced along the tunnel axis (15% of the dataset). Penetration rate and advance rate were computed by dividing the daily excavated length by the net boring time and the total boring time (24 h), respectively. 2.1.3. Varzo The Varzo tunnel is excavated entirely in the Antigorio Gneiss Formation, a massive or weakly foliated rock generated by high-grade metamorphism of granite and granodiorite rocks [37,38]. Metaaplite and metabasite dikes locally traverse the tunnel axis, but the area may be considered essentially homogenous. The geological structure is a monocline gently dipping (10–201) in a southerly direction, slightly complicated by folds and minor fault zones related to the SempioneCentovalli fault, a major tectonic structure located 2 km to the south [39]. In general, the schistosity follows the attitude of the overall structure and, is therefore, mainly parallel to the longitudinal axis of the tunnel (plunging direction N080E–N070E).

Table 2 Main characteristics of excavated rocks Tunnel

Rock type

Uniaxial compressive strength (MPa)

Tensile strength (MPa)

Hardness Indenter (u.c.)

Knoop hardness (GPa)

Drillability (mm1)

Tangent Young’s modulus (GPa)

Maen

Serpentinite

124 (64–174) 180 (104–289) 17

—

—

—

—

—

15 (9–29) —

26 (13–40) —

6.2 (4.3–8.3) —

0.04–0.10 —

65 (37–94) —

10–12

13

5.1

—

39

—

—

—

—

—

124–215 171–221 146–296

5–9 8–13 0.7–7

7.5–9.7 11 7.1–7.4

5.2–8.5 6.2–7.0 7–10

0.11–0.22 0.03–0.05 0.06–0.09

28 46–100 24–38

161 (90–260) 115 (82–217)

16 (9–24) 17 (7–25)

3.7 (2.2–4.8) 3.8 (2.5–3.3)

—

—

9 (6–13)

—

—

Metabasite Chlorite schist

(0.9–39) 138 (113–163) 75 (29–134)

Metagabbro Calc schist

Pieve

Micaschist Metadiorite Metagranite

Varzo

Gneiss

> Schist. //Schist.


Also, in this latter case, the use of double shield TBM with segmental lining prevented a continuous surveying of excavated rocks. Geomechanical classification was then performed during the maintenance downtimes (almost every day), and the surveyed quality was extended to the whole section bored in that day as described for Varzo. Resulting dataset consists of 103 daily records featuring rock mass quality (RMR=all sections; Q ¼ 16 sectionsE16% of the dataset), mean head thrust, net boring time, and daily excavated length (15 m on average).

775

quency distributions are negatively skewed (relatively fewer frequencies at low RMR values) with most of the values falling in the good-quality classes (I and II RMR classes). Low quality reaches (IV–V RMR class) are related to fault zones, composed of highly fractured rocks, softened chlorite and talc schists and groundwater dripping from major planes. RMR is well correlated with Q (Fig. 3) and the experimental distribution follows very closely the correlation line proposed by Bieniawski [26] for tunnels.

2.2. Rock mass classification 3. Empirical relationships Most of the excavated rock masses exhibited good strength and a relatively low degree of fracturing. Rarely more than three discontinuity sets were encountered, and usually only two were found at any location, typically characterized by planar, smooth and tight, unweathered or slightly weathered joint walls. The general good quality of the rock masses is evident by the frequency distributions of rock mass rating depicted in Fig. 2. RMR values are based on the 1989 version of the classification [26] taking into account the adjustment factor for discontinuity orientation. Fre-

3.1. Penetration rate 3.1.1. Testing the regression model Typical relation between PR and RMR is depicted in Fig. 4. As can be seen the scatter is rather wide, leading to uncertainties about which regression model, for example quadratic or linear, is appropriate to fit experimental data. Published works typically show that empirical relations seem to follow a bell-shaped curve more than a linear trend, with maximum performance

Fig. 2. Frequency distributions of the excavated rocks in the three tunnels.

776


Fig. 3. Correlation between RMR and Q values logged in the three tunnels. Dotted lines include 80% of the 111 case histories analyzed by Bieniawski [26].

Fig. 4. A statistical analysis of variance was performed in order to attain a significant regression models of performance data. Example refers to Maen tunnel.

for medium-quality rock masses and lower penetration for poor and very hard rock masses [3,8,10,40]. In order to attain a significant regression model of performance data, a statistical analysis of variance was performed [41]. The analysis consists of a set of three F tests aimed to verify: (i) the significance of the linear fit; (ii) the significance of the quadratic fit; (iii) the significance of increase of quadratic over linear fit. If the computed F value for each of the three tests falls in the critical region, that is if it exceeds the critical

value of F (Fcrit ) at the selected level of significance (for example, a ¼ 0:05; see [41]), we conclude that our model is correct. On the other hand, if F oFcrit we must accept the null hypothesis stating that the variance about the regression is no different to the variance in the observations, and conclude that our model is not correct. In the example of Fig. 4, all the three tests give F > Fcrit ; so we can state that: (i) the linear regression is significant; (ii) the quadratic regression is also significant; (iii) the quadratic term is making a significant


contribution to the regression model and should be retained. This means that the quadratic equation fits performance data more closely than a straight line does. If performance data were linearly distributed with RMR; we would obtain a significant regression (F > Fcrit ) both for the linear and the quadratic model (a quadratic equation may also fit a linear distribution), but the third test on the contribution of the quadratic model over the linear one would have given negative response (F oFcrit ). It was then suggested that we adopt the linear regression model. In the right chart of Fig. 4, data have been grouped in 10 RMR classes and plotted as bar charts, the central point of each bar indicating the mean and the length two times the standard deviation of the values falling in each class. This simple averaging technique allows the trend to be seen more clearly, and it will be used throughout the paper to enhance charts readability. However, statistical analyses and correlation coefficients will always refer to unaveraged values. 3.1.2. Empirical relations for different rocks The analysis of variance has been performed for the predominant rock types encountered in the three tunnels and both for RMR and Q classification methods. Fig. 5 summarizes the results obtained for RMR system. In general, the penetration rate increases with decreasing rock mass quality until RMR values of about 50–70. The performance drop below that range reflects bad boreability in adverse rock masses, where mucking problems and face instability reduce the potentially high penetration rate. On the contrary, low PR recorded in very good rock masses (RMR>80–90) depend on the high strength of the intact rock and by the low discontinuity frequency in the rock mass, which reduce the ability of roller cutter indentation and chips formation by a fracture mechanism. An approximate quadratic trend also characterizes the correlation between penetration rate and Q (Maen only) on a logarithmic scale, with maximum performance in the range Q ¼ 5215 and slower penetration for both higher and lower Q-values. In most cases, the curvilinear regression model fits performance data better than the linear one, with the only exceptions of mostly bad (Chlorite and Talc Schists—Maen Tunnel) or good rocks (Metadiorite— Pieve Tunnel) characterized by a range of RMR values too narrow to depict the whole curvilinear trend. The more or less quadratic relation between penetration rate and RMR is seen despite the steady linear increase of TBM thrust with rock mass strength (Fig. 6), indicating that the observed trend does not imitate the applied force but that it is the result of the TBM–rock mass interaction. A similar trend of decreasing penetration with increasing thrust has been observed by Grandori et al. [22] for Hong Kong granites, in which the available

777

thrust per cutter was insufficient because of the very strong rock. 3.1.3. Average trend Performance data for all the excavated rocks in the three tunnels are summarized in Fig. 7 (upper) as a function of Rock Mass Rating. Once again, a quadratic relation between PR and RMR is suggested, both for single tunnels and the cumulative dataset. The correlations are significant from the statistical point of view, and almost identical results have been obtained correlating the penetration rate with the basic RMR index, that is RMR unadjusted for discontinuity orientations [26]. However, the high dispersion of recorded data should be noted (shaded area in Fig. 7). Although some of the scatter is obviously due to the cumulative analysis of different rocks excavated by different machines, we believe the dispersion is an intrinsic feature of penetration data, and that it mostly arises from the difficulty in maintaining a constant thrust. In fact, similar scatter may be also seen considering individual rock types (Fig. 5) or normalizing the penetration rate according to the net thrust per cutter and rpm of a specific TBM machine (Fig. 7 lower). Relevance of data scatter to performance prediction will be discussed in Section 5. As regards, the applicability of our results to other TBM projects, correlations depicted in Fig. 7 are probably significant in terms of shape (best performance in medium-quality rocks) but not for numerical prediction. The RMR-system, in fact, does not account for rock–machine interaction parameters, so any empirical relation based on this system is inevitably limited to the rock–machine combinations considered in the original dataset. 3.2. Utilization coefficient The fraction of total construction time that the TBM has been utilized for boring (utilization coefficient, U) is given by the ratio of AR and PR: As pointed out by Barton [24] the advance rate declines with time following a rather uniform logarithmic trend, so that declining utilization is seen as the unit of time (day, week, month) increase (see also [11]). The trend is described by the equation U ¼ T m ; where T is expressed in hours and the negative gradient m is a function of rock and machine parameters (see Section 4.2), and it indicates the increasing likelihood that unfavorable extreme conditions (both exceptionally poor and exceptionally good) are encountered as tunnel length progresses. In our case, TBM utilization has been derived from daily data (T ¼ 24) and mean values for the three tunnels are depicted in Fig. 8 as a function of Rock Mass Rating. The three lines show that even in

778


Fig. 5. Relationships between RMR and penetration rate for the predominant rock types encountered in the three tunnels. The small table in the corner of each plot summarizes the results of the analysis of variance: F > Fcrit states that the model is correct (the null hypothesis must be accepted (A); F oFcrit states that the model is not correct (the null hypothesis must be rejected (R).

favorable conditions the utilization coefficient is less than 55% and that values as low as 5–10% may be experienced in bad conditions. The corresponding mean advance rate ranges from 0.7 to 1.0 m/h in good rocks and from 0.2 to 0.3 m/h in highly jointed faulted rocks.

These values are well in the range of published daily utilizations [9,10,40], although the average gradients m back-calculated in the three cases (Maen=0.43; Pieve=0.30; Varzo=0.33) are lower than the typical gradient m ¼ 0:20 indicated by Barton. That is to say


Fig. 6. Mean TBM thrust linearly increase with Rock Mass Rating for individual rocks (Maen tunnel).

779

Fig. 8. Utilization coefficient derived from daily average data.

cutter wear from abrasive rocks at Pieve and Varzo, and to the steeply inclined excavation at Maen.

4. Comparison with existing predictive models Actual penetrations may be compared with those predicted by the empirical equations proposed by Innaurato et al. [3,21] and Barton [23,24], which relate TBM performance with rock classification indexes. Purpose of the comparison is to test the predictive capabilities of these models when detailed data, closely surveyed at the excavation face, are available. To some extent we are dealing with ideal conditions, so we expect good predictions. 4.1. Penetration rate 4.1.1. The RSR model Innaurato et al. [3,21] found a strong correlation between PR; rock structure rating (RSR) [42] and uniaxial compressive strength (UCS) of intact rock: PR ¼ 40:41UCS0:44 þ 0:047RSR þ 3:15;

Fig. 7. Relation between TBM penetration and Rock Mass Rating. Excavated rocks include serpentinite, metabasite, chlorite schist, talc schist, calc schist, metagabbro, mica schist, metadiorite, metagranite, and gneiss, involving a total length of about 14 km.

that we experienced slower performance compared to a typical project. Rather low-utilization coefficients might be due to the non-optimal cutter spacing and the severe

ð1Þ

where PR is in mm/round and UCS in MPa. For a given rock with constant UCS the relation predicts penetration as a linear function of RSR; faster boring being expected in low-quality rock masses. The database used by Innaurato consists of five tunnels (total lengthD19 km) excavated in igneous, sedimentary, and metamorphic rocks with average UCS in the range 50–150 MPa. The RSR is related to RMR by the following [26]: RSR ¼ 0:77RMR þ 12:4

ð2Þ

780


which has been used by Innaurato to derive RSR when not available. Using Eq. (1) and the mean UCS values listed in Table 2, the theoretical penetrations for Maen, Pieve and Varzo have been computed and compared with the recorded ones. The comparison is shown in Fig. 9 in terms of difference between simulated and measured penetrations (DPR) as a function of RMR: As can be seen, predicted penetrations are consistently higher than the measured ones, the difference increasing in poor rock where the mean error rises up to 100% of the real value. It is likely that the poor agreement is due to the absence of any TBM-related factor in the predictive model, which limits the applicability of Eq. (1) to rock– machine combinations similar to those considered in the original database. This is especially true in poor rock, where the TBM–rock mass interaction is of paramount importance. 4.1.2. The QTBM model The method recently proposed by Barton [23] is based on an expanded Q-system of rock mass classification, in which the average cutter force, abrasive nature of the rock, and rock stress level is accounted for. The new parameter QTBM is a function of 20 basic parameters, many of which can be simply estimated by an experienced engineering geologist: QTBM ¼

RQD0 Jr Jw SIGMA 20 q sy ; Jn Ja SFR F 10 =209 CLI 20 5

ð3Þ

where RQD0 is the conventional RQD interpreted in the tunneling direction; Jn ; Jw ; and SFR are unchanged from conventional Q; Jr and Ja are also unchanged but they should refer to the joint set that most assists (or hinders) boring; SIGMA is the rock mass strength (MPa); F is

the average cutter load (tnf); CLI is the cutter life index; q is the quartz content (on percentage); sy is the average biaxial stress on tunnel face (MPa). From the analysis of numerous projects (145 cases), Barton derived a simple relationship between penetration rate and QTBM : PR ¼ 5ðQTBM Þ0:2

ð4Þ

which predicts a power increase of penetration with decreasing of QTBM : As clearly stated by the author (see [24], pp. 73 and 99), the relation gives meaningful results only for QTBM > 1; as in very poor rock masses the operator would usually reduce the penetration rate due to the bad rock conditions. At the time of the construction of the tunnels (from early 1998 to middle 2000) we were not aware of the new method developed by Barton, therefore geomechanical data were collected according to conventional RMR and Q systems (see Section 2). The problem behind a late-inthe-project QTBM analysis is that the new term QTBM has additional rock–machine–rock mass interaction parameters that should be explicitly evaluated for TBM tunneling, while conventional classification procedures are focused on tunnel stability and support measurements. In our case, however, at least for Maen tunnel in which conventional Q-values were continuously logged, the available dataset seems adequate for a posteriori evaluation of QTBM : This belief is supported as follows: * *

*

Jn ; Jw ; SFR are unchanged from conventional Q: RQD0 coincides with conventional RQD, since scanlines for spacing measurements were oriented along tunnel alignment. Jr and Ja ; are essentially unknown, but the error related to the use of conventional joint factors can be estimated.

Fig. 9. Difference between recorded and computed penetration rate as a function of RMR. Predictions are based on the empirical equations proposed by Innaurato et al. [21] and Barton [23].


*

* *

In principle, the use of conventional Jr and Ja values is a potential source of large error in a late-inthe-project QTBM analysis. Logged values for conventional Q; in fact, refer to the joint set that most influence tunnel stability, which is usually the set whose strike is parallel to the tunnel axis, while QTBM draws the attention to the joint set that most influence boring, which is typically a dominant jointing or anisotropic structure parallel to the tunnel face [24,43]. Based on our geomechanical surveys, the worst scenario we could have faced in Maen is that, at a given tunnel section, the difference in Jr =Ja ratio between the joint set critical for stability (logged) and that critical for boring (required by QTBM ) was very high, let us say Jr =Ja ¼ 5 for the first and Jr =Ja ¼ 0:13 for the latter. This unfavorable combination would have caused QTBM to be modified by a factor up to 40. Reanalyzing the original data sheets we have estimated that such a large error should not affect more than 10% of the dataset, while it should range from 0 to 20 in a further 20%, and it is almost negligible in the remaining 70%, both because only one set or a dominant set were present (55%) and because the rock mass was so highly fractured that average logged values were suitable both for boring and stability analysis (15%). SIGMA was estimated on the basis of Q0 (the conventional Q with oriented RQD0 ) and rock density as proposed by Barton [24] (Table 3). F was continuously recorded during excavation. CLI values were defined with reference to the typical values published by NTH for 12 different rock types [18]. Obviously, the NTH table does not deal with a great variety of rocks texture and composition, so the choice of appropriate values was sometimes ambig-

*

781

uous. To overcome this problem and in order to reduce subjectivity, an estimate of CLI was supported by petrographic analyses and laboratory tests performed on numerous rock samples collected at the tunnel face during excavation. In particular, mean Mohs’ hardness and rock abrasivity were useful for this purpose. The first was estimated by determining the proportional of each mineral in the rock and then multiplying the hardness value assigned to that mineral by the Mohs’ scale [44]; the latter from the relation between mean Mohs’ hardness and steel point abrasiveness test value [44]. Estimated hardness and abrasivity values are listed in Table 3 with corresponding CLI: As can be seen, the maximum uncertainty range of CLI is about 40 (serpentinite and calc schist), which might cause QTBM to be modified by a factor of 2. Relevance of this uncertainty to QTBM predictions has been investigated with a sensitivity analysis. The quartz content q was obtained from petrographic analysis. Values are less than 20–30% for most of the excavated rocks, as they result from metamorphism of igneous and sedimentary rocks with low quartz content. However, severe cutters wear was observed in garnet-rich rocks (metabasite) and in rocks containing more than 60–70% amphiboles and olivine (metagabbro), suggesting that an equivalent quartz content would be more suitable for our purposes. Three different values of q were then considered for each lithotype: (i) the ‘‘true’’ quartz content; (ii) the equivalent quartz content, computed on the basis of the quartz-equivalence of the rock-forming minerals [45]; (iii) the percentage of minerals with Mohs’ hardness grade higher than 7, which is the nominal hardness of quartz. Computed values are summarized

Table 3 Relevant parameters for QTBM analysis. Italic values are those giving the best agreement between recorded and computed penetrations from sensitivity analyses. (1)–(3) refer to the three methods for estimating the quartz-content described in the text Tunnel

Rock type

SIGMA (MPa)

Mean Mohs’ hardness

Abrasiveness (1/10 mm)

CLI

q (%) (1)

(2)

(3)

Maen

Serpentinite Metabasite Talc and chlorite schists Metagabbro Calc schist

41716 72731 874 75727 42712

3.6 6.2 2.8 6.0 3.6

1.9 5.0 1.0 4.8 1.9

30–70 10–20 60–90 15–25 30–70

5 8 5 5 20

28 63 23 56 37

5 26 5 5 20

Pieve

Micaschist Metadiorite Meta quartzdiorite Metagranite and metaaplite

50718 65723 68724 56724

4.1 5.1 6.4 6.6

2.5 3.7 5.2 5.5

15–70 15–40 15 10

30 5 15 40

51 53 80 85

30 5 15 40

Varzo

Gneiss

48726

5.8

4.5

15–25

40

75

40

782


in Table 3 and have been used as inputs for a sensitivity analysis. The comments above apply to Pieve and Varzo as well, but with three important differences: (i) conventional Q was not logged except in a few sections (see Section 2); (ii) RQD0 was derived from joints spacing measurements at the tunnel face; (iii) due to continuous segmental lining, the rock mass is known with less accuracy than in Maen. Of these three, the first is the most important limitation, since we should derive Q (Q0 ) from RMR (RMR0 ) [26] in order to compute QTBM ; and although RMR and Q are well correlated (Fig. 3) the procedure is questionable and results cannot be used for testing model capabilities. However, QTBM was computed in the cases of Pieve and Varzo as well, with the purpose of evaluating model response when the available dataset is neither comprehensive nor tailored for performance analysis. It is logical to expect the model would perform worse than in the case of Maen. A first series of sensitivity analyses was done on CLI; q and Jr =Ja : The results showed that, over the selected ranges, QTBM is only slightly influenced by CLI and q; while it is very sensitive to Jr =Ja changes. We then decided to use single values for CLI and q (chosen to obtain best predictions; see Table 3) and error bars on the graphs to help capture the uncertainty in Jr =Ja ratio. The difference between predicted and measured penetrations at Maen is plotted in Fig. 9 as a function of RMR: Unlike the Innaurato model, QTBM apparently gives good results, the difference in penetration rate varying around zero on the average. However, when we compare actual and theoretical penetrations as a function of QTBM (Fig. 10 upper) the apparent good match disappears into statistical noise: measured points elongate over an almost horizontal axis, indicating low sensitivity of QTBM : We can explain this different outcome looking at the term SIGMA=ðF 10 =209 Þ of Eq. (3). Following Barton [24], this ratio should allow QTBM to predict PR in poor rocks, expressing the possibility of reduced penetration (high QTBM values) with decreased rock mass strength (SIGMA) if cutter force (F ) decreases more consistently. From this point of view, the ratio performed well in our case: much higher values were obtained in poor rocks (up to 105) than in hard rocks (102 and lower), with a progressive decrease of the ratio for increasing QTBM : However, an unwelcome reduction in QTBM sensitivity was observed, as it is evident by plotting mean Q and QTBM values as a function of RMR (Fig. 11). The slope of the QTBM 2RMR correlation line, in fact, is remarkably higher than the slope of the conventional relation between Q and RMR [26], with the result that a wide range of our RMR values (10oRMRo70) falls into a narrow range of QTBM indexes (100oQTBM o700). In

this narrow range, the theoretical curve in Fig. 10 cuts the experimental distribution close to its mean axis, which is why the model seems to predict the mean PR in Fig. 9 well. It may be tempting to explain these unsatisfactory results with the uncertainties inherent in our late-in-theproject analysis, but we must remember that the error is probably negligible at least for 70% of the Maen dataset (single points in Fig. 10); we probably could not do much better even logging QTBM during tunnel excavation. On the other hand, the new Barton model is based on data from 145 TBM projects and its reliability cannot be judged by an individual case. A short discussion on this point will be given later in the paper. 4.2. Advance rate The QTBM -system also allows the estimate of advance rate (AR) as follows [24]: AR ¼ PR T m ;

ð5Þ

where T is the time in hours and m is a negative gradient which express the decelerating average advance rate as the unit of time increase. The gradient m is a function of cutter life index (CLI), quartz content (q), porosity of the rock (n), tunnel diameter (D) and of a parameter (m1 ) tabulated as a function of Q [24]: 20 0:15 q 0:10 n0:05 : ð6Þ m ¼ m1 CLI 20 2 From his case record analysis, Barton obtained a typical value m ¼ 0:20 and an approximate ranges from 0.15 to 0.45, the least negative value referring to good rock conditions. In the case of Maen, the mean value is m ¼ 20:17 and 95% of the computed gradients fall between 0.30 and 0.10; similar results have been obtained for Pieve and Varzo, the mean gradients being –0.18 and –0.22, respectively. As expected given the data in Fig. 10 for PR; the correlation between QTBM and daily AR (T ¼ 24 h) is unsatisfactory as well, the experimental points spreading parallel to the abscissa without a significant trend (Fig. 12). Moreover, the majority of experimental points fall below the two theoretical curves computed using the extreme values of the gradient m; thus the predicted advance rate is somehow overestimated. However, it is very probable that our data are not suitable for this comparison because of the non-optimal design of the machines (Pieve and Varzo) and the steeply inclined excavation (Maen) already mentioned for explaining the low-utilization coefficients (Section 3.2). 4.3. Specific penetration Alber [28] proposed an interesting correlation between uniaxial rock mass strength, derived from RMR


783

Fig. 10. Comparison of recorded penetrations in the three tunnels (Maen, upper; Pieve and Varzo, lower) with predictive equation proposed by Barton [23]. Classes indicate relative difficulty of ground for TBM use.

using the Hoek-Brown failure criterion, and the specific penetration SP, which is more suitable than the penetration rate for comparing different TBM projects. The correlation is based on the analysis of 55 km TBM tunneling involving five different TBMs (1700 disc size) and may be used for a probabilistic estimate of project economics. Unfortunately, the comparison of recorded and predicted penetrations is rather unsatisfactory, actual data falling below the correlation line of the 10% percentile (Fig. 1, [28]). The presence of high abrasive rocks and the non-optimal cutter spacing may

possibly explain lower penetration velocities experienced in our cases.

5. Discussion As previously described, empirical relations between mean penetration rate and rock mass rating clearly reveals the strong dependence of TBM performance on rock type (Fig. 5). Even considering the same TBM machine and the same RMR class, lower penetration

784


Fig. 11. Relationship between Q; QTBM and RMR for Maen tunnel.

Fig. 12. Comparison of advance rate in the three tunnels with predictive equation proposed by Barton [23].

rates are experienced in stronger rocks, as shown, for example, by the comparison of the two predominant rock types encountered in Maen and Pieve tunnels (Fig. 13). Reductions in mean penetration rate are seen despite the increased thrusts that were utilized for stronger rocks, suggesting that rock-related factors (joint spacing, tensile strength, joint or fabric orientation) may dominate the mechanism of rock crushing and chip formation in hard rock. Based on this simple observation we can conclude that the conventional RMR system is inadequate for TBM performance prediction, which is not surprising if we consider that rock mass rating, like most of the

geomechanical classifications used in daily practice, has been developed to provide support guidelines for underground openings excavated with drill-and-blast method. A logical development would be to define a normalized RMR index with reference to the basic factors affecting penetration rate, for example, uniaxial compressive strength, tensile strength, brittleness, abrasion, or rock hardness, that is factors controlling rock resistance to cutter penetration and fracture propagation: ideally, different rocks would depict a unique curve on a PR-normalized RMR plot. Our data do not allow us to define a suitable normalization factor but some indications can be given.


785

Fig. 13. Different penetrations are experienced in different rocks for the same RMR. Examples refer to the two predominant rock types in Maen and Pieve tunnel.

Fig. 14. The variation of mean penetration rate (DPR) is much better correlated with mean Mohs’ hardness than with uniaxial compressive strength of the intact rock (UCS). The ten data points plotted in each chart derive from the one-by-one comparison of the five rock types encountered in Maen tunnel (serpentinite, metabasite, chlorite and talc schists, calc schist, metagabbro).

Fig. 14 compares the difference of mean penetration rates recorded for predominant rock types in Maen at the same RMR value (RMR ¼ 60) with the difference in UCS and mean Mohs’ hardness (Section 4.1). Interestingly, the variation of mean penetration rate is much better correlated with mean Mohs’ hardness (r ¼ 0:81) than with UCS (r ¼ 0:16), and in the former case the regression line passes close to zero indicating that two rocks with same mean Mohs’ hardness should ideally give the same penetration rate (for the same RMR). Similar results have been obtained for RMR values in the range 40–90 and by normalizing UCS and Mohs’ hardness with reference to the mean TBM thrust, F : Mohs’ hardness scale, however, is neither linear, nor do the minerals selected provide a uniform scale of

hardness increase when the minerals are evaluated using modern hardness testing instruments, so Mohs’ hardness is not really the ideal candidate for RMR normalization. Beside the most logical choice of using some measurable drillability parameter, for example, the Drilling Rate Index [46], the Rock Drillability Index [47], or the Stamp Test [48], also quantitative measure of rock texture describing grain shape, orientation, interlocking and relative proportions with matrix (e.g. the Texture Coefficient proposed by Howarth and Rowlands [49,50]) are worthy of attention. As stated by Sanio [43] these parameters can be linked to the fracture propagation mechanism caused by the TBM rolling cutters, which is strongly dependent on rock fabric orientation. Numerous rock samples collected during

786


tunnels excavations are freely available to everyone wishing to start a collaborative research on this topic. The last point of discussion concerns the large scatter showed by recorded penetrations. As said above, we believe this scatter mostly depends on the difficulty in maintaining a constant thrust during excavation, which causes the net penetration to vary up to 50% of the mean values for the same rock type and the same RMR value (Fig. 4, 5 and 7; see also [51]). In Maen, for example, the 50 m tunnel section from 1+100 to 1+150 m is characterized by a low-fractured, homogenous serpentine rock mass in which 10 identical RMR values have been logged (RMR ¼ 89; continuous surveying with 5 m steps), but despite this apparent homogeneity, the mean thrust averaged over the 5 m steps (nominal data recorded every 0.2 m) varies from 4500 to 6000 kN, and the penetration rate from 2.0 to 2.6 m/h (note that the variation of TBM thrust is too large to be explained by an unnoticed variation of rock mass quality; see Fig. 6). Operator sensitivity and hardto-capture interactions between rock mass and TBM cutterhead are the possible source of data scatter, which seems to be unavoidable even for an experienced team. In fact, similar dispersions have been obtained in many different tunneling projects [22,52,53], apparently in the form of a random error superimposed to a simple trend. Assuming the scatter is normally distributed around the mean, performance prediction might be focused on the mean trend, neglecting the complicated pattern of real data. But if we just deal with a rough estimate of the average penetration, do we really need a large number of parameters in our prediction models? Table 4 gives a preliminary answer to this question. Following the approach recently proposed by Sundaram et al. [53], the table summarizes TBM performance data and corresponding correlation levels with main geomechanical classifications and basic rock mass and intact rock properties. As can be seen, even if the strongest correlation coefficients are those related with rock mass conditions (RMR; Q; rock mass uniaxial strength) rather good correlation is also shown by a basic parameter like the uniaxial compressive strength of the intact rock. A large number of parameters is probably essential when the relative importance of discontinuities over intact rock properties is high, but we should consider the difficulties involved when many rock mass parameters are involved. The correlation coefficient of QTBM ; for example, which contains factors of special relevance to TBM penetration, is even slightly lower than conventional Q: As the objective of the prediction (penetration rate) exhibits such a large random scatter, simple parameters probably give similar or even better results than comprehensive indexes. This conclusion agrees with the results presented by Morgan et al. [56] on the TBM construction of the

Table 4 Correlation values (r) of machine parameters with average intact rock (UCS) and rock mass properties Machine parameters

UCS

UCSRMa

UCSRMb

Penetration rate Field penetration index

0.36 0.40

0.46 0.48

0.44 0.40

Machine parameters

RMR

Log(Q)

Log(QTBM )

Penetration rate Field penetration index

0.42 0.44

0.41 0.50

0.37 0.26

a

Rock mass uniaxial compressive strength following Hoek and Brown [54]. b Rock mass uniaxial compressive strength following Singh [55].

Kielder tunnel, where it was found that Schmidt hammer rebounds were much better correlated with TBM performance than conventional classification indexes, and that better correlations emerge using an averaging method over ‘geological lengths’ of the tunnel, a way to smooth out the inherent scatter of penetration data.

6. Conclusions Data from the three tunnels excavated in predominately hard metamorphic rocks support the following conclusions: (1) The correlation between penetration rate and Rock Mass Rating is significant from a statistical point of view and can be approximated by a second-degree polynomial curve. Best performances have been recorded in fair rock (RMR ¼ 40270) whilst slower penetrations were experienced both in too bad (RMRo30 40) or too good (RMR > 70 80) rock masses, as a consequence of thrust reduction in the former case and reduced ability of cutter indentation and chips formation in the latter. (2) Despite the significant correlation, empirical relations are of very limited use in terms of predicting machine performance, even for a specific rock– machine combination. The scatter about the mean trend is in fact remarkably high, the penetration rate varying up to 50% of mean value for a given RMR: Literature review confirms this scatter is not a limitation of our dataset; rather, it is a common feature in many TBM projects, and it is probably related to the difficulty in maintaining a constant thrust during excavation. (3) Several improvements should be made to the conventional RMR-system if it is to predict TBM performance. Different penetrations have been obtained in different rocks for the same RMR value, suggesting the need of RMR normalization


with reference to parameters of special relevance to bored tunnels. As regard rock properties, simple analyses of our data showed that rock hardness could be suitable for this purpose. (4) Comparison of actual penetrations with those predicted by the Innaurato [21] and Barton [23] model showed poor agreement. As regards the Innaurato model, the mismatch is probably due to the absence of machine-related factors, which limits its application to rock–machine combinations similar to those considered by the author. In the case of the Barton model the poor result is much more difficult to explain, as the new term QTBM has additional rock–machine interaction parameters of special relevance for TBM applications. In particular, QTBM shows low sensitivity to penetration rate, and the correlation coefficient with recorded data is even worse than conventional Q or other basic parameters like the uniaxial compressive strength of the intact rock. Obviously, the reliability of the Barton model cannot be judged by an individual case, but the mismatch underlines the difficulties involved in performance prediction when so many factors (rock mass condition, machine and muck removal system characteristics, human experience) are involved. Finally, it is important to note that empirical relations discussed above are based on rock mass surveying during the excavation, that is considering the rock mass conditions at depth. At the design stage instead, especially for deep tunnel, performance prediction mostly deal with geomechanical surveys of outcropping rocks, whose characteristics may be significantly worse as a consequence of superficial weathering and stress removal effects [57]. A preliminary analysis involving more than 20 km of TBM tunnels has shown that an increase of rock mass quality is experienced both in terms of Q and RMR: For example, an increase up to 15–20 RMR points may be expected at depth, the entity of the variation being a function of the RMR value itself. The detailed analysis of this effect is still in progress and it will be the topic of a future paper. In order to promote refinements of existing predictive models and to facilitate the comparison with other experiences, the authors are happy to place the data set used in this paper at everyone’s disposal. Data files may be downloaded from our web page: www.geomin.unibo.it/ORGV/geoappl/TBM Performance.htm.

Acknowledgements Authors wish to thank the colleagues of Maen, Pieve Vergonte, and Varzo sites for their help in the collection of machine performance data and their support in

787

fieldwork. We are also grateful to the reviewers for their careful reading of our manuscript and their many helpful comments.

References [1] Wallis S. Record setting TBMs on China’s long Yellow River drives. Tunnel 1999;1:19–26. [2] Barla G, Pelizza S. TBM tunnelling in difficult ground conditions. GeoEngineering 2000, International Conference on Geotechnical and Geological Engineering, Melbourne, Australia, 2000. [3] Innaurato N, Mancini R, Stragiotti L, Rondena E, Sampaolo A. Several years of experience with TBM in the excavation of hydroelectric tunnels in Italy. International Congress on ‘‘Tunnel and Water’’, Madrid, 1988. [4] Lislerud A. Hard rock tunnel boring: prognosis and cost. Tunnelling Underground Space Technol 1988;3(1):9–17. [5] Nelson PP. TBM performance analysis with reference to rock properties. In: Hudson JA, editor. Comprehensive rock engineering, vol. 4 (10). New york: Pergamon Press, 1993. p. 261–91. [6] US Army Corps of Engineers. Manual No. 1110-2-2901— Engineering and Design—TUNNELS AND SHAFTS IN ROCK, 1997, http://www.usace.army.mil/inet/usace-docs/engmanuals/em1110-2-2901/toc.htm. [7] Norwegian Tunnelling Society (NFF). Norwegian TBM Tunnelling. Publication No. 11, 1998. [8] Alber M. Prediction of penetration, utilization for hard rock TBMs. Proceedings of the International Conference of Eurock ’96, Balkema, Rotterdam, 1996. p. 721–5 [9] Nielsen B, Ozdemir L. Hard rock tunnel boring prediction and field performance. In: Proceedings of the RECT. Boston, Littleton, Society for Mining, Metallurgy, and Exploration, 1993. p. 833–52. [10] Dolcini G, Fuoco S, Ribacchi R. Performance of TBMs in complex rock masses. North American Tunnelling ’96, Balkema, Rotterdam, 1996. p. 145–54. [11] Boniface A. Tunnel Boring Machine performance in basalts of the Lesotho Formation. Tunnelling Underground Space Technol 2000;15(1):49–54. [12] Graham PC. Rock exploration for machine manufacturers. In: Proceedings of the Symposium on Exploration for Rock Engineering, vol. 1, Johannesburg, Balkema, 1976. p. 173–80. [13] Hughes HM. The relative cuttability of coal measures rock. Mining Sci Technol 1986;3:95–109. [14] Farmer IW, Glossop NH. Mechanics of disc cutter penetration. Tunnels Tunneling 1980;12(6):22–5. [15] Nelson PP, Ingraffea AR, O’Rourke TD. TBM performance prediction using rock fracture paramaters. Int J Rock Mech Min Sci Geomech Abstr 1985;22(4):189–92. [16] Sharp W, Ozdemir L. Computer modelling for TBM performance prediction and optimization. Proceedings of the International Symposium on Mine Mechanization and Automation, CSM/ USBM, vol. 1(4), 1991. p. 57–66. [17] Wanner H, Aeberli U. Tunnelling machine performance in jointed rock. Proceedings of the Fourth ISRM Congress, vol. 1, Montreaux, Balkema, 1979. p. 573–9. [18] Johannessen O, et al. NTH Hard rock tunnel boring. Project Report 1–94, NTH/NTNU Trondheim, Norway, 1994. [19] Johannessen O, Jacobsen K, Ronn PE, Moe HL. Tunnelling costs for drill and blast. Project Report 2C-95, Unit-NTH, Department of Building and Construction Engineering, 1995. [20] Cassinelli F, Cina S, Innaurato N, Mancini R, Sampaolo A. Power consumpion and metal wear in tunnel-boring machines:

788

[21]

[22]

[23] [24] [25]

[26] [27]

[28]

[29]

[30] [31]

[32]

[33]

[34]

[35]

[36]

M. Sapigni et al. / International Journal of Rock Mechanics & Mining Sciences 39 (2002) 771–788 analysis of tunnel-boring operation in hard rock. Tunnelling ’82, London, Inst. Min. Metall., 1982. p. 73–81. Innaurato N, Mancini R, Rondena E, Zaninetti A. Forecasting and effective TBM performances in a rapid excavation of a tunnel in Italy. Proceedings of the Seventh International Congress ISRM, Aachen, 1991. Grandori R, Jaeger M, Antonini F, Vigl L. Evinos-Mornos Tunnel, Greece—construction of a 30 km long hydraulic tunnel in less than three years under the most adverse geological conditions. Proceedings of the RECT, San Francisco, Society for Mining, Metallurgy, and Exploration, 1995. p. 747–67. Barton N. TBM performance in rock using QTBM. Tunnel Tunnelling Int, Milan, 1999;31:41–8. Barton N. TBM tunnelling in jointed and faulted rock. Rotterdam: Balkema, 2000. Barton N, Lien R, Lunde J. Engineering classification of rock masses for the design of tunnel support. Rock Mech 1974; 6(4):189–236. Bieniawski ZT. Engineering rock mass classifications. New York: Wiley, 1989. Eusebio A, Grasso P, Mahtab A, Innaurato A. Rock characterization for selection of a TBM for a railway tunnel near Geneva, Italy. Proceedings of the International Symposium on Mine Mechanics and Automation, CSM/USBM, vol. 1(4), 1991. p. 25–35. Alber M. Advance rates of hard rock TBMs and their effects on project economics. Tunnelling Underground Space Technol 2000; 15(1):55–64. Park CW, Park C, Synn JH, Sunwoo C, Chung SK. TBM penetration rate with rock mas properties in hard rock. Proceedings of the AITES-ITA 2001 World Tunnel Congress, Milan, 2001. p. 413–9. Dal Piaz GV. Revised setting of the piedmont zone in the northern Aosta valley, Western Alps. Ofioliti 1988;13:157–62. Reinecke T. Very-high-pressure metamorphism and uplift of coesite-bearing metasediments from the Zermatt-Saas zone, Western Alps. Eur J Mineral 1991;3:7–17. Bethaz E, Fuoco S, Mariani S, Porcari P, Rosazza Bondibene E. Scavo in rimonta con TBM: l’esperienza del cunicolo di Maen. Gallerie e Grandi Opere Sotterranee 2000;61:67–75. Bethaz E, Peila D, Innaurato N. Prevision of hard rock TBM boring time. Proceedings of the AITES-ITA World Tunnel Congres, 2001. Reinhardt B. Geologie und Petrographie der Monte Rosa-Zone, der Sesia-Zone und des Canavese im Gebiet zwischen Valle Ossola und Valle Loana (Prov. Di Novara, Italien). Schiez Mineral Petrol Mitt 1966;46:553–678. Dal Piaz GV, Hunziker JC, Martinotti G. La Zona Sesia-Lanzo e l’evoluzione tettonico—metamorfica delle Alpi nordoccidentali interne. Mem Soc Geol Ital 1972;11:433–66. Compagnoni R, Dal Piaz GV, Hunziker JC, Gosso G, Lombardo B, Williams PF. The Sesia-Lanzo Zone, a slice of continental crust with alpine high pressure-low temperature assemblages in the Western Italian Alps. Rend Soc Ital Mineral Petrol 1977;33: 281–334.

[37] Milnes AG. Structural reinterpretation of the classic Simplon Tunnel Section of the Central Alps. Geol Soc Am Bull 1973; 84:269–74. [38] Steck A. Le massif du Simplon. Reflexions sur la cin!ematique des nappes de gneiss. Schweiz Mineral Petrogr Mitt 1987;67:27–45. [39] Mancktelow N. The Simplon line: a major displacement zone in the western Lepontine Alps. Eclogae Geol Helv 1985;78:73–96. [40] Scolari F. Open-face borers in Italian Alps. World Tunnelling 1995:8;361–6. [41] Davis J. Statistics and data analysis in geology. New York: Wiley, 1973. [42] Wickham GE, Tiedmann HR, Skinner EH. Ground control prediction model—RSR concept. Proceedings of the Rapid Exc. and Tunnelling Conference, AIME, New York. 1974. p. 691–707. [43] Sanio HP. Prediction of the performance of disc cutters in anisotropic rock. Int J Rock Mech Min Sci Geomech Abstr 1985;22(3):153–61. [44] West G. A review of rock abrasiveness testing for tunnelling. International Symposium on Weak Rock, vol. 1, Tokyo, 1981. p. 584–5. [45] West G. Rock abrasiveness testing for tunnelling. Int J Rock Mech Min Sci Geomech Abstr 1981;26(2):151–60. [46] Suana M, Peters Tj. The Cerchar Abrasivity Index and its relation to rock mineralogy and petrography. Rock Mech 1982;15:1–7. [47] Movinkel T, Johannesen O. Geological parameters for hard rock tunnel boring. Tunnels Tunneling 1986:45–8. [48] Bruland A, Dahl TS, Nilsen B. Tunnelling performance estimation based on drillability testing. Proceedings of the Eighth ISRM Congress, vol. 1, Tokyo, Balkema, 1995. p. 123–6. [49] Wijk G. The stamp test for rock drillability classification. Int J Rock Mech Min Sci Geomech Abstr 1989;26(1):37–44. [50] Howarth DF, Rowlands JC. Quantitative assessement of rock texture and correlation with drillability and strength properties. Rock Mech Rock Eng 1987;20:57–85. [51] Innaurato N, Oggeri C, Oreste P. Validation techniques in tunnelling: the complementary approach of modelling, monitoring and excavation performance evaluation. Proceedings of the AITES-ITA 2001 World Tunnel Congres, 2001. p. 217–26. [52] McKelvey JG, Schultz EA, Helin TAB, Blindheim OT. Geomechanical analysis in S. Africa. World Tunnelling, 1996;9: 377–90. [53] Sundaram NM, Rafek AG. The influence of rock mass properties in the assessment of TBM performance. Proceedings of the Eighth IAEG Congress, Vancouver, Balkema, 1998. p. 3553–9. [54] Hoek E, Brown ET. Underground excavations in Rock. The Institute of Mining and Metallurgy, Longon, 1980. [55] Singh B. Norwegian method of tunnelling workshop. New Delhi: CSMRS, 1993. [56] Morgan JM, Barratt DA, Hudson JA. Tunnel Boring Machine performance and ground properties. Transportation and Road Research Laboratory, Supplementary Report 469, 1979. [57] Gonzalez De Vallejo LI. A new rock classification system for underground assessment using surface data. Proceedings of the International Symposium on Eng Geol Underground Constr, Balkema, Rotterdam, 1983. p. 85–94.