Contents lists available at

Tunnelling and Underground Space Technology

j ourna l home page :

Analysis and estimation of gripper TBM performances in highly fractured

and faulted rocks
⇑
E. Paltrinieri a, , F. Sandrone a,b, J. Zhao a,c
a
Ecole Polytechnique Fédérale de Lausanne (EPFL), School of Architecture, Civil and Environmental Engineering, Laboratory of Rock Mechanics (LMR), Lausanne, Switzerland
b
Ecole Polytechnique Fédérale de Lausanne (EPFL), School of Architecture, Civil and Environmental Engineering, Laboratory of Experimental Rock Mechanics (LEMR),
Lausanne, Switzerland
c
Department of Civil Engineering, Monash University, Melbourne, Victoria 3800, Australia

A R T I C L E I N F O
A B S T R A C T
Article history:
Received 9 January 2015 This paper focuses on performance analysis of Tunnel Boring Machine (in particular gripper TBMs) in
Received in revised form 28 October 2015 highly jointed rock masses and fault zones. In order to investigate possible relationships between these
Accepted 24 November 2015 difficult excavation conditions and TBM performance, the data of several tunnel projects have been col-
Available online 14 December 2015 lected in a specific database (TBM-performance database). Despite the difficulties in gaining complete
TBM data sets and detailed geological information, this database compiles data from the field, laboratory
Keywords: tests and literature. Preliminary analyses have been carried out in order to find possible correlations
Weak rocks between TBM performance parameters (penetration and advance rates) and rock mass characteristics,
Fault zone classification such as rock strength and fracturing degree, generally used in common TBM performance prediction
TBM-performance database models. Although some trends could be identified, the scattered results confirm the difficulties in predict-
TBM-performance prediction
ing the machine performance in complex geological environments starting from characterising indexes
proper to good rocks. Then, a classification system for highly fractured rock masses and fault zones
has been developed, based on representative parameters, providing a more complete geomechanical
description of disturbed zones. Four ‘‘fault zone” classes have been identified. For each class a reduction
rate of the selected TBM-performance parameters (i.e. cutterhead rotation speed, penetration rate and
daily advance rate) has been evaluated with respect to the tunnelling performances recorded in good
ground conditions. The results of these analyses allow effectively quantifying the effects of degrading
rock masses on the TBM performances.
© 2015 Elsevier Ltd. All rights reserved.

1. Introduction E-mail addresses: [email protected], [email protected] (E.

Paltrinieri).

Tunnelling activities and in particular mechanised excavation

are more and more widespread in recent years. The great develop-
ment of Tunnel Boring Machines (TBMs) in tunnelling industry is
mainly due to its advantages with respect to conventional excava-
tion methods, such as continuous operation, safer working condi-
tions, reduced damage at ground surface and grater rate of
advance. However, because of the lack of versatility of the
machine, the TBM performance could be significantly reduced in
changing ground conditions, and in extreme geological situations
like very abrasive or very poor rock masses, where lower
advance rates are generally recorded.
One of the most problematic events that may occur in TBM tun-
nelling (especially for what concerns gripper TBMs) is crossing

* Corresponding author.
weakness zones. Although the presence of important fault An understanding of the geological development, especially
zones is generally identified along the tunnel alignment in the about the occurrence of weakness zones, is therefore important
project phase, in some cases it may still represent an in the planning and construction of a project, although they
unexpected incident due to either a lack of warning during generally occur along only 1–15% of a tunnel (Palmströ m and
excavation or an underesti- mation of the problem seriousness Berthelsen, 1988).
(Barla and Pelizza, 2000). Tun- nelling in such difficult ground Weakness ground conditions, besides referring to weak
conditions could cause an increase in the project cost (in (sandstone, siltstone, shale, mudstone, marl and chalk, tuff,
terms of additional ground improvements and consequent agglomerate) and weathered (hydrothermal and chemical) rocks
substantial delays) and plays a central role for the safe of all types, could be caused by jointing and shear zones or faults
completion of the works. present in the rock mass (Klein, 2001).

http://dx.doi.org/10.1016/j.tust.2015.11.017
0886-7798/© 2015 Elsevier Ltd. All rights reserved.
 E. Paltrinieri et al. / Tunnelling and Underground Space Technology 52 (2016) 44–61
4
As reported by several authors (Habimana et al., 2002; Loew Table 2
et al., 2010; Ramoni and Anagnostou, 2010; Schubert and Main specifications of the TBMs used for the tunnel projects analysed in this study.
Riedmü ller, 1997, 2000), the most common geotechnical problems
Diameter [m] 6–10
resulting from tunnelling in faulted/highly fractured zones are: Maximum thrust force [kN] 15,000–18,150
Maximum revolutions per minute (RPM) [rev/min] 5.1–8.2
● Instability of the excavation face. Maximum torque [kN m] 2300–9860
● Excessive overbreak of the tunnel contour. Number of cutters 42–68
Cutter diameter [mm] 419–432
● Excessive deformation when squeezing ground is encountered.
● Frequent changes (i.e. both magnitudes and direction) of stres-
ses and displacements.
● Large water inflows. Different tunnel categories have been considered and all of
them have been excavated by gripper TBMs. An open-type
The possible instability phenomena (such as collapses, ravel- machine is generally used in hard rock and good geological condi-
ling, slabbing, softening and slaking, swelling, squeezing and flow- tions or where considerable deformations are expected. In less
ing ground) are generally due to the low strength and to the high favourable ground conditions, such as fault zones and highly frac-
deformability of the involved weak and weathered rocks, further- tured rocks, the required additional supports, the potential grout-
more, they can be influenced by groundwater conditions and by ing and significant problems for the crew security make a gripper
discontinuities (bedding planes, joints, foliation and faults) in the TBM less efficient if compared to a shield-machine. Therefore, it
rock mass (Klein, 2001). has been possible to analyse the performance of the TBM in such
Besides an adequate ground investigation activity (both prior difficult environments, by considering the worst excavation condi-
and during construction), the prediction of the TBM performance tions represented by the disadvantages of an open machine. The
in such difficult zones becomes extremely important for planning main specifications of the gripper TBMs considered are reported
purposes. Due to the great importance of the TBM progress in in Table 2.
the final scheduling of a tunnelling project, many TBM perfor-
mance prediction models have been developed in the last decades,
but the majority of them still remain not suitable to be used in case 2.1. TBM performance database structure and description
of difficult ground conditions.
The goal of this paper is to assess the influence of difficult envi-
After defining the tunnel characteristics (i.e. average overbur-
ronments, such as fault zones and highly fractured rocks, on TBM
den, azimuth and diameter) for each project, the TBM performance
operations. This is done by investigating possible correlations
database is composed by two main sections:
between machine performance and geomechanical properties of
disturbed rock masses. With this aim, data from real projects
(1) TBM specifications and TBM performance parameters.
where TBM encountered difficult ground conditions have been col-
(2) Geological–geotechnical parameters of the rock mass.
lected and compiled in a specific database.
Concerning the compilation of geological–geotechnical parame-
ters, it is important to emphasise that they have been directly
2. TBM performance database
gath- ered from the projects or estimated by using existing
formulations commonly used in rock mechanics. The data quality
In order to investigate the correlations existing between
varies from one project to another, as a matter of fact the degree of
param- eters describing TBM performance and
detail is not the same and the information may be quantitative or
geological/geotechnical characteristics of a highly fractured and
qualitative.
faulted rock mass, a TBM-performance database has been
developed by compiling data coming from several tunnel projects.
In particular, three of them are briefly described in Table 1.
The attention has been focused on longitudinal tunnel sections 2.1.1. TBM specifications and performance parameters
in which difficult environments for the TBM operations have been The information related to boring operations, such as the bored
encountered, such as weak and fault zones. The sections are length, the effective boring time and the total time needed to com-
characterised by homogeneous geological/geotechnical conditions plete the excavation of a given tunnel section, is included in the
and mostly constant TBM performance over the entire length that database (Table 3). The operational parameters (recorded by on-
varies from 8 to 10 m to several tens of meters. A detailed descrip- board acquisition system) are varied by the TBM operator accord-
tion of the approach adopted to select the tunnel sections is pro- ing to the encountered ground conditions. Based on these data sets
vided in the following sections. it has been possible to estimate other important parameters usu-
ally taken into account for characterising TBM performances
(Eqs. (1)–(7), Table 4).

Table 1
Reference tunnel projects in this research work.

Project name Project location Tunnel Tunnel type Tunnel Tunnel main geology
construction length [km]
period

AlpTransit Lö tschberg Switzerland (Canton of Bern, 1999–2006 High speed 36.4 Limestone, shale, marl and schist (Autochton
Base Tunnel Canton of Valais) railway tunnel Gampel-Baltschieder); granite, granodiorite, granitic
gneiss and massive/schistose gneiss (Aar Massif)
Manapouri Power New Zeland (Fiordland 1997–2002 Tailrace tunnel 9.6 Non-banded and banded gneiss, calc-silicate rocks,
Station National Park, Southland) gabbro/diorite, diorite-gneiss, amphibolite,
amphibolite-gneiss, pegmatite and granite
Cleuson-Dixence Switzerland (Canton of Valais) 1991–1996 Headrace tunnel 15.8 Metamorphic rocks (nappes penniques of Grand-St-
Hydroelectric Bernard), greenstone (micaschist), gneiss (phyllite)
Complex and tectonised rocks (kakirite)
Table 3
Boring information, TBM specifications and TBM performances recorded for each section.

Boring information TBM specifications Operational parameters

Project name Section (first chainage-last chainage) Bored length [m] Cutterhead rotation speed, RPMmax [rev/min] RPM [rev/min]
Total time [h] Thrust force, Thmax [kN] Th [kN]
Boring time [h] Torque, Tqmax [kN m] Tq [kN m]

Table 4
Definition of the TBM parameters included in the TBM-performance database.

TBM-parameter Equation

Utilisation factor (U) of the TBM: amount of time in which the machine
Boring time ½h] (1)
has been effectively used for boring U¼ Total time ½h]
Penetration rate (PR): TBM advance speed computed without considering time required
for installing supports, TBM maintenance, etc. (Barton, 2000) (2a)
½m]
PR m
h ¼ Bored
Boringlength
time ½h]

Penetration per revolution (p): TBM advance speed computed without considering (2b)
time required for installing supports, TBM maintenance, etc. p rev
mm ¼ RPM×6
PR×1000
0
Daily advance rate (daily AR): TBM advance speed computed by considering
the TBM delays due to rock supporting, maintenance, etc.
Daily AR hdaym i ¼ ðPR × UÞ× 24 (3)

Field penetration index (FPI) (Hassanpour et al., 2011; Klein et al., 1995) FPI kN·rev
cutter·mm ¼ NTc h ; Nc ¼ number of cutters (4)
×p
TBM power consumption/use 2p×T q ×RPM
Power ½kW] ¼ h i (5)
60
Net production rate m3 RPM×p×60×p×£2
Net Production Rate ¼ ; (6)
h 1000×4
£ ½m] ¼ tunnel diameter
h i
Specific energy: energy needed for excavating a unit volume of rock MJ 3:6×Power (7)
Specific Energy m3
¼ Net Production Rate

2.1.2. Rock mass parameters challenging environments, as fault zones and highly fractured
For what concerns the rock mass characteristics, attention has rocks, laboratory test could not be performed. Thus, in those cases
been focused on the parameters reported in Table 5, directly the ranges of strength values reported in the literature have been
gathered from the projects or estimated by using commonly used taken into account.
formula in rock mechanics from laboratory tests results
(Eqs. (8)–(11b), Table 5).
2.3. Selection criteria for data collection

2.2. Data quality In order to select the tunnel sections to be included in the TBM-
performance database, attention has been focused firstly on
The distributions of the different sources of compiled data are stretches identified as difficult/very difficult ground conditions,
reported in Tables 6 and 7, both for TBM performance and geolog- such as highly fractured or faulted rocks, over a length of 8–10 m
ical–geotechnical parameters. at least. The selected zones are generally characterised by RMR
As it can be clearly seen in Table 6, though the calculation of the values lower than 40, characterising poor (IV class) to very poor
basic machine parameters has been generally performed on the (V class) rock quality according to the classification provided by
basis of the measures carried out by the TBM on-board acquisition Bieniawski (1976). These sections are also generally characterised
system during construction, it was not possible to gain complete by large to very large water inflows (>60 l/min). For what
TBM data for all tunnel projects. This is due to the fact that the concerns the joint spacing (sj), five classes have been considered
contractors are often unable to easily share data especially when (Table 8) and only the sections belonging to classes C, D and E
on-going claims have not been solved yet. However, although have been included in the database. When only a qualitative
important TBM performance parameters (such as penetration per description of the rock mass conditions was available (i.e. from
revolution, thrust force and torque unavailable for up to 70% of ‘‘slightly frac- tured” to ‘‘disintegrated” in Table 8), sections
the collected data) could not been compiled for all projects, it characterised by ‘‘Highly Fractured” (corresponding to classes C
has been possible to obtain the daily advance rate (meters per and D) and ‘‘Disin- tegrated” (corresponding to class E) rocks have
day) for each section included in the analysis. This has been done been included in database.
by starting from available time-way diagrams, and by using infor- On the basis of the joint spacing the volumetric joint count (Jv)
mation gathered from the literature (e.g. delays and stoppages of has been computed. Though Jv is always rather high in the data-
the machine) about each specific case. base, some sections are characterised by lower values of Jv. This
With regard to the rock mass characterisation, it is important to is due to the fact that other factors (such as the weathering degree
underline that, mainly due to the fact that they belong to the tun- or water inflows) were considered while choosing the sections to
nel owner and not to the contractor, geological–geotechnical be compiled. Due to the available data quality, the identification
descriptions of tunnel projects are easier to obtain than TBM per- of classes that quantitatively describe the weathering degree of
formance data. As it is possible to see, several sources have been the joint surfaces was not an easy task. As a matter of fact, the
considered: geological–geotechnical reports, laboratory tests, characteristics of the joints (roughness, opening, alteration/filling)
stereographic projections, tunnel profile surveys, face and side- are not always clear or well-defined. This is the case for example of
wall pictures and literature. Except for the joint orientation extremely fractured rock masses where it is no more possible to
(38% of data were not available), all parameters could have been identify the discontinuity sets. In those cases the most relevant
collected for each tunnel project, though with a variable detail information is represented by the rock conditions, where the
and reliability. It is also important to consider that in extremely material alteration and weakness can be considered as results of
Table 5
Definition of the rock mass parameters included in the TBM-performance database.

Geological–geotechnical parameter Source

Uniaxial compressive strength, UCS [MPa]

Brazilian Tensile Strength, BTS [MPa] Laboratory tests on rock samples collected during excavation,
Joint orientation detailed geological/geotechnical reports, literature
Joint surface weathering conditions Face mappings, face/side-walls pictures, tunnel profile surveys
Joint spacing drawn during construction
Water inflows Project geological–geotechnical reports, tunnel surveys
Rock Mass Rating, RMR (Bieniawski, 1976) Computed (or evaluated) on the basis of collected data
Q Index (Barton et al., 1974)
Geological Strength Index, GSI (Hoek, 1983; Hoek and Brown, 1988)
P h
Volumetric joint count (Jv) N rð5Þ (8)
J v ½joints=m3 ] ¼ s
1
þ 5
i i

si ½m] ¼ average spacing of the ith joint set

Nrð5Þ ¼ number of random joints alonga 5 m scan line
Rock Quality Designation (RQD) RQD½%] ¼ 100e—0:1k ð1 þ 0:1kÞ (9)
(Hudson and Priest, 1979) k ¼ total joint frequency
Uniaxial compressive strength of the rock mass (UCSrm) (Hoek et al., 2002) UCSrm ½MPa] ¼ UCSsa (10a)
Hoek and Brown parameters (s and a) s½—] ¼ expðGSI—100Þ (10b)

with D 9—3D
¼ 0ðTBM tunnellingÞ
1
a½—] ¼ þ 1 ðe—GSI=15 — e—20=3Þ (10c)
2 6
Uniaxial compressive strength for cataclastic rocks (Habimana, 1999; Habimana et al., UCSH ½MPa] ¼ tUCS (11a)
2002)
Reduction coefficient (t). It varies between 1, in case of intact rock, and 0, in case of a
GSI 0:55 (11b)
soil-like material (Habimana, 1999; Habimana et al., 2002) t½—] ¼ 100

Table 6
TBM performance parameters source distribution for the collected data. the rock strength (including uniaxial compressive strength, brittle-
ness as well as rock hardness) and the fracturing degree (including
TBM-performance TBM on-board Time-way No joint spacing, RQD and Jv) are the most used parameters in the
parameter acquisition system (%) diagram/literature data existing TBM performance prediction models.
Penetration rate 62 – 38% The preformed analyses (Paltrinieri and Sandrone, 2014) show
Advance rate 62 38% – the difficulties in selecting, within existing prediction models, the

(delays) parameters affecting the most the TBM performance while degrad-
RPM 62 – 38%
Thrust force 62 – 38%
ing the ground conditions. As a matter of fact, in highly fractured
Torque 31 – 69% and faulted zones, the delays associated to the need of heavier
temporary supports and additional drainage system, the potential
TBM jams, the gripper bearing failure, etc., are partially balanced
faulting and folding processes. Table 9 summarises the criteria by the rather high TBM penetration rates generally observed in
used for describing the rock conditions. On the basis of the nature fractured/very fractured rock masses.
and detail degree of the available data, at least one of the three cri- Since from the first results it came out that the rock strength, in
teria has been used for identifying the class. terms of UCS, was often too high (Fig. 2a) to describe the real
Finally, to complete the description of the rock mass conditions, beha- viour of the considered rocks, the uniaxial compressive
also the available information about the (temporary) supports has strength (UCSH) reduced by Habimana (1999) and by Habimana et
been taken into account: only tunnel sections with heavy to very al. (2002) has been introduced for analysing possible correlations
heavy supporting measures have been considered. between rock properties and TBM performance. By taking into
account the weathering degree of rock mass (both rock and joint
surfaces alteration) via the Geological Strength Index (GSI), the
3. Preliminary analyses UCSH allows a better description of the behaviour of weak materi-
als. Despite a significant lowering (Fig. 2b) of strength values (thus
Based on the selection criteria reported in previous section, more representative of the real conditions), a large data scattering
more than 100 tunnel sections have been included in the TBM- remains and the correlations with p and PR proved to be almost
performance database. First analyses have been performed in meaningless (Fig. 3a and b).
order to investigate existing correlations among the rock mass In order to cope with the uncertainties related to the rock mass
parame- ters commonly taken into account in the existing TBM- characterisation (representing one of the major cause of the large
performance prediction models and TBM advancement (i.e. daily data scattering), it has been decided to adopt another approach
AR) and rate of penetration (i.e. p and PR). As reported in Fig. 1, by considering the rock mass conditions as a combination of

Table 7
Geological–geotechnical parameters sources distribution for the collected data.

Geological–geotechnical Geological–geotechnical Laboratory Literature Face and side-walls Tunnel profile Stereographic No
parameter reports tests photgraphs surveys projections data
Rock strength – 62% 38% – – – –
Joint orientation 31% – – – – 31% 38%
Joint spacing – – – – 100% – –
Joint surface/rock weathered 31% – – 31% 38% – –
conditions
Water inflows 62% 38%
Table 8
Quantitative and qualitative descriptions of the rock mass fracturing degree. have been identified (by referring to the average performance val-
ues for different classes of rock mass quality), the RMR seemed to
Fracturing degree class Quantitative description Qualitative description be still not the most appropriate system for characterising difficult
(according to the joint spacing, sj [m] ground conditions in the TBM performance estimation (Paltrinieri
quantitative description)
and Sandrone, 2014). As a matter of fact, it has been observed that
A sj > 5 Slightly fractured some parameters which are basic factors for the evaluation of the
B 1 < sj < 5 Fractured
RMR, (e.g. the rock strength) and which are extremely significant
C 0.1 < sj < 1 Highly fractured
D sj 6 0.1
for the study of the TBM behaviour in hard rock tunnelling, may
E sj 6 0.01 Disintegrated actually play a very marginal role in the definition/estimation of
the machine performance in highly fractured and faulted zones.
Furthermore, it is necessary to remind that the conventional RMR
Table 9 system has been developed to provide support guidelines for
Rock conditions classes according to three different criteria. underground construction excavated by the Drill-and-Blast
Class Rock behaviour Rock weathering Rock strength,
method. This means that it does not take into account rock–ma-
UCS [MPa] (ISRM, 2014) chine interaction parameters, and therefore, it might be
1 Good Unweathered Very strong (>100)
inadequate for TBM performance prediction.
2 Fair Slightly weathered Strong (60–100) The preliminary results obtained in this and previous works
3 Mediocre Weathered Medium strong (40–60) highlight the great difficulty in characterising the degraded and
4 Very poor Highly weathered Weak (<40) altered rock mass because of the complexity in identifying and
quantifying the most relevant factors within the parameters usu-
ally taken into account in TBM performance analyses. Due to this,
in order to find relationships between highly fractured/faulted
rock masses and TBM performances (and potential construction
delays), it has been decided to better focus on the rock mass (i.e.
fault zones) characterisation.

4. Fault zone classification system and correlations with TBM

performances

Due to the very complex structure and nature of the fault zones,
thus particularly difficult to be characterised, detailed criteria have
been chosen for developing a specific classification method.
Though in literature fault rocks (generally termed ‘cataclastic
rocks”) result from a mechanical and chemical weathering related
to tectonic activities (Christe, 2009), in this framework, the term
‘‘fault zone” refers, with a more general approach, to difficult
Fig. 1. Geological–geotechnical parameters generally evaluated for TBM-perfor-
mance prediction. The frequency refers to a sample of 20 existing prediction ground conditions encountered in TBM tunnelling, from extremely
models. jointed rock masses up to completely faulted/sheared rocks. From
a geomechanical point of view, they can be characterised as weak
and weathered rocks.
factors. The relationships existing between the TBM performance Starting from an existing compilation of data done by Santi
parameters (i.e. PR and AR) and the Rock Mass Rating (RMR) have (2006), Table 10 summarises geological and geomechanical
been therefore investigated (Paltrinieri and Sandrone, 2014). The param- eters, as well as their relative ranges of values and
results proved to be consistent with previous works made by bibliographical sources, generally used for describing weathered
Sapigni et al. (2002) where the Authors approximated the correla- and weak rocks, which are the basis for the proposed
tion between PR and RMR by a second-degree polynomial curve classification.
where the maximum TBM performance was reached for rock In order to identify the ‘‘fault zone” classes, two major aspects
masses of medium quality (i.e. RMR = 40–60). Since only sections have been considered: the fracturing degree and the weathering
characterised by RMR lower than 40 have been included in the degree of the rock mass.
database, only the increasing of PR with the rock mass quality For what concerns the fracturing degree of the rock mass, it has
improvement could be investigated. Though a certain trend could been decided to consider only blocks with small to very small size,
according to values reported in Table 8. Due to the well-defined

Fig. 2. Classes of strength of the rocks included in the database: (a) Uniaxial compressive strength (UCS); (b) uniaxial compressive strength (UCSH).
Fig. 3. Correlation between the rate of penetration (p and PR) and the uniaxial compressive strength (UCSH) reduced by Habimana (1999) and by Habimana et al. (2002).

Table 10
Summary of engineering properties of weak rocks, after Santi (2006).

Parameter Range of values for weak rocks Reference

Compressive strength 1–40 MPa Afrouz (1992), Santi (2006), and Thuro and Scholz (2003)
Rock quality designation <25–75% Santi and Doyle (1997) and Santi (2006)
Ratio between weathered matrix and >75% matrix Geological Society Engineering Group Working Party (1995)
unweathered blocks
Volumetric joint count, Jv (joints/m 3) Small blocks: 10 < Jv < 30 joints/m3
Santi and Doyle (1997)
Very small blocks: Jv > 30 joints/m3
Dearman weathering classification category PCategory 4 Santi (1995)
RMR (Bieniawski, 1976) <35–60 Santi (1995)
Norwegian Geotechnical Institute ‘‘Q” rating <2 Santi (1995)
Extended GSI <30 Hoek et al. (1998), Marinos and Hoek (2001) and Marinos et al. (2007)

joint sets described by most of the available data, the volumetric

(rock and soil-like materials), the more recent Geological Strength
joint count (Jv) has been selected as referring parameter. A value
Index, developed for heterogeneous rock masses (Marinos and
of Jv equal to 10 joints/m3 has been adopted as lower limit, below
Hoek, 2001; Marinos et al., 2007), has been introduced. This index
which the rock mass cannot be described anymore as highly frac-
refers to the composition and structure of the mass, in terms of
tured or faulted. Moreover, a reference value of 35 joints/m 3 has
tec- tonic disturbance and alternation of the weaker layers, rather
been considered as the highest Jv for a rock mass where individual
than to the blockiness and interlocking of rock pieces, while the
joint sets can be still identified.
surface conditions of the discontinuities predominantly refer to
The Geological Strength Index (GSI) (Hoek, 1983; Hoek and
the bed- ding planes.
Brown, 1988) has been considered for describing the different
Finally, it is important to underline that the weathering degree
alteration/weathering degree suffered by the rock mass (and by
of the rock (and of the joint surfaces) could be a reference param-
the discontinuities surfaces). This system provides an estimation
eter in order to describe the potential fluid flow in a fault zone
of the rock mass properties based on qualitative descriptions of
since the products of rock alteration (generally fine-grained mate-
the blockiness of the mass and of the conditions of the joint sur-
rial) could significantly reduce the permeability. Possible amount
faces. The GSI system has been modified over the years (Hoek
of water may vary from low to extremely high in all the identified
et al., 1998; Marinos and Hoek, 2001; Marinos et al., 2007) in order
classes and it is linked to the permeability of the rock mass. Fur-
to be adopted in many different geological situations, including
thermore, the hydrological properties of a fault zone can be
sheared rock masses characterised by very poor quality and
strongly influenced by the rheology of the host rock. As a matter
heterogeneous and tectonically disturbed lithological formations.
of fact, the brittle deformation of competent and incompetent
According to the description given in Hoek et al. (1998), the ‘‘foli
rocks leads to different permeability characteristics of the crushed
ated/laminated/sheared” class consists of non-blocky structure
material. According to the results of Sausgruber and Brandner
rock masses with a low strength and a high deformability, where
(2003), brittlely deformed incompetent host rocks create zones
a well-defined persistent and closely spaced lamination (or folia-
with soil-like features, thus characterised by a lower permeability.
tion) is dominant with respect to any other discontinuity set and
On the contrary, brittlely deformed competent host rocks create
shows slickenside surfaces characterised by an important
nearly non-cohesion zones which are extremely permeable and
weather- ing degree. Unlike the disintegrated category, where the
characterised by a higher potential of water inflow. In this frame-
rock-to- rock contacts govern strength and deformability, the
work, water inflows of faulted/fractured zones are considered
mechanism of deformation in foliated/laminated/sheared class is
only as an additional factor because it contributes to the reduction
controlled by the displacements along the foliation planes and by
of the TBM performance independently of the class.
the shear strength of the fines along the surfaces (Hoek et al.,
In this study, four ‘‘fault zone” classes have been proposed,
1998). The pre-sheared nature of the rock mass is a condition
which can be distinguished in two groups according to the govern-
easily encoun- tered in fault zones. Based on this consideration,
ing weathering process:
extremely low values of GSI (i.e. lower than 5) have been
considered in this framework. Moreover, in order to take into
(1) Mostly affected by a mechanical weathering, representing
account the complex nature of a fault zone, often constituted of a
the fracturing and the potential opening of the joints (Class
combination of mate- rial with different characteristics of strength
I and Class II).
and deformability
50 E. Paltrinieri et al. / Tunnelling and Underground Space Technology 52 (2016) 44–61

(2) Characterised by a chemical weathering (i.e. the decomposi-

tion of the rock material) due to faulting and folding pro-
cesses (Class III and Class IV).

For what concerns the geomechanical behaviour, the uniaxial

compressive strength of the rock mass (UCS rm) has been consid-
ered for describing Classes I and II, due to the significant role
played by the blockiness of the mass and by the weathering degree
of the joint surfaces. For Classes III and IV, due to the fact that the
blockiness (expressed by Jv) has not sense anymore, the uniaxial
compressive strength proposed for cataclastic rocks (UCS H,
Eq. (11a)) by Habimana (1999) and Habimana et al. (2002), has
been introduced for describing the chaotic structure (Class III)
rep- resented by rock fragments (clasts) dispersed in a fine-
grained matrix (Bü rgi, 1999) and for characterising the crushed
rock masses (Class IV).
Fig. 4 summarises the procedure adopted for classifying the col-
lected data according to the description given in the previous
sections.
Fig. 5 shows the distribution of the ‘‘fault zone” classes within Fig. 5. Percentages of tunnel sections representing the different ‘‘fault zone” classes
the TBM-performance database. As it is possible to observe, Class in the TBM-performance database.
I and Class II (i.e. highly fractured rock masses characterised by
different degrees of weathering) represent the most common con-
ditions encountered in the analysed tunnel projects. best and the most frequent (i.e. mode) performance recorded in
good tunnelling conditions. In order to have information about
TBM performance in ‘‘normal conditions”, data about stretches
4.1. TBM performance reduction analyses where a daily advance rate greater than 15 m/day had been
recorded have been included in the existing database. Based on
Though it is quite difficult to evaluate with exactitude the TBM the available project data, this value seems to be a reasonable
behaviour in a fault zone, the performance decrease of the lower limit for the advancement speed in good tunnelling condi-
machine in difficult ground conditions has been studied with tions. The sections with this advance rate are characterised by
respect to the

Fig. 4. ‘‘Fault zones” classification procedure.

 E. Paltrinieri et al. / Tunnelling and Underground Space Technology 52 (2016) 44–61
5
Table 11
Average, best and mode daily AR [m/day], PR [m/h] and p [mm/rev] for different lithotype classes with good tunnelling conditions.

Main Daily AR [m/day] PR [m/h] p [mm/rev]

lithotype
Average Best Mode Average Best Mode Average Best Mode
performance performance performance performance performance performance

Gneiss 22.1 27.7 20.1 1.70 2.02 1.63 5.26 6.25 5.38
Schist 16.8 23.1 19.9 2.59 2.82 2.74 7.44 8.13 7.86
Limestone 19.1 24.3 17.4 2.21 2.57 2.05 6.25 7.32 5.96
Granodiorite 28.5 32.4 28.5 2.27 2.35 2.27 6.33 6.56 6.15
Greenstone 23.8 31.8 21.9 – – – – – –
Amphibolite 29.7 31.7 31.1 1.76 2.02 1.73 5.74 6.60 5.14

unweathered rock masses with low or medium fracturing degree

formation, the water ingress and the experience of the contractor
and insignificant water inflows. The decrease of the machine
and operator. Though aware of the fact that the daily AR is proba-
performance, with respect to the good tunnelling conditions, has
bly the parameter more affected by the project characteristics (e.g.
been assessed for each lithotype already included in the TBM-
muck transportation by train or belt conveyor) and location, its
performance database for the highly fractured and faulted rocks.
estimation allows a more general view of the TBM performance
This allowed evaluating the behaviour of the machine in the same
reduction in difficult ground conditions.
lithotype with changing ground conditions.
Unfortunately, as previously said, while the computation of the
Table 11 shows the average values of daily advance rate
daily advance rate has been possible for all tunnel sections
(m/day) and penetration rate (m/h, mm/rev) recorded for
included in the TBM-performance database, the values of p, RPM
different lithotypes in favourable tunnelling conditions. Since
and PR are not on hand for some sections compiled in the
these values refer to several tunnel projects, each lithotype
database. However, it is possible to observe that the analyses
actually represents a ‘‘category” which can be slightly different
performed can be always considered representative since data are
from project to project. The lithotype ‘‘gneiss”, for instance, might
available for more than 50% of the total data recorded per each
correspond to crys- talline gneiss, gneiss with subordinate
class. Table 12 shows the ‘‘representativeness” of each ‘‘fault zone”
micaschist or gneiss with minor amounts of quartzite, calc-
class per type of TBM parameter analysed.
silicates and/or pegmatite. The best performance is given by the
average between the two highest values recorded for each
lithotype during construction, while the definition of the ‘‘mode”
4.1.1. Cutterhead rotation speed, RPM
mainly based on the identification of ranges for each TBM
The cutterhead rotation speed strongly depends on the dimen-
parameter. As a matter of fact, a defined single ‘‘mode” does not
sion of the machine, therefore, to do a comparison among several
exist for any lithotype due to the fact that each section is
different tunnel projects (with different diameters) can prove to
characterised by different performances. The mode value of the
be irrelevant. For this reason, it has been decided to take into con-
TBM-performance parameters, shown in Table 11, refers to the
sideration the percentage of reduction with respect to the real (i.e.
weighted average of each range mode. As reported in the pre-
maximum) value of RPM of the TBM (RPMmax). In Fig. 7 the
vious sections (Table 6), TBM parameters such as PR and p are not
decrease of RPM, with respect to RPM max, is represented for each
available for all tunnel projects, therefore some information is not
‘‘fault zone” class.
available (e.g. greenstone).
In Class I (Fig. 7a) the RPM keeps constant for most of the sec-
Existing prediction models have been analysed in order to iden-
tify whether it is possible to reliably predict the TBM performance tions where a minor reduction has been recorded. As a matter of
in good rocks. It has been observed that the average penetration fact, more than 80% of sections is characterised by RPM greater
and daily advance rates recorded in good tunnelling conditions than, or equal to, 80% of RPMmax, while less than 10% is reduced
fit quite well (Fig. 6) the performance predicted by the NTNU up to 60% of the maximum value.
(NTH) model, developed by Bruland (1998), where the fracturing A similar behaviour has been observed in Class II (Fig. 7b),
degree of the rock plays a very important role as well as the thrust where the high weathering degree of the joint surfaces seems
force of the cutterhead. However, it is important to underline that not affecting the RPM of the TBM. However, a certain reduction
the predicted values are strongly affected by two main factors: the with respect to Class I has been recorded: less than 70% of
reduced thrust per cutter and the utilisation factor U. These sections is characterised by a RPM greater than (or equal to) 80%
param- eters have been estimated according to the most frequents of RPMmax, while more than 10% shows significant decrease with
values used in literature. respect to the maximum value (down to 40% of RPMmax).
Knowing the TBM behaviour in good excavation conditions, it is In Class III, i.e. faulted rocks, (Fig. 7c) the reduction of RPM is
possible to estimate the decrease of performances due to difficult more evident. Approximately 70% of sections is indeed charac-
conditions. The data about the TBM-performance parameters, i.e. terised by a RPM lower than 80% of RPM max and approximately
the penetration per revolution p (mm/rev), the cutterhead 40% of them is reduced to values comprised between 40% and
rotation speed RPM (rev/min) and the penetration rate PR (m/h) 60% of the maximum value.
have been ‘‘normalised” by considering the average values of the An important reduction has been observed also in Class IV,
best and the most frequent performance (as expressed in Table described by crushed rock masses (Fig. 7d). In this case about
11) for each class. In addition, also the decrease of the daily 40% of tunnel sections keeps the RPM close to the maximum
advance rate AR (m/day) has been investigated, although this value, while a same number of sections is characterised by RPM
parameter is affected by a significant degree of uncertainty. As a between 40% and 60% of RPMmax.
matter of fact, as reported by Rostami (2008), the TBM The results obtained for the cutterhead rotation speed clearly
advancement takes into account several rock mass conditions and show the influence of highly fractured and faulted rocks on the
operational aspects as well as unpredictable factors such as TBM performance, especially in the last two ‘‘fault zone” classes,
downtimes and stops of the machine mainly due to the extra which represent indeed the worst tunnelling conditions. The
support measurements, the change of the cutters because of the RPM shows an inevitable reduction in challenging environments
abrasiveness of the rock
52 E. Paltrinieri et al. / Tunnelling and Underground Space Technology 52 (2016) 44–61

Fig. 6. Comparison between predicted (by NTNU model, where U is supposed equal to 0.4 and the thrust per cutter is reduced to 80% of the maximum value) and measured
TBM performance. (a) Penetration per revolution p; (b) penetration rate PR; (c) daily advance rate (daily AR).

(such as crushed rock masses) that hinder operational issues

related, for instance, to the efficiency of the muck removal system. Table 12
Percentages of tunnel sections considered in the estimation of the TBM-performance
parameters (per ‘‘fault zone” class).
4.1.2. Penetration per revolution, p
TBM parameter Class I (%) Class II (%) Class III (%) Class IV (%)
In Fig. 8 the decrease of the penetration per revolution (p), with
p, RPM, PR 76 52 61 71
respect to the best performance (pmax) and the most frequent
Daily AR 100 100 100 100
performance (pmode) recorded in good tunnelling conditions, is
represented for each ‘‘fault zone” class.
In Class I, i.e. highly fractured rock masses (Fig. 8a), the
majority of the tunnel sections (i.e. more than 60%) is
result seems to confirm the theory, already introduced in the pre-
characterised by a p-value approximately equal to the best
vious sections, that the high number of joint sets generally does
performance recorded in good tunnelling conditions (p greater
not represent a hindrance to the boring process. Regarding the
than 80% of pmax). A reduc- tion of about 40–60% is observed for
most frequent (mode) performance recorded in good tunnelling
only 10% of sections. This
 E. Paltrinieri et al. / Tunnelling and Underground Space Technology 52 (2016) 44–61
5

Fig. 7. The reduction of the cutterhead rotation speed RPM (rev/min) with respect to the RPMmax of the machine. The broken line represents the average trend in each class.
(a) Class I; (b) Class II; (c) Class III; (d) Class IV.

conditions, about 50% of the tunnel sections is characterised by a

possible to justify the increasing of the penetration with a high
p-value equal (or even greater) to pmode. A decreasing trend can
number of joint sets, the apparently easier boreability of the rock
be observed down to the range corresponding to 40–60% which
can be explained by a significant strength reduction due to an
interests less than 10% of sections. This result does not differ
intense faulting and folding process. Moreover, concerning the
signif- icantly from the p-reduction trend obtained by considering
mode performance analysis, a significant peak (i.e. about 70% of
the best performance (pmax).
tunnel sections) has been observed for the values equal to or
Similar results are observed for Class II. However, in this case,
greater than pmode recorded in good rocks and approximately 10%
more variable values for p have been recorded. As it is possible
of sections is characterised by a reduction down to 80%. This
to observe (Fig. 8b), about 20% of the tunnel sections is affected
trend is almost equivalent to the one obtained for the analysis
by an important performance decrease, i.e. down to 20% of the
done by considering the best performance values, where more
best performance value. Unlike Class I, only approximately 30% of
than 10% of tunnel sections is characterised by the greatest
sec- tions exceeds the 80% of the best performance recorded in
reduction (from 20% to 40% of pmax).
good tunnelling conditions. Lower values of p can be explained by
High values of p have been recorded also for the crushed rock
the decreased surface quality of the joints (poor and very poor),
masses described by Class IV (Fig. 8d). As it is possible to observe,
where a high weathering degree might signify soft clay joints-
about 50% of the tunnel sections is characterised by p greater than
fillings. Actually, this can represent a problem for the interaction
or equal to 80% of pmax, while, only about 20%, by a reduction com-
between rock mass and TBM cutters because the excavation may
prised between 20% and 40% of the best performance. In this
not pro- ceed anymore via the usual chipping process. Also in the
class the rock is characterised by the highest degree of mechanical
histogram describing the p-reduction with respect to the mode
and chemical weathering, reaching a soil-like state. This almost
performance, about 20% of the tunnel sections is affected by an
total loss of strength of the crushed material, seems, on one hand,
important decrease. Approximately 30% of sections exceeds the
lead- ing to a better boreability of the rock mass but, on the other
100% of the reference performance recorded in good tunnelling hand, could reduce the chipping efficiency. In the histogram
conditions and more than 20% of sections is characterised by a concerning to the mode performance analysis, about 50% of the
reduction com- prised between 60% and 80%. Compared to what tunnel sections is characterised by p greater than or equal to 100%
has been obtained for the best performance, a peak shift to the of pmode, and about 30% is reduced to 40% of the mode
right occurs for Class II. This is obviously due to the fact that the performance. The p-reduction trend obtained for the best
reference value is lower (pmode < pmax). However, a similar range performance is almost con- firmed. Only a slight increase of the
distribution can be observed in both situations. tunnel sections characterised by a value greater than or equal to
Regarding Class III a totally different behaviour has been
100% of the reference perfor- mance can be observed as well as
observed (Fig. 8c). As a matter of fact, more than 70% of sections
the shift to the right of the lowest reduction range. This is again
is characterised by a p-value greater than (or equal to) 100% of
explained by a lower reference value (pmode < pmax).
pmax. It is important to say that, in this class, the fracturing degree
Contrary to the effects that bad ground conditions have on the
is no longer a reference parameter. Therefore, though it is not
cutterhead rotation speed (RPM) which is affected by the
54 E. Paltrinieri et al. / Tunnelling and Underground Space Technology 52 (2016) 44–61

Fig. 8. The reduction of the penetration per revolution p (mm/rev) with respect to the best performance pmax (on the left) and the most frequent performance pmode (on the
right) recorded in the same lithotype with good ground conditions. The broken line represents the average trend in each class. (a) Class I; (b) Class II; (c) Class III; (d) Class IV.
 E. Paltrinieri et al. / Tunnelling and Underground Space Technology 52 (2016) 44–61
5
excavation problems at the TBM scale, the results of this analysis
than) the mode performance. Some sections show an important
clearly show that the penetration of the rock by the cutters (i.e.
reduction, down to reach PR lower than 20% of the PRmode.
at the tool scale) does not represent a particular issue in highly
In Class IV the values of PR do not reflect what has been
fractured and fault zones. Thus, the recorded p-values are not
observed for p (Fig. 9d). Despite approximately 30% of tunnel sec-
extremely low since a significant fracturing degree (together with
tions is characterised by PR greater than or equal to 80% of PR max,
the decrease of the rock strength) generally assists the crushing of
for more than 40% of sections PR shows a strong decrease with
the material by the cutting tools at the tunnel face. A similar effect
respect to the best performance recorded in good conditions,
has been observed by Delisio et al. (2013) and Delisio and Zhao
down to values lower than 20% of PRmax. Moreover, the clear
(2014) in the case of TBM excavation in blocky rock conditions.
reduction of RPM in this class seems to affect more PR than in
Class III. As a mat- ter of fact, it has been observed that in crushed
4.1.3. Penetration rate, PR
rocks a greater percentage of tunnel sections is affected by a
The penetration rate (PR), expressed as meters per hour, repre-
significant decrease (40–60% of RPMmax) of the cutterhead
sents the distance excavated by the TBM in a continuous excava-
rotation speed compared to the one in faulted rocks (i.e. Class III).
tion phase. This parameter is a combination of the penetration
Regarding the mode per- formance analysis, more than 50% of
per revolution (p) and the cutterhead rotation speed (RPM) as
tunnel sections is charac- terised by PR greater than or equal to
expressed by the Eq. (2b) (Table 4). In particular, if RPM increases
80% of PRmode and approximately 50% of sections by PR lower
(or decreases), PR can significantly increase (or decrease) aside
than 60% of the mode performance recorded in good conditions.
from the value of p. This means that PR does not depend only on
The results obtained for the reduction with respect to the mode
the real penetration at the cutter scale (mm/rev), but, being
performance do not differ from what has been observed for the
strongly related to the rotational speed of the cutterhead, it repre-
best performance analysis. As a matter of fact, as it is possible to
sents somehow the penetration at the TBM-scale (without consid-
see, though a shift to the right of the histogram peak can be
ering downtimes and possible stops of the machine). In Fig. 9 the
observed, due to the difference between PRmax and PRmode
decrease of PR, with respect to the best performance (PR max) and
(PRmode < PRmax), the PR-reduction maintains the same trend for
the most frequent performance (PRmode) recorded in good tun-
each ‘‘fault zone” class.
nelling conditions, is represented for each ‘‘fault zone” class.
In Class I most of the tunnel sections (about 80%) is charac-
4.1.4. Daily advance rate, daily AR
terised by PR greater than or equal to 60% of PR max (Fig. 9a). For
The daily advance rate (daily AR) is expressed as meters per
more than 50% of them, PR greater than or equal to 80% of PR max
day and represents the real advancement of the TBM. As written
has been recorded, while about 10% of sections shows a reduction
previ- ously, since it takes into account also downtimes and stops
down to 40% of the best performance. The same considerations
of the machine, it is probably the performance parameter more
already made for the penetration per revolution can be applied
affected by uncertainties. Though, by referring to time-
in this case: the penetration does not represent a big issue in
construction plans of the projects, it has been possible to exclude
highly fractured rocks and good performance can be achieved also
from the analyses holiday terms and TBM-maintenance intervals,
in these conditions. Moreover, RPM shows a negligible reduction
the daily AR remains a function of the utilisation factor (U) of the
in this class, keeping a value close to RPM max for the majority of the
TBM (as expressed in Eq. (3), Table 4) which is affected by several
sections. The high values of p are therefore confirmed by the high
factors (e.g. geological conditions and human component)
values of PR. Regarding the analysis performed with respect to the
extremely diffi- cult to be predicted and evaluated. However, a
mode performance recorded in good conditions, most of the
certain trend can be defined for the daily AR reduction in each
tunnel sections (more than 90%) is characterised by PR greater
‘‘fault zone” class.
than or equal to 60% of PRmode. For more than 60% of them, PR
Fig. 10 shows the decrease of daily AR, with respect to the daily
greater than or equal to 80% of PRmode has been observed, while
ARmax (i.e. the average value between the two best performances
less than 10% of sections is characterised by values comprised
recorded in good tunnelling conditions) and the most frequent
between 40% and 60% of the mode performance. The high values
per- formance (i.e. daily ARmode), for each ‘‘fault zone” class.
of p are again con- firmed by the high values of PR and the p- and
As reported in Table 12, the daily AR has been estimated
PR-reduction trends are almost the same.
consid- ering all the tunnel sections included in the database.
In Class II where more variable values of p have been recorded
Therefore, it can be considered more representative of the
(see Fig. 8b) the variability is observed also for PR (Fig. 9b). Also in
collected data than p, RPM and PR. In order to better describe the
this case a low percentage of tunnel sections (about 20%) exceeds
performance reduction in each class, and thus reach a higher
the 80% of the best performance achieved in good tunnelling con-
degree of detail, smaller intervals have been considered for this
ditions. However, a greater reduction is observed for PR with
analysis.
respect to the reduction observed for p. As a matter of fact, more
In Class I (Fig. 10a) about 30% of tunnel sections is characterised
than 30% of sections shows a significant decrease and about 10%
by a daily AR comprised between 20% and 30% of the best perfor-
of them reaches values lower than 20% of PRmax. This reduction
mance (daily ARmax) while more than 10% of sections shows a sig-
is due to the influence of the cutterhead rotation speed (RPM). In
nificant reduction, down to values lower than 20%. Daily AR
the mode performance analysis, 20% of tunnel sections exceeds
greater than 50% of daily ARmax is recorded for approximately 30%
the 100% of the reference value recorded in good tunnelling
of sec- tions, but it does not exceed 70% of the best performance.
condi- tions but the general reduction trend is very similar to the
In the histogram concerning the mode performance analysis, about
one observed for the best performance.
30% of tunnel sections is characterised by a daily AR comprised
In Class III the tunnel sections are characterised by rather high
between 30% and 40% of daily ARmode while more than 10% of
PR (Fig. 9c). The reduction of RPM observed in this class (see
sec- tions undergoes a significant reduction, reaching values lower
Fig. 7c) affects the percentage of sections interested by the highest
than 20% of the mode performance. Daily AR greater than 50% of
values of PR. For about 30% of sections, values between 80% and
daily ARmode is recorded for approximately 50% of sections, also
100% of PRmax have been recorded, while for p the values even
exceed- ing 70% of the mode performance.
exceed 100% for more than 60% of sections. By considering the
Class II shows a different trend (Fig. 10b), where more than 60%
most frequent PR (i.e. mode PR) recorded in good conditions, 30%
of tunnel sections is characterised by a highly reduced daily AR
of sections is characterised by values equal to (or even greater
(lower than 30% of daily ARmax). About 40% of sections reaches
per- centages lower than 20% of the best performance, while
about 20% has daily AR comprised between 40% and 60% of daily
ARmax. There
56 E. Paltrinieri et al. / Tunnelling and Underground Space Technology 52 (2016) 44–61

Fig. 9. The reduction of the penetration rate PR (m/h) with respect to the best performance PR max (on the left) and the most frequent performance PR mode (on the right) recorded
in the same lithotype in good ground conditions. The broken line represents the average trend of PR in each class. The grey lines represent the average trend of p and RPM. (a)
Class I; (b) Class II; (c) Class III; (d) Class IV.
 E. Paltrinieri et al. / Tunnelling and Underground Space Technology 52 (2016) 44–61
5

Fig. 10. The reduction of the daily advance rate AR (m/h) with respect to the best performance daily ARmax (on the left) and the most frequent performance daily ARmode (on
the right) recorded in the same lithotype in good ground conditions. The broken line represents the average trend of daily AR in each class. (a) Class I; (b) Class II; (c) Class III;
(d) Class IV.
is a very low percentage (about 5%) of tunnel sections where the
can be observed in the last two classes (which correspond to the
daily AR exceeds the 90% of the best performance recorded in
worst tunnelling conditions), while in Classes I and II the values
good conditions. However, this percentage is so low and so distant
are very close to the maximum RPM of the machine (Fig. 11b).
from the mean value of the class, that it is not taken into
However, it is important to remind that RPM is also affected by
consideration for defining the average trend of the daily AR.
the skill of the TBM operator, as well as by the possibility to adapt
Regarding the mode performance, approximately 50% of tunnel
the rotation speed according to the changing ground conditions
sections is charac- terised by values lower than 30% of daily
right on time. Thus, for what concerns PR, the performance
ARmode. About 25% of sec- tions reaches percentages lower than
decreasing trend with decreasing ground conditions is more evi-
20% of the mode performance, while about 20% shows daily AR
dent (Fig. 11c), mainly due to the influence of the cutterhead rota-
comprised between 30% and 40% of daily ARmode. More than 10%
tion speed (RPM) of the machine.
of tunnel sections is characterised by daily AR comprised between
Finally, the daily advance rate (daily AR) is the parameter
60% and 70% of the mode performance and, also in this case,
which shows the most evident reduction. The decreasing trend
about 5% of sections shows values greater than 90% of the
could be observed from the first to the fourth ‘‘fault zone” class
reference performance in good rock conditions.
(Fig. 11d). A quite important reduction of the daily AR (down to
In Class III (faulted rocks), the reduction of the performance is
40% of the best performance) can be observed with just an
still evident but differently spread with respect to the first two
increasing degree of fracturing (Class I). Then, the difference
classes (Fig. 10c). A specific trend is not visible because the per-
between the reduction percentages becomes smaller from one
centages are quite evenly distributed among the tunnel sections.
class to the next one. The minimum is clearly reached with Class IV
The daily AR does not exceed 50% of the best performance and
(representing the worst excavation conditions) where the majority
more than 60% of sections is characterised by values lower than
of tunnel sections is characterised by daily AR lower than 10% of
30% of daily ARmax. The lowest values (lower than 10% of the best
the best reference performance (see also Fig. 10d). Despite of this
performance) have been recorded for about 30% of sections.
result, however, it is important to underline that the advance rate
Unlike what has been observed for the best performance, a
of the TBM strongly depends on the utilisation factor U for which it
decreasing trend is visible in the reduction with respect to the
is extremely difficult to establish a direct correlation with the
daily ARmode, up to reach values comprised between 60% and 70%
geological/- geotechnical conditions of the rock mass, since it is
of the mode value recorded in good tunnelling conditions. Also in
affected by several other factors, including experience of the crew,
this case, the lowest values (lower than 10% of the daily AR mode)
contractual decisions, working conditions, etc. and thus it can be
have been recorded for about 30% of tunnel sections.
only esti- mated with a certain uncertainty degree (Einstein,
The most significant reduction has been recorded in Class IV
1996).
where the daily AR proves to be lower than 40% of daily AR max in
Some considerations can be also done by considering how the
each tunnel section (Fig. 10d). However, unlike the faulted rocks,
tunnel size can affect the performance of a (gripper) TBM espe-
the vast majority of the sections is characterised by values lower
cially in difficult ground conditions, such as highly fractured and
than 10% of the best performance (more than 45% of sections).
faulted rock masses. As it has been observed by Barton (2000),
About 90% of tunnel sections shows values that do not exceed
the overbreak is a function of the tunnel diameter as well as the
30% of daily ARmax. Referring to the mode value, the daily AR
quantities of support to be installed. These problems also mean
proves to be lower than 50% of daily AR mode in each tunnel section.
higher delays, thus slower TBM advancement. Since the projects
A quite important number of sections (i.e. more than 40% of the
analysed in this study are characterised by different tunnel diam-
total) is characterised by values lower than 10% of the mode per-
eters, in order to compare the results, the recorded performances
formance. About 90% of tunnel sections shows values that do not
have been normalised with respect to the maximum possible value
exceed 30% of daily ARmode.
(i.e. RPMmax) or with respect to the best performance observed for
The reduction trends of the daily AR obtained by referring to the
each project in good tunnelling conditions (i.e. pmax, PRmax and
mode performance recorded in good tunnelling conditions are very
daily ARmax). As already mentioned (see Section 2.2), it is important
similar to the ones observed in the analysis performed respect to
to underline that some performance parameters (i.e. RPM, p and
the best performance. Only in Class I a shift to the right of the peak
PR) were available only for two projects (i.e. the Lö tschberg Base
occurs (as already said, mainly due to the fact that daily AR mode < -
Tunnel and the Second Manapouri Tailrace Tunnel) which are
daily ARmax), while in Class III a more definite trend could be iden-
char- acterised by similar sizes (i.e. diameter respectively of 9.4
tified (in the best performance analysis the reduction percentages
m and
are more equally distributed among the sections).
10.1 m). On the other hand, for what concerns the advance rate
of the TBM, also tunnel sections with smaller diameter have been
4.2. Results discussion
considered. As it can be observed in Fig. 12, the reduction of the
daily AR shows the same trend (within similar reduction ranges)
The results allow affirming that the TBM rate of penetration
from Class I to Class IV for the different tunnel sizes analysed in
(expressed as p or PR) does not seem to be the major issue in
this work.
highly fractured and faulted rock masses. The penetration per
By referring to the developed ‘‘fault zone” classification and to
revolution
the results obtained from the TBM-performance reduction analy-
(p) does not undergo a significant reduction with respect to the
ses, the following important observations can be pointed out:
good tunnelling conditions, quite the contrary, an increase has
been even observed in Classes III and IV (Fig. 11a). As already
● For characterising highly fractured and fault zones, it is neces-
men- tioned, the high fracturing degree, together with the strength
sary to take into account the mechanical and chemical weather-
reduction undergone by the rock material (especially for Class III
ing of the rock mass. This can be done by introducing specific
and Class IV), seem facilitating the boring process rather than hin-
geological and geotechnical parameters which allow a more
dering it. The RPM seems to be more affected by the worsening of
effective description and characterisation of the behaviour of
the ground conditions. This is certainly connected to the different
the rock mass. In this study, the new categories of GSI (better
scale considered for the analyses. In particular, at the tool scale
representing the altered state of the rock subject to folding
(i.e. p) the face stability issues do not represent a hindrance to
and faulting processes) together with significant fracturing
cope with as they are when the size of the TBM is introduced at the
degrees (expressed by high Jv, up to values greater than
tun- nel diameter scale (i.e. RPM). As a matter of fact, a certain
30 joints/m3) and the reduced strength of the rock UCSH have
decrease
been considered.
Fig. 11. The average reduction of the TBM parameters, in each ‘‘fault zone” class, with respect to the performances recorded in good tunnelling conditions. (a) The average
reduction of the penetration per revolution (p) with respect to the best (pmax) and the most frequent performance (pmode); (b) the average reduction of the cutterhead rotation
speed (RPM) with respect to the maximum RPM of the TBM (RPM max); (c) the average reduction of the penetration rate (PR) with respect to the best (PR max) and the most
frequent performance (PR mode); (d) the average reduction of the daily advance rate (daily AR) with respect to the best (daily AR max) and the most frequent performance (daily
ARmode).

Table 13
The rates of reduction for each ‘‘fault zone” class to be applied to the TBM-
performance parameters evaluated by existing performance prediction models in
ordinary tunnelling conditions. The most frequent reduction range (MFR) has been
also reported for each class.

TBM parameter Rate of reduction (to be applied to the best

performance: RPMmax, PRmax)

Class I Class II Class III Class IV

RPM [rev/min] 80–100% 60–100% 40–100% 40–100%

MFR MFR MFR MFR
80–100% 80–100% 40–60% 40–
60%

PR [m/h] 40–100% 20–100% 20–100% 20–100%

MFR MFR MFR MFR 80–
Fig. 12. The average daily AR reduction in each ‘‘fault zone” class, for different
100% 60–80% 60–80% 20–40%
tunnel diameters, with respect to the best performance recorded in good tunnelling
conditions (daily ARmax). Grey line: 9.4–10.1 m (i.e. Lotschberg Base Tunnel and The
Second Manapouri Tailrace Tunnel); black line: 5.8 m (i.e. Cleuson-Dixence).

For what concerns the evaluation of the daily AR in difficult

● The data compiled in the database have been grouped in four ground conditions, the reduction rates obtained for each ‘‘fault
‘‘fault zone” classes. For each class a specific reduction of TBM zone” class (Table 14) require further investigations. As a matter
performance has been estimated by taking into account the of fact, the analysed data have been collected from different tunnel
best and the mode TBM performances characterising the projects and in order to create a model able to predict the TBM
excavation in good tunnelling conditions. The reduction rates advancement, the delays associated to the construction phases
applicable to the TBM-performance parameters (i.e. RPM and (e.g. rock supporting, ground treatment and drainage systems, cut-
PR) in each class are summarised in Table 13. Since the ter replacement, muck transportation, etc.) should be in the same
reduction observed for each class is similar if considering the magnitude for each project. In this study the available information
best performance or the mode (the difference is mainly related did not allow to carry out a detailed analysis of times related to the
to the gap between the best performance and the relative mode excavation activities. However, it is important to underline that the
performance), the table includes only the decrease in the four tunnel sections have been selected on the basis of specific criteria
classes with respect to the best performance. Moreover, in and, according to them, they are characterised by very similar con-
order to better distin- guish the influence of each class on the ditions for what concerns the rock mass (e.g. water inflows, rock
TBM performance, the most frequent reduction (MFR) is also weathering, etc.) and the required supports.
reported.
Table 14
The rates of reduction for each ‘‘fault zone” class to be applied to the TBM daily tured weathered rock mass; Class III – cohesive fault rocks and
advance rate evaluated by existing performance prediction models in ordinary heterogeneous rock mass; Class IV – crushed fault rocks. In each
tunnelling conditions. The most frequent reduction range (MFR) has been also class, for the penetration per revolution, p (mm/rev), the cutter-
reported for each class.
head rotation speed RPM (rev/min), the penetration rate PR (m/
TBM parameter Rate of reduction (to be applied to the best h) and the daily advance rate AR (m/day), a reduction rate with
performance: Daily ARmax) respect to the best and the most frequent performance recorded
Class I Class II Class III Class IV in good rocks has been estimated. The results show how the rate
Daily AR [m/day] 10–70% 10–60% 10–50% 10–40% of penetration of the TBM seems not to be affected at the tool-
MFR MFR MFR MFR scale (i.e. p) where the face stability problems do not represent a
20–30% 10–20% <10% <10% hindrance to cope with. On the contrary, when the tunnel diameter
scale is considered (i.e. PR), thus including issues related to face
instability, the performance is reduced and a decreasing trend
Though the obtained results are linked to the quality of the data can be observed with the worsening of ground conditions. The
collected in the database, and thus prone to be improved by intro- same trend is even clearer by analysing the daily advance rate
ducing more case studies, it is important to underline that the out- AR; however, a more detailed time analysis of all construction
comes of this analysis provide useful insights and give a first activities remains necessary for this latter parameter.
qualitative overview on the reduction of the TBM performance in Although the limitations due to the accessibility of TBM data as
difficult ground conditions such as highly jointed and fault zones. well as geological information for projects crossing unexpected
weakness zones, the results of this study allow obtaining a better
5. Summary and conclusions evaluation of the TBM behaviour in challenging advancing condi-
tions, such as highly jointed and faulted rocks. In particular, by
Despite the great advantages in good rocks, TBM operations starting from a better characterisation of the ‘‘fault zone” environ-
may be seriously affected by difficult ground conditions, such as ments, the proposed study allows estimating the TBM
highly fractured and fault zones, with a consequent significant performance reduction while excavating in such difficult ground
reduction of the TBM performance with respect to the expected conditions.
ones (i.e. good rock conditions).
In order to better analyse the TBM performance in difficult Acknowledgements
grounds and investigate the relationships existing between
machine performance parameters and geological/geotechnical The research is financed by ASTRA (Bundesamt fü r Strassen),
parameters of the rock mass, a TBM-performance database has the Swiss Federal Roads Authority. The authors would like to
been created. The data compiled have been selected from different thank BG SA for having provided TBM data as well as all the
tunnelling projects where highly jointed rock masses and/or fault information about tunnel design and construction. They would
zones were encountered. Though sometimes the contractors were also like to thank the Swiss National Geological Service and
unable to share easily the data from the site (e.g. on-going claims Kallerhals+Haefeli AG for having provided geological/geotechnical
not yet resolved), for the majority of the cases it has been possible data. Special thanks to Dr. Jafar Hassanpour for his help during
to estimate TBM-performance parameters on the basis of the mea- data collection.
sures carried out by the TBM on-board acquisition system during
construction. Although the degree of detail of data varies from
References
one project to another, enough information has been collected
for analysing the most important parameters related to the boring Afrouz, A.A., 1992. Practical Handbook of Rock Mass Classification Systems and
process and thus affecting the TBM advancement. Modes of Ground Failure. CRC Press, Boca Raton, FL.
With the aim to identify the correlations between the perfor- Barla, G., Pelizza, S., 2000. TBM tunnelling in difficult ground conditions. In: Paper
Presented At the GeoEng2000 – An International Conference on Geotechnical
mance parameters of the machine and the geological/geotechnical and Geological Engineering, Melbourne, Australia.
characteristics, preliminary analyses have been carried out based Barton, N., 2000. TBM Tunneling in Jointed and Fault Rock. Balkema, Rotterdam.
on the information collected in the TBM-performance database. Barton, N., Lien, R., Lunde, J., 1974. Engineering classification of rock masses for the
design of tunnel support. Rock Mech. 6 (4), 189–239.
The first results confirm the difficulties in selecting within existing Bieniawski, Z.T., 1976. Rock mass classification in rock engineering. In: Bieniawski,
prediction models, the parameters affecting the most the TBM per- Z.T. (Ed.), Paper Presented At the Exploration for Rock Engineering, Proc. Symp.,
formance while degrading the ground conditions. As a matter of Cape Town.
Bruland, A., 1998. Hard Rock Tunnel Boring. PhD, Norwegian University of Science
fact, in highly fractured rocks and fault zones, the delays and Technology, Trondheim.
associated to the increasing need of heavy temporary supports, Bü rgi, C., 1999. Cataclastic Fault Rocks in Underground Excavations: A Geological
additional drainage system, further ground treatments, etc., are Characterisation. PhD, EPFL.
Christe, P., 2009. Geological Characterization of Cataclastic Rock Samples Using
partially bal- anced by the rather high TBM penetration rates
Medical X-ray Computerized Tomography: Towards a Better Geotechnical
generally observed in fractured/very fractured rock masses. Description. PhD, EPFL.
In order to consider the complex structure of fault zones, and Delisio, A., Zhao, J., Einstein, H.H., 2013. Analysis and prediction of TBM
bearing in mind that a more global approach is required for char- performance in blocky rock conditions at the Lö tschberg Base Tunnel.
Tunnell. Undergr. Space Technol. 33, 131–142.
acterising their behaviour (i.e. combination of factors instead of Delisio, A., Zhao, J., 2014. A new model for TBM performance prediction in blocky
specific parameters approach), the next step has been the develop- rock conditions. Tunnell. Undergr. Space Technol. 43, 440–452.
ment of a classification system of these particular environments. In Einstein, H.H., 1996. Risk and risk analysis in rock engineering. Tunnell. Undergr.
Space Technol. 11 (2), 141–155.
order to identify the specific classes describing highly fractured Geological Society Engineering Group Working Party, 1995. The description and
and faulted rocks, two main factors have been taken into account: classification of weathered rock for engineering purposes. Quart. J. Eng. Geol. 28
the fracturing and the weathering degree of the rock (and of the (3), 207–242.
Habimana, J., 1999. Caracté risation gé omé canique de roches cataclastiques
joints surfaces). The choice of specific characteristics has been
rencontré es dans des ouvrages souterrains alpins. PhD, EPFL.
done on the basis of most common parameters used in the Habimana, J., Labiouse, V., Descoeudres, F., 2002. Geomechanical characterisation of
literature for the description of fault (weak) rocks. Four classes cataclastic rocks: experience from the Cleuson-Dixence project. Int. J. Rock
have been defined: Class I – highly fractured rock mass; Class II – Mech. Min. Sci. 39, 677–693.
Hassanpour, J., Rostami, J., Zhao, J., 2011. A new hard rock TBM performance
highly frac- prediction model for project planning. Tunnell. Undergr. Space Technol. 26 (5),
595–603.
Hoek, E., 1983. Strength of jointed rock masses. Gé otechnique 23 (3), 187–223.
Hoek, E., Brown, E.T., 1988. The Hoek–Brown failure criterion – a 1988 update. In:
Paper Presented At the 15th Canadian Rock Mech. Symp., Toronto. Paltrinieri, E., Sandrone, F., 2014. A study of the TBM performance in fault zones and
Hoek, E., Carranza-Torres, C., Corkum, B., 2002. Hoek–Brown criterion – 2002 highly fractured rocks. In: Paper Presented At the World Tunnel Congress,
edition. In: Paper Presented At the 5th North American Rock Mechanics Symp., Iguassu Falls, Brazil.
Toronto. Ramoni, M., Anagnostou, G., 2010. Tunnel boring machines under squeezing
Hoek, E., Marinos, P., Bennissi, M., 1998. Applicability of the Geological Strength conditions. Tunnell. Undergr. Space Technol. 25 (2), 139–157.
Index (GSI) classification for the very weak and sheared rock masses. The case of Rostami, J., 2008. Hard rock TBM cutterhead modeling for design and performance
the Athens schist formation. Bull. Eng. Geol. Environ. 57, 151–160. prediction. Geomechanik und Tunnelbau (Geomech. Tunnell.) 1 (1), 18–28.
Hudson, J.A., Priest, S.D., 1979. Discontinuities and rock mass geometry. Int. J. Santi, P.M., 1995. Classification and Testing of Weak and Weathered Rock Materials:
Rock Mech. Min. Sci. Geomech. Abstr. 16, 339–362. A Model Based on Colorado Shales. PhD, Colorado School of Mines, Golden CO.
ISRM, 2014. Ulusay, R. (Ed.). The ISRM Suggested Methods for Rock Santi, P.M., Doyle, B.C., 1997. The Locations and Engineering Characteristics of Weak
Characterization, Testing and Monitoring: 2007–2014 (The Orange Book). Rock in the U.S. Denver, CO: Association of Engineering Geologists, Special
Springer. Publication #9.
Klein, S., 2001. An approach to the classification of weak rocks for tunnel projects. Santi, P.M., 2006. Field methods for characterizing weak rock for engineering.
In: Paper Presented At the Proc. Rapid Excavation and Tunn. Conf., San Diego, Environ. Eng. Geosci. 12 (1), 1–11.
CA, USA. Sapigni, M., Berti, M., Bethaz, E., Busillo, A., Cardone, G., 2002. TBM performance
Klein, S., Schmoll, M., Avery, T., 1995. TBM performance at four hard rock tunnels in estimation using rock mass classifications. Int. J. Rock Mech. Min. Sci. 39 (6), 771–
California. In: Proceedings of the Rapid Excavation and Tunnelling Conference. 788.
Loew, S., Barla, G., Diederichs, M., 2010. Engineering geology of Alpine tunnels: past, Sausgruber, T., Bradner, R., 2003. The relevance of brittle fault zones in tunnel
present and future. In: Paper Presented At the 11th IAEG Congress, Auckland. construction – lower inn valley feeder line north of Brenner Base tunnel, Tyrol,
Marinos, P., Hoek, E., 2001. Estimating the geotechnical properties of heterogeneous Austria. Mitt. Osterr. Geol. Ges. 94, 157–172.
rock masses such as flysch. Bull. Eng. Geol. Environ. (IAEG) 60, 85–92. Schubert, W., Riedmü ller, G., 2000. Tunnelling in fault zones-state of the art in
Marinos, P.G., Marinos, V., Hoek, E., 2007. The geological strength index (GSI): a investigation and construction. Felsbau 18 (2), 7–15.
characterization tool for assessing engineering properties for rock masses. In: Schubert, W., Riedmü ller, G., 1997. Influence of faults on tunnelling. Felsbau 15 (6),
Paper Presented At the International Workshop on Rock Mass Classification in 483–488.
Underground Mining. Thuro, K., Scholz, M., 2003. Deep weathering and alteration in granites – a product
Palmströ m, A., Berthelsen, O., 1988. The significance of weakness zones in rock of coupled processes. In: Paper Presented At the International Conference on
tunnelling. In: Paper Presented At the International Conference on Rock Coupled T-H-M-C in Geosystem: Fundamentals, Modelling, Experiments and
Mechanics and Power Plants, Madrid, Spain. Application, GeoProc 2003, Stockholm, Royal Institute of Technology (KTH).