ARTICLE IN PRESS

International Journal of Rock Mechanics & Mining Sciences 46 (2009) 8–18

Contents lists available at ScienceDirect

International Journal of
Rock Mechanics & Mining Sciences
journal homepage: www.elsevier.com/locate/ijrmms

Development of a rock mass characteristics model for TBM penetration

rate prediction
Q.M. Gong a,, J. Zhao b
a
College of Architecture and Civil Engineering, Beijing University of Technology, Beijing 100022, China
b
Ecole Polytechnique Federale de Lausanne (EPFL), Rock Mechanics Laboratory, CH-1015 Lausanne, Switzerland

a r t i c l e in f o a b s t r a c t

Article history: The TBM tunneling process in hard rock is actually a rock or rock mass breakage process, which
Received 21 June 2007 determines the efficiency of tunnel boring machine (TBM). On the basis of the rock breakage process, a
Received in revised form rock mass conceptual model that identifies the effect of rock mass properties on TBM penetration rate is
7 March 2008
proposed. During the construction of T05 and T06 tunnels of DTSS project in Singapore, a
Accepted 10 March 2008
comprehensive program was performed to obtain the relevant rock mass properties and TBM
Available online 29 April 2008
performance data. A database, including rock mass properties, TBM specifications and the correspond-
Keywords: ing TBM performance, was established. Combining the rock mass conceptual model for evaluating rock
Rock breakage process mass boreability with the established database, a statistical prediction model of TBM penetration rate is
Rock mass conceptual model
set up by performing a nonlinear regression analysis. The parametric studies of the new model showed
Rock mass boreability
that the rock uniaxial compressive strength and the volumetric joint count have predominantly effects
Database
TBM penetration rate on the penetration rate. These results showed good agreement with the numerical simulations. The
model limitations were also discussed.
& 2008 Elsevier Ltd. All rights reserved.

1. Introduction Table 1. The machine factors applied to predict the PR include

Because the planning of tunnel projects and selection of
construction methods require effective prediction of tunnel boring
machine (TBM) performance, the prediction of TBM performance
in a given rock mass has been a longstanding research topic. The
complexity of the interaction between rock mass and TBM makes
TBM performance prediction very difficult theoretically. During
the last 30 years, the prediction models have been developed from
a single factor model [1–5] to multiple factors model [6–11], as
summarized in Table 1. Among these models, some are based on
laboratory cutting tests, and some based on practical tunnel
construction data. Generally speaking, laboratory cutting tests on
rocks are necessary, but they neither give a full picture of rock
mass nor represent the full-scale operation of the TBM. However,
the empirical and statistical models scarcely take the rock mass
breakage mechanisms into account. A realistic prediction model
for TBM performance should combine the laboratory test data,
site geological and construction conditions.
The main factors of rock mass properties used to predict the
penetration rate (PR) in these models include the compressive
strength and tensile strength of the rock material, and
the frequency and orientation of the rock joints, as shown in
 Corresponding author. Tel.: +86 10 67396357; fax: 86 10 67391645.
E-mail address: gongqiuming@bjut.edu.cn (Q.M. Gong).
cutter spacing, cutter tip width, cutter diameter, average thrust
per cutter and revolution per minute (RPM). At present,
the extensively used models include the CSM model [9], the
NTNU model [7], probabilistic model [8] and QTBM [10].
These models, of course, have their advantages and disadvantages
because of their origin and background. The CSM model does
not originally consider the influence of joints on TBM PR. The
modified CSM model only regards joints as a factor that affects
the rock mass strength. The NTNU model requires special
experiments originated from the drilling. These tests are not
commonly available outside Norway. The probabilistic model
only takes into account the similarity of tunnels and neglects
the interaction between TBM and rock mass. QTBM originates from
Q system and includes too many parameters for practical
application. In addition, some parameters are overlapped in this
model.
Some authors also correlated TBM performance to rock mass
classification systems [12–15]. In these studies, the rock mass
classification systems used to characterize rock mass quality
mainly include RSR (rock structure rating), RMR, Q system and
IMS (integrated mass system). The results show that with the
increase of rock mass quality, the PR generally decreases.
However, very poor rock mass does not facilitate the PR. Although
the tendency is obvious, the TBM PR distributes in a large range
corresponding to the rock mass classes.

1365-1609/$ - see front matter & 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijrmms.2008.03.003
 ARTICLE IN PRESS

Q.M. Gong, J. Zhao / International Journal of Rock Mechanics & Mining Sciences 46 (2009) 8–18 9

Table 1
Penetration rate prediction models and the corresponding factors.

Prediction value Reference Rock mass factors Machine factors and tunnel geometry

PRev [2] st Fn
PRev [1] sc Fn
PRev [4] sc Fn, cutter diameter
Rf (kN/cutter/mm/rev) [3,5] HT (total hardness) Fn
PRev CSM [6] sc, st Cutter spacing, cutter tip width, cutter radius, Fn
PRev Modified CSM [9] st, RMBI (elasticity modulus, size reduction factor Cutter spacing, cutter tip width, cutter radius, Fn
and possion’s ratio)
PR NTNU model [7] Drilling rate index (DRI), fracture frequency and Fn, cutterhead RPM, cutter spacing, cutter size and
orientation, porosity shape, installed cutterhead power
PR, AR, U Probabilistic model [8] The regional geological structure, rock type, quartz Maximum thrust, torque, cutterhead RPM, disc
content, intact rock strength, the tunnel water cutter number and diameter, TBM diameter, type
conditions (level 2) of mucking system, tunnel length and gradient
PR, AR QTBM [10] RQD0, Jn, Jr, Ja, Jw, SRF, the rock mass strength, CLI, Average cutter load, TBM diameter
quartz content, induced biaxial stress, porosity
PR Neuro-fuzzy model [11] The core fracture frequency, UCS RPM, normal force per cutter, cutter diameter

In order to develop a rock mass characteristics prediction Normal Force

model for TBM PR, this paper reviews the rock fragmentation
Rolling Force
process, and thereby reveals the rock breakage mechanisms and
the main properties influencing the PR. Furthermore, the
existence of joints in rock mass changes the rock breakage process
and chipping pattern. On the basis of the rock and rock mass Cutter Spacing
fragmentation mechanism, a conceptual model of rock mass Side Force
properties affecting TBM PR is proposed. In order to validate the
conceptual model, a database is established by a comprehensive
program during the construction of deep tunnel sewerage system
(DTSS) project in Singapore. Then, the prediction model is set up
Penetration
by performing a nonlinear statistical analysis. The effect of each
rock mass parameter on TBM PR is analyzed. The limitations of
this new model are identified.
Fig. 1. Rock chipping by TBM cutters and kerf cutting geometry.

2. Rock breakage mechanism

2.1. Rock breakage process by TBM cutter
The rock breakage process under TBM rolling cutters can be
divided into two continuous stages. The first is that the rolling
cutter intrudes into the rock, and then generates the small and
large fragments as well as internal cracks. This process is termed
as indentation. The second stage is that cracks between two
adjacent cutters propagate to each other and coalesce, and then
large chips are formed between two cutters.
The loaded process of rock indentation can be divided into four
stages based on the experiment [16], namely, building up of a
stress field, formation of a crushed zone, formation of subsurface
and surface chipping, and formation of crater. The indenter shape,
size and the confinement force affect the formation of the crushed
zone and the internal cracks initiation and propagation [17,18].
The indentation tests and numerical modeling results showed
beneath the indenter three different zones can be distinguished
[16,19,20], namely, the crashed zone, the crack zone, the elastic
zone. The side cracks developed in the crack zone decide rock
breakage efficiency of TBM. In a multi-pass cutting process, the
penetration process is composed of successive indentation cycles.
Each of indentation cycle consists of a crushing and chipping
phase [21,22]. In the process of chipping formation, the applied
force is generally not reduced to zero, because not all the rock
fragments are expelled from the region around the indenter.
For the tunnel excavation, when TBM cutters roll across the
tunnel face, they continuously expand the crushed zone beneath
themselves, and cracks are initiated and propagated, as shown in
Fig. 1. One or more cracks under the action of the rolling cutter
may reach the free surface or propagate to meet the cracks of the
neighboring cuts. In these two cases, chipping occurs. The first
case is similar to the chip formation of single indentation process.
The latter is decided by the interaction between two adjacent cuts,
which is directly relevant to the efficiency of TBM excavation.
In practice, two possible mechanisms are applicable for rock
fragmentation under rolling cutters. The first is the single pass
cutting process, and the second is the multiple pass cutting
process [23]. The multiple cutting process occurs due to the
following two reasons. One is that the cutter load is not enough to
crush the rock and initiate and propagate the cracks. The other is
that due to the rotation of the rolling cutter, the available time to
create a stress field sufficient for crack initiation and propagation
is limited [24]. The single pass cutting mode almost occurs in all
TBM excavated tunnels except those with very high rock strength.
2.2. Influence of machine specifications on rock breakage process
By utilizing a linear rock cutting rig, Snowdon et al. [25]
performed a series of linear cutting tests. The results showed that
the cutter spacing and cutter force strongly affect the specific
energy of cutting, namely the cutting efficiency. The influence of
cutting geometry on rock chip formation in linear cutting tests
was identified by Rostami and Ozdemir [26]. The cutter tip
geometry influences the size of the crushed zone immediately
under the cutter. The cutter spacing and cutter force affect the
induced cracks propagation and rock chip formation between
adjacent cuts. On the basis of the numerical simulations on the
cutter indentation process and rock chipping between adjacent
cutters, Gong and Zhao [27] concluded that from the rock
 ARTICLE IN PRESS
10 Q.M. Gong, J. Zhao / International Journal of Rock Mechanics & Mining Sciences 46 (2009) 8–18
breakage process, the thrust force, the cutter diameter and cutter
tip width affect the stress field in the rock, and then the indention
process. The thrust force and cutter line spacing affect the
chipping process which includes the crushed zone, cracks
initiation and propagation and chipping angle.
Generally, the rock breakage process is closely related to the
machine parameters, such as TBM diameter, cutter line spacing,
cutter diameter and tip width, total thrust and torque. Particularly,
the design constants of cutter spacing and cutter diameter and tip
width directly affect the rock breakage process, which should be
regarded as the important machine parameters affecting the PR.
The thrust and torque usually changes with variation of rock mass
conditions and are regarded as TBM operational parameters that
affect the PR.
3. Rock mass conceptual model
On the basis of the above-mentioned rock breakage process,
the influence of main rock mass properties on TBM PR is identified
here. A rock mass conceptual model that stands for rock mass
boreability is proposed.
3.1. Rock strength
The rock strength affects the rock behavior under compression.
When the rolling cutter indents the rock, the stress exerted must
be higher than the rock strength. Thus, the rock strength is
directly relevant to the performance of TBM. This fact has been
observed by many researchers. Some models for predicting PR
show that the PR is directly relevant to rock uniaxial compressive
strength [1,2,5,26]. Generally, the PR decreases with the increase
of rock uniaxial compressive strength.
3.2. Rock brittleness
The effect of rock brittleness index on TBM penetration process
was studied by Gong and Zhao [27]. The brittleness index is
defined as the ratio of rock compressive strength to tensile
strength. The result shows that with increasing rock brittleness
index the cutter indentation process gets easier. Generally, the PR
increases with increasing rock brittleness index.
3.3. Joint spacing
Because rock masses are composed of rock material and joints,
the existing joint conditions certainly affect the rock breakage
process. It is easy to be understood that discontinuities can
facilitate rock breakage, because cracks induced by TBM cutters
easily develop along with the existing discontinuities. In practice,
it has been well recognized that joints or fractures have an
important effect on the TBM performance [7,9,28,29]. On the basis
of a large number of case histories, Bruland [7] concluded that
with the decrease of joint spacing, the TBM penetration increases
distinctly.
Rock mass fragmentation processes at different joint spacings
were simulated by using UDEC [20]. For joints with different
spacings, two patterns of chipping process are found, which is
different from that in rock material. One is that the crack initiates
from the joint plane and propagates toward to the free surface.
The other is that the crack initiates from the crushed zone and
propagates towards to the joint plane. With the increase of the
joint spacing, the PR decreases. The maximum PR in a highly
jointed rock mass is more than nine times of that in rock material.
3.4. Joint orientation
The influence of the joint orientation on TBM PR was widely
observed in the tunneling projects [28]. Aeberli and Wanner [30]
observed that the advance rate of TBM increases with the increase
of the angle between TBM axis and the planes of schistosity in a
homogeneous zone of schistose phyllite. Similar phenomena were
also observed by Thuro and Plinninger [31] in phyllite and
phyllite-carbonate-schist inter-stratification. A theoretical analy-
sis of the interaction between cutter and rock with foliation by
Sanio [32] showed a similar tendency. Bruland [7] summarized
the effects of joint orientation of different classes of joints. The
similar rule was obtained. However, the maximum PR occurred
when the angle was equal to 601. He also noted that with the
increase of joint spacing, the effect of joint orientation on TBM
penetration decreases. The rock chipping process induced by the
TBM cutter is simulated by DEM modeling to examine the effect of
joint orientation [33]. The results show good agreement with
Bruland’s results. In the recent models, the significance of joint
spacing and orientation on TBM performance are emphasized and
regarded as important factors influencing the TBM performance
[7,9,10].
On the basis of the above analysis, the main rock mass
parameters influencing the TBM PR are rock compressive strength,
rock brittleness index, joint spacing and joint orientation. The four
parameters of rock mass will be adopted to analyze the rock mass
boreability.
4. Database
4.1. Contracts T05 and T06 of DTSS project in Singapore
The Deep Tunnel Sewerage System (DTSS) project is a great
infrastructure project as the long-term solution to meet the
island’s needs in wastewater conveyance, treatment and disposal
in Singapore. In DTSS project, only contracts T05 and T06 that go
through hard rock will be studied. Contract T05 is 12.6 km in
length, with an internal diameter of 3.6 m. The Contract T06 is
9.6 km in length with a finished diameter of 3.3 m. Depth of this
tunnel varies from 22 to 41 m. Both the geophysical method and
boreholes were adopted for site investigation. A total number of
49 and 43 boreholes were, respectively, drilled to provide the site
geological conditions in T05 and T06 besides the geotechnical data
obtained from the vicinal project sites. These boreholes are
approximately evenly distributed along the tunnel alignment.
The T05 tunnel only goes through Bukit Timah granite with
different weathering grades and its residual soil. The predominant
ground is fresh to slightly weathered granite. It covers 47% of the
total tunnel length. Approximately, 3.2 km is in completely
weathered granite to residual soil, 1.9 km in mixed face ground
and 1.6 km in moderately to highly weathered granite. Along the
whole 9.6 km length of the T06 tunnel alignment, 3.3 km is in
moderately fractured and slightly weathered to fresh granite or
highly fractured and highly to moderately weathered granite,
3.1 km in completely weathered granite to residual soil, and 1.1 km
is located in mixed face ground. About 0.9 km of the last southern
section of the tunnel is located in multilayer sedimentary rock.
1.1 km of the last northern section of the tunnel is located in the
weathered to fresh Old Alluvium.
The mineralogical content of granite and diorite was analyzed
based on the core sample for each rock type. The results are
summarized in Table 2. The granite content is consistent with the
previous experiment results [34]. The Cerchar abrasivity index
(CAI) of the granite is summarized in Table 3 based on the rock
core samples. All the results show that the CAI is more than 4.0,
 ARTICLE IN PRESS

Q.M. Gong, J. Zhao / International Journal of Rock Mechanics & Mining Sciences 46 (2009) 8–18 11

and the rock is classified as extremely abrasive in accordance to
the Cerchar definition [35]. The rock samples were also collected
for the strength tests. However, the rock strength varies in a large
range for different weathering grades, as shown in Table 4. The
results are not consistent with the general granite weathering
classification in Singapore [36]. It shows that the rock classifica-
tion at the site is too rough or the rock may fail along weak planes
in the rock strength test.
The site investigation provided an overall image of the ground
condition along the tunnel alignment. The detailed rock mass
properties cannot be interpreted from the geotechnical investiga-
tion reports due to the nature of granite weathering in the tropical
climate. The rock head is generally a few meters below the ground
surface. The joint system developed in the rock mass cannot be
observed in the outcrops. In order to establish a detailed database
for the PR prediction, and study the effect of rock mass properties
on the TBM PR, a comprehensive program for obtaining the joint
properties and rock strength parameters would be performed
during the tunnel construction.
TBMs used in T05 and T06 are four hard rock type shield
machines manufactured by Herrenknecht AG. They can be
operated in earth pressure balanced (EPB) mode. TBMs are
equipped with hard rock single and double cutters and soft
ground drag bits. The TBM outer diameter, total cutter number
and cutter line spacing of the TBMs in T05 and T06 is 4880, 4450,
35, 29, 100, and 90 mm, respectively. The maximum designed
cutter thrust is 250 kN/cutter. The shield comprises of three main
sections, namely front shield, center and airlock shield, and tail
shield. The propulsion cylinders with a stroke of 2300 mm can be
operated individually or in groups. All important operation
Table 2
Mineralogy of the Bukit Timah granite.
Mineral type Diorite (dark gray fine Granite (light gray coarse grained)
grained) (%) (%)
Feldspar 57.1 (plagioclase) 57.8–64.5 (plagioclase+k-feldspar)
Amphibole (green 22.4 0.6–4.1
hornblende)
Quartz 10.3 26.4–29.2
Mica (biotite and 8.8 5.3–7.0
some chlotite)
Calcite (mostly vein) 0.2–6.4
Others 1.4 0.2–5.0
parameters of the TBM can be monitored by an acquisition and
recording system. Some of the major parameters recorded are:
(a) face pressure, (b) thrust force, (c) steering force, (d) backup
towing loads, (e) revolution per minute, (f) advance speed, (g)
time for advance, (h) volume of excavated material via belt
weighter, (i) time for segment placing and standby/stoppage, and
(j) some other system operating data.
4.2. Field data collection program
During the excavation of T05 and T06 tunnels, an extensive
field and laboratory test program was performed to develop a
database of rock mass properties and TBM performance in various
rock mass conditions. The site work includes the tunnel face
mapping for joint spacing and orientation, rock coring in the
tunnel face for rock strength tests in the laboratory, collection of
TBM data.
4.2.1. Rock face mapping
Due to intensive weathering of the bedrocks, rock exposures
are hardly observed except at the rock quarries. The statistics of
the joint system (including all discontinuities, such as joints,
fissures, factures and so on) in the rock mass cannot be conducted
in the ground surface. Thus, the tunnel face mapping was
conducted during tunnel excavation.
As full observations of the rock mass conditions were not
possible in the tunnel wall, the tunnel face mapping was only
carried out in the TBM cutterhead chamber. During the cutter
inspection and cutter replacement, the cutterhead chamber is
opened. It gives a chance to survey the rock mass conditions in the
tunnel face. Before the observation, the rock powder and
fragments attached on the rock face was washed out. Then, the
tunnel face can be mapped through four openings used for the
conveyance of the mucks. Usually, the tunnel face mapping
includes: (1) rock type; (2) color; (3) rock weathering grade;
(4) joint descriptions: number of joint sets, approximate joint
length in the tunnel face, joint aperture, joint alteration;
(5) measurements of joint sets: for each joint set, the joint
spacing and the angle between the tunnel axis and the joint plane
are estimated; (6) estimation of RQD in the tunnel face: 1 m ruler
used for the measurement of RQD at the same tunnel face position
and direction, RQD and fracture number are recorded for a

Table 3
Results of abrasivity testing on the Bukit Timah granite.

Borehole no. Depth (m) Geotechnical unit Rock type RQD Cerchar abrasivity index (CAI)

NR-8 (T06) 40.5 G1 Granite 95 4.78

NR-8 (T06) 43.6 G1 Granite 43 5.24
NR-9 (T06) 51.0 G1/G2 Granite 27 4.72
G-106 (T05) 25.0 G1 Diorite 100 4.53
H-108 (T05) 53.75 G1 Granite 100 4.44

The rock classified as extremely abrasive in accordance to the Cerchar definition.

Table 4
Laboratory rock strength test results of cored samples.

Rock type Geotechnical unit RQD UCS (MPa) Point load test (Is50, MPa)

Range Average Range Average

Bukit Timah granite G1 0–94 19–209 85 1.2–14.5 7

G2 0–25 – 51 0.7–5.1 1.8
 ARTICLE IN PRESS
12 Q.M. Gong, J. Zhao / International Journal of Rock Mechanics & Mining Sciences 46 (2009) 8–18
supplement to the joint spacing in the face mapping; and (7)
water inflow from the tunnel face.
In order to continuously record the various rock mass
conditions encountered by TBM, the face mapping and rock
sampling were frequently conducted. During the tunnel construc-
tion, tunnel face mapping and muck sampling or rock core
sampling at the same tunnel face were carried out about once per
30–50 m or when the TBM performance changed dramatically in
the same operation conditions, for example, the thrust suddenly
decreased or increased. Totally, more than 80 tunnel face
mappings were taken in T05 and T06 tunnels. The rock mass
conditions vary from fresh granite rock mass to very highly
fractured and highly weathered rock mass.
4.2.2. Field rock coring program
In order to identify the rock mass conditions, the core rock
samples at the tunnel face were taken by drilling during tunnel
face mapping. Due to the limited space in the cutterhead chamber,
a drilling machine with a small inner diameter of 34 mm was used
for core rock sampling. The drilling process is described as
follows: (1) fix the drilling machine in the cutterhead; (2) rotate
the cutterhead to select a suitable location, in order to avoid the
joints and its weathered zones; (3) take the rock sample from the
drilling machine; and (4) repeat the above process to obtain
enough core samples.
The rock core samples must be enough to conduct three
uniaxial compressive strength tests and five Brazilian tensile
strength tests. Because coring at the tunnel face was a time-
consuming process, sampling in some rings was carried out in the
tunnel wall during the machine maintenance in order not to affect
the project progress. Since sampling was designed to be
conducted in the mapped tunnel face, the frequency of sampling
was the same with the tunnel face mapping. Thus, the rock mass
conditions are completely obtained. But, in some rings, it was very
difficult to drill core samples due to highly weathered and
fractured rock mass conditions. Only hand rock samples were
taken from the cutterhead chamber.
4.3. Rock mass parameters
4.3.1. Joint spacing
Joint spacing can be represented by RQD, joint frequency and
volumetric joint count. For a particular set of joints, joint normal
spacing, the mean perpendicular distance between the adjacent
joints, is usually used to express the joint spacing. Joint spacing is
normally measured along a specific direction (scan line) for all
discontinuities and represented by the mean spacing of all
discontinuities along the scan line. The volumetric joint count
(Jv), defined as the sum of the number of joints per cubic meter for
each joint set present [37,38], is also used in practice. According to
this definition, Eq. (1) is derived; it takes into account all the joints
and fractures encountered:
X 1  Nrð5Þ or Nrð10Þ 
Jv ¼ þ (1)
J si 5
where Jsi is the joint spacing in meters for the actual joint set, and
Nr(5) or Nr(10) is the number of random joints along 5 or 10 m
perpendicular to the sampling lines.
Because the shield prevents the observation of the tunnel wall,
the scan line measurement along the tunnel wall is not possible.
However, the tunnel face mapping was conducted during cutter
inspection and replacement. The joint sets and their spacing were
measured and the random joints were recorded in tunnel face
mapping. Thus, the volumetric joint number can be derived from
the tunnel face mapping. The joint spacing is hereafter denoted by
the volumetric joint number for the TBM PR prediction model.
4.3.2. Joint orientation
The rock mass conceptual model shows that joint orientation is
another important factor affecting the performance of TBM.
Usually, a rock mass contains several major sets of joints plus
some random joints. Each joint set may have different effects on
the TBM PR. The higher the joint density or frequency, the larger
the effect of the joint set on the TBM PR. Therefore, the orientation
of the most frequent joint or the most closely spaced joint set
would be considered for the PR prediction model. The angle (a)
between the tunnel axis and the joint plane can also be
determined from the tunnel face mapping.
4.3.3. Uniaxial compressive strength test
Uniaxial compressive strength is one of the most basic rock
strength parameters for rock mass condition evaluation. It is
commonly used to assess rock mass boreability. The uniaxial
compressive strength tests were performed in accordance with
the procedure recommended in ISRM standard [39]. A ratio
between length and diameter of the sample is mostly 2.5 and
some range from 2 to 2.5 in light of the core sample length. The
loading rate used in the tests is 1.0 MPa per second.
Because the rock sample diameter is not the standard size,
50 mm. Thus, the measured UCS is corrected to the strength for a
sample with diameter of 50 mm, using the following equation
proposed by Hoek and Brown [40]:
 0:18
50
sc50 ¼ scd (2)
d
where scd is the uniaxial compressive strength of a rock specimen
with a diameter of d mm, sc50 is the compressive strength of a
50 mm diameter sample.
4.3.4. Rock brittleness index
The brittleness index is defined as the ratio of the uniaxial
compressive strength to Brazilian tensile strength [27]. It is an
indirect test parameter and used to evaluate the rock mass
boreability. Together with the above measured uniaxial compres-
sive strength, the Brazilian tensile strength is used to calculate the
rock brittleness index. Rock Brazilian tensile strength is also
measured according to the procedure recommended by ISRM [39].
The ratio of diameter to thickness of the disk is about 2:1. A
loading rate of 200 N per seconds is adopted. The specimen
ruptures usually along a single tensile fracture aligned with the
loading axis. If the fracture plane deviates significantly from a
straight line between platen contacts, the test is regarded invalid.
4.4. TBM performance data
Generally, TBM was designed based on the given site ground
conditions. It took into account all possible ground conditions
including hard rock, soft ground as well as mixed face ground in
some cases. TBMs are well known to run thrust-limited in hard
rock formations, because it takes more thrust to penetrate the
rock. On the other hand, TBMs run power limited in the soft
ground due to the higher rolling force needed for deeper
penetration for a given level of thrust [9]. In the mixed face
ground, TBMs cannot always operate efficiently due to cutterhead
vibration and face instability.
To fully understand the effect of geological conditions on the
machine performance, the level of thrust and torque used to reach
a certain PR has to be measured. In other words, the magnitudes
of cutter load, torque, PR have to match the rock mass properties.
 ARTICLE IN PRESS

Q.M. Gong, J. Zhao / International Journal of Rock Mechanics & Mining Sciences 46 (2009) 8–18 13

To analyze the effect of variable rock mass conditions on the TBM 25 100%
performance, the operating parameters (thrust force, torque and 90%
RPM) of the machine should be monitored at rings where the
20 80%
tunnel face mapping and intact rock strength tests were Frequency
conducted. Thus, the effect of the rock mass properties on Cumulative % 70%

Frequency
machine PR can be analyzed and evaluated. Fortunately, the 15 60%
machine operating parameters can be monitored automatically by 50%
the TBM acquisition system in T05 and T06. The monitored 10 40%
parameters include thrust force, torque, RPM, PR, and other useful
30%
state parameters. These parameters are recorded once every ten
seconds or an average per ring. The recorded files can be opened 5 20%
by EXCEL spreadsheet. 10%
0 %
4.5. Database 0 5 10 15 20 25 30
Jv
TBM specifications were collected from the documents Fig. 3. Histogram of Jv variation in the database.
provided by the manufacturer. TBM operating parameters can be
extracted directly and analyzed from the TBM acquisition system.
The corresponding rock mass properties were obtained by tunnel
25 100%
face mapping and laboratory rock tests. After extensive data
collection in T05 and T06, a database including rock mass Frequency 90%
properties, TBM specifications and the corresponding TBM 20 Cumulative % 80%
operating parameters was established. The dataset is shown in 70%
Table 5. The database is analyzed progressively to further set up a Frequency
15 60%
function for the prediction of TBM PR.
50%
Although many tunnel face mappings have been conducted
during tunnel construction, only part of these were used to 10 40%
establish the database, because no corresponding core rock 30%
samples were obtained in some locations. Some rock experiment 5 20%
results were not used too, because these samples failed along 10%
certain structural weak planes. Only a total of 47 datasets were
0 %
used for the model set-up.
100 120 140 160 180 200 220 240 260
Figs. 2–5 show histograms of the rock mass properties in the
UCS (Mpa)
established database. These graphs present the range and
Fig. 4. Histogram of UCS variation in the database.

20 100%
Table 5
18 Frequency 90%
Dataset for the prediction model. Cumulative %
16 80%
TBM specifications TBM operating parameters Rock mass parameters
14 70%
Frequency

TBM diameter Thrust Uniaxial compressive strength 12 60%

Cutter number Torque Brittleness index
Cutter spacing Revolution per minute Joint spacing 10 50%
Cutter size and shape Penetration rate Joint orientation 8 40%
6 30%
4 20%
12 100% 2 10%
Frequency
Cumulative % 90% 0 %
10
80% 8 10 12 14 16 18 20 22
70% BI
8
Frequency

60% Fig. 5. Histogram of brittleness index variation in the database.

6 50%
40%
4
30% cumulative distribution curve of each parameter. Fig. 2 shows
20% the range of the angle between the tunnel axis and the joint plane,
2 which is mainly distributed between 01 and 801. As the angle is
10%
close to 901, it is difficult to be observed at the tunnel face.
0 % Generally, the angle is evenly distributed from 01 to 801 except the
10 20 30 40 50 60 70 80 90
range between 201 and 301. Fig. 3 shows the range of the
Angle (degree) volumetric joint count (Jv), which is between 0 and 30. In most
Fig. 2. Histogram of variation of the angle between tunnel axis and joint plane in cases, the volumetric joint count is less than 15 in the mapped
the database. granitic rock masses. The results are consistent with the joint
 ARTICLE IN PRESS

14 Q.M. Gong, J. Zhao / International Journal of Rock Mechanics & Mining Sciences 46 (2009) 8–18

investigation in Mandai granite [41]. Fig. 4 shows the range of the 14 100%
rock uniaxial compressive strength, which is between 100 and Frequency
90%
260 MPa. The UCS mainly concentrates between 160 and 200 MPa, 12 Cumulative %
80%
which is the normal compressive strength of fresh granite found
in Singapore. Fig. 5 shows the range of the rock brittleness index 10 70%
which ranges from 8 to 22. The high frequency zone is from 14 to

Frequency
8 60%
16. The rock compressive strength and rock brittleness mainly
50%
concentrate on a certain range respectively, because the encoun- 6
tered rock in the project is the granite with different weathering 40%
grades. 4 30%
20%
2
10%
5. Rock mass characteristics prediction model for TBM
penetration rate 0 %
110 130 150 170 190 210 More
5.1. Rock mass boreability analysis Specific Rock Mass Boreability Index (kN/cutter/mm/rev.)

Fig. 6. Histogram of the specific rock mass boreability index variation in the
Due to the influence of machine specifications and TBM database.
operation parameters, TBM PR can be different at the same rock
mass condition. In order to eliminate these effects and find out a
parameter to assess the rock mass boreability, a specific rock mass
boreability index (SRMBI), defined as a boreability index at the PR 230
equal to 1 mm per revolution, was proposed to evaluate the rock Predicted Specific Rock Mass Boreability Index 210
mass boreability in different rock mass conditions [42]. The
specific rock mass boreability index eliminates the influence of 190
TBM operation parameters (e.g., RPM, force per cutter). Never-
theless, the effect of TBM specifications (e.g., cutter spacing, cutter 170
(kN/cutter/mm/rev.)

diameter and cutter tip width) on TBM PR still exists. However,

the rock mass boreability can be compared and evaluated in the 150
same tunnel or in the tunnels excavated by similar TBMs. The
130
relationships among the rock mass boreability index, the specific
rock mass boreability index and TBM PR are estimated as follows: 110
BI  BIð1Þ P 0:75 , (3) 90
where BI is the rock mass boreability index, BI(1) is the specific
rock mass boreability index, and P is the penetration per 70
revolution.
50
On the basis of the established database, the corresponding 50 70 90 110 130 150 170 190 210 230
specific rock mass boreability indices with relevance to the
Measured Specific Rock Mass Boreability Index
obtained rock mass conditions were calculated. Fig. 6 shows the (kN/cutter/mm/rev.)
range of the specific rock mass boreability index, which varies
from 110 kN/cutter/mm/rev to more than 210 kN/cutter/mm/rev. It Fig. 7. Comparison between the measured and predicted specific rock mass
boreability index.
mainly distributes in the range between 130 and 210 kN/cutter/
mm/rev.

On the basis of the statistical results, the proportion of variance

5.2. Regression analysis results
A commercial statistical software, Nonlinear Regression Ana-
lysis Program (NLREG), was used to carry out the multi-variables
regression analysis [43]. After a series of modeling, the best
combination of rock mass parameters to predict the specific rock
mass boreability index is identified with reference to the
established database. The prediction model is as follows:
BIð1Þ ¼ 37:06UCS0:26 Bi0:10 ð0:84 e0:05Jv þ e0:09 sinðaþ30Þ Þ,
where BI(1) is the specific rock mass boreability index, UCS is the
rock uniaxial compressive strength (MPa), Bi is the rock brittle-
ness index, Jv is the volumetric joint count and a is the angle
between the tunnel axis and the joint plane.
The F-test related to the utility of the overall regression model
was carried out. The model statistic value F and Prob(F) are 24.53
and 0.00001, respectively. The null hypothesis can be rejected.
That means at least one of the independent variables, namely the
four rock mass parameters significantly affects the specific rock
mass boreability index.
explained (R2) is 74.94% and the adjusted coefficient of multiple
determination (R2a ) is 71.89%. It indicates that the above regression
model explains 74.94% of the total variance of the 47 datasets. The
model adequacy was also assessed. The correlation coefficient (r)
between the residuals and their expected values is 0.996, which is
very close to 1. The measured specific rock mass boreability index
is compared with the predicted one, as shown in Fig. 7. Most of the
residuals are less than 10 kN/cutter/mm/rev.
(4)
6. Effect of rock mass parameters on penetration rate
Combined the statistical prediction model with the relation-
ships among the rock mass boreability index, the specific rock
mass boreability index and the thrust force, the effect of rock mass
parameters on the PR are revealed in this section. The results are
compared with that of the numerical modeling. Since the
regression model is based on the TBM performance in T05 and
T06, the thrust force per cutter is assumed to be 200 kN in order to
basically reflect the actual TBM operation state. The rock mass
 ARTICLE IN PRESS

Q.M. Gong, J. Zhao / International Journal of Rock Mechanics & Mining Sciences 46 (2009) 8–18 15

12.0
11.0 angle = 30, Jv = 5, BI = 15
angle = 30, Jv = 10, BI = 15
10.0
angle = 30, Jv = 20, BI = 15
9.0
Penetration (mm/rev) 8.0
7.0
6.0
5.0
4.0
3.0
2.0
1.0
0.0
60 80 100 120 140 160 180 200 220 240 260
UCS (MPa)

Fig. 8. Penetration variation with different rock strengths at thrust force of 200 kN/cutter.

conditions are also selected to stand for the typical rock mass 8.0
conditions revealed in T05 and T06. angle = 30, Jv = 10, UCS = 100
7.0 angle = 30, Jv = 10, UCS = 200
angle = 30, Jv = 10, UCS = 300
6.1. Rock uniaxial compressive strength 6.0
Peneteration (mm/rev)

5.0
With the thrust force held at 200 kN/cutter, the relationship
between UCS and penetration per revolution is shown in Fig. 8. In 4.0
the graph, the maximum PR is taken as 12 mm/rev, because of the
3.0
limits of the machine torque and muck removal capacity.
Generally, the PR decreases with increasing UCS. The PR 2.0
distributes in a large range from about 1 mm/rev to more than
1.0
12 mm/rev due to the effect of UCS. It shows the UCS has a
decisive influence on TBM PR. It agrees with the analysis in the 0.0
conceptual model. 0 5 10 15 20 25 30
Brittleness Index
6.2. Rock brittleness index
Fig. 9. Penetration variation with different rock brittleness indices at thrust force
of 200 kN/cutter.
Fig. 9 shows the relationship between the rock brittleness
index and penetration per revolution, as the thrust force was fixed 12.0
at 200 kN/cutter. The PR increases almost linearly with increasing
11.0 angle = 30, BI = 15, UCS = 100
rock brittleness index. The result coincides with the numerical angle = 30, BI = 15, UCS = 200
10.0
simulation result [27]. It is obvious that the increase of the rock angle = 30, BI = 15, UCS = 300
Peneteration (mm/rev)

9.0
brittleness index facilitates the cutter indentation process.
8.0
7.0
6.3. Volumetric joint count 6.0
5.0
The effect of volumetric joint count on the PR is shown in 4.0
Fig. 10, as the thrust force was fixed at 200 kN/cutter. The PR
3.0
increases with increasing volumetric joint count. The curve can be
2.0
divided into three sections as follows: the first section is when Jv is
1.0
less than 5, the PR almost increases linearly with increasing Jv; the
0.0
second section is when Jv between 6 and 15, the PR is sensitive to
0 5 10 15 20 25 30
Jv and the incremental rate of the PR increases with increasing Jv;
JV
the third section is when Jv is more than 15, the PR almost
increases linearly, but the incremental rate is higher than that in Fig. 10. Penetration variation with different volumetric joint counts at thrust force
the first section. The results are consistent with the numerical of 200 kN/cutter.
simulations [20]. The joint spacings of 80 and 200 mm are two
critical points in the graph. While the volumetric joint count is the numerical modeling and the Brulnad’s results. All of the
transferred into the equivalent joint spacing, the effect of joint results reveal that due to the influence of the joint spacing, the PR
spacing on the PR are shown in Fig. 11, together with the results of may increase more than nine times. The tendency of the four
 ARTICLE IN PRESS
16 Q.M. Gong, J. Zhao / International Journal of Rock Mechanics & Mining Sciences 46 (2009) 8–18
10.0
9.0 simulated results
in situ measurements (St)
8.0 in situ measurements (Sp)
statistical results
7.0
6.0
PS /P0
5.0
4.0
3.0
2.0
1.0
0.0
0 100 200 300 400 500 600 700 800
Joint Spacing (mm)
Fig. 11. Effect of joint spacing on penetration rate (Ps and P0 denotes, respectively,
penetration rate in the rock mass with joint spacing S and with no joint).
curves is consistent. The prediction model result is almost
consistent with that of the numerical modeling, and slightly
smaller than the results by Bruland.
The difference between the results of the new model and the
Bruland’s results may be derived from the joint classification. The
discontinuities are divided into two types in NTNU model [7],
namely joints (Sp) and fissures (St), and five classes. These two
types of discontinuities were independently measured and taken
into account. The effect of each discontinuity class and its
orientation on the TBM PR was obtained. In this model, all
discontinuities were regarded as joints and measured together. In
addition, during the tunnel face mapping, the joint properties
such as the weathering grade, fillings and aperture were not
recorded in details due to the time limit in the cutterhead
chamber. The joint length cannot be measured due to the small
openings in the cutterhead and thus not taken into consideration.
In most cases, the joints in the granitic rock mass are tight and
well matched, and the joint planes are fresh. These joint proper-
ties, which are the same with the numerical simulation, may also
be the main reasons why the influence of joint spacing on the PR
in this model is less than that of the results by Bruland.
6.4. Joint orientation
Fig. 12 shows the relationship between the joint orientation
and penetration per revolution, as the thrust force was fixed to
200 kN/cutter. The prediction model results present the same
tendency with the numerical simulation and the results by
Bruland. The PR increases with increasing angle between tunnel
axis and joint plane as the angle is less than 601, and then
decreases with increasing angle. The PR reaches the maximum
value as the angle is equal to 601. But, the influence of the angle on
the PR is smaller than that of the numerical simulation
and the results by Bruland. In the database, the angle adopted is
the angle between tunnel axis and the minimum spacing joint set.
It does not take the angle of the other joint sets and random joints
into consideration because it is very difficult to obtain the
equivalent angle for several joint sets theoretically. So, the
influence of these angles between tunnel axis and other joint
sets with larger spacing and random joints is neglected in this
model. Another reason may be the error of the measurements
during the tunnel face mapping. Due to the small openings in the
cutterhead and time limit for surveying, it is not possible to
measure the angle directly. In most cases, the angle was estimated
visually.
4.0
3.0
Peneteration (mm/rev)
2.0
1.0 UCS = 200, BI = 15, Jv = 5
UCS = 200, BI =1 5, Jv = 10
UCS = 200, BI = 15, Jv = 15
0.0
900 1000 0 10 20 30 40 50 60 70 80 90
Angle (degree)
Fig. 12. Penetration variation with different joint orientations at thrust force of
200 kN/cutter.
7. Model limitations
7.1. Parameters of tunnel boring machine
Although the thrust force per cutter is a value normalized by
the cutter number in the prediction model, the stress level acting
on the rock face is greatly affected by the cutter diameter and
cutter tip width even if the force per cutter is the same. In fact, it is
the stress that influences the rock fragmentation process. The
optimal cutter spacing in different rock masses also affects the
machine efficiency significantly. In this model, the machine
parameters are held constant. When the machine parameters
are changed, especially cutter diameter, cutter tip width and
cutter spacing, the model should be modified in order to consider
the effects of these parameters.
7.2. Parameters of rock mass
All of the statistical data are based on the granitic rock mass
found in Singapore in this prediction model. Although these
parameters of rock masses vary in a large range, the rock type is
only granite formation. When the model is applied to rock masses
other than granite, special attention must be paid due to the
singularity of the original data of the prediction model.
7.3. Boundary conditions
The boundary conditions of rock masses mainly include the in
situ stress and groundwater. The groundwater may primarily
impact the advance rate [3]. Tunnel instability accidents always
occur with inflow of groundwater. But it has little effect on the PR.
The in situ stress impacts both the PR and advance rate. If the in
situ stress is not enough to promote stress slabbing at the face, the
stress that acts on TBM cutters to form and propagate cracks
increases with increasing confining stress [16,18] and the
corresponding PR decreases. If the in situ stress is high relative
to rock strength to yield stress slabbing and the rock is not greatly
overstressed or susceptible to bursting, the PR may result in an
increase [44]. Myrvang et al. [45] and Boniface [46] observed that
the suitable in situ stress increases TBM PR in tunneling projects.
When the in situ stress is high and the rock is overstressed, the
stress slabbing and spalling, raveling, face overbreak and ground
squeezing may occur, which must be controlled with shield or
cutterhead modifications and rock supports [45,47]. In this case,
the PR and advance rate are greatly affected. In this model, the in
 ARTICLE IN PRESS
Q.M. Gong, J. Zhao / International Journal of Rock Mechanics & Mining Sciences 46 (2009) 8–18
situ stress in the tunnel sites is low relative to the rock strength.
Hence, the model does not take the high in situ stress into
consideration. The effects of in situ stress on PR should be further
studied based on this model.
8. Conclusions
On the basis of the rock fragmentation process, the rock mass
properties influencing TBM PR were analyzed. A rock mass
conceptual model that includes four rock mass parameters,
namely rock material compressive strength, rock brittleness, joint
spacing and joint orientation, is proposed to evaluate rock mass
boreability. In order to quantify the rock mass conceptual model, a
comprehensive program was performed to develop a database of
rock mass properties and TBM performance during the construc-
tion of the DTSS project in Singapore. By analyzing the face
mappings and screening the rock test results, a total of 47 datasets
were obtained.
By using multiple variables nonlinear regression analysis, the
rock mass characteristics prediction model was created. Of the
four rock mass parameters in the new model, rock uniaxial
compressive strength and volumetric joint count have predomi-
nant effects on the PR. With the increase of rock uniaxial
compressive strength, the PR decreases. The PR increases
approximately linearly with increasing rock brittleness index.
The PR increases with increasing volumetric joint count. The PR
increases with the increase of the angle between tunnel axis and
joint plane, when the angle is less 601, then decreases with
increasing angle. Due to the limitation of the obtained database,
attention must be paid when the model is used to predict the PR
with critical different tunneling conditions.
Acknowledgments
The authors thank Sembcorp, Public Utility Board of Singapore
for allowing these data obtained in the tunneling sites to be
published. The first author thanks the fund provided by Nanyang
Technological University during his Ph.D. study. This study was
also partially supported by the China National Natural Science
Foundation (Grant no. 50479071).
References
[1] Graham PC. Rock exploration for machine manufacturers. In: Bieniawski ZT,
editor. Exploration for rock engineering. Johannesburg: Balkema; 1976. p.
173–80.
[2] Farmer IW, Glossop NH. Mechanics of disc cutter penetration. Tunnels
Tunnell Int 1980;12:22–5.
[3] Nelson PP. Tunnel boring machine performance in sedimentary rock. PhD
dissertation, Cornell University, Ithaca, NY, 1983.
[4] Hughes HM. The relative cuttability of coal measures rock. Min Sci Technol
1986;3:95–109.
[5] O’Rourke JE, Spring JE, Coudray SV. Geotechnical parameters and tunnel
boring machine performance at Goodwill Tunnel, California. In: Nelson PP,
Laubach SE, editors. Proceedings of the 1st North American rock mechanics
symposium, Austin, Texas. Rotterdam: Balkema; 1994.
[6] Rostami J. Development of a force estimation model for rock fragmentation
with disc cutters through theoretical modeling and physical measurement of
crushed zone pressure. Doctoral dissertation, Colorado School of Mines,
Golden, CO, 1997.
[7] Bruland A. Hard rock tunnel boring. Doctoral thesis, Norwegian University of
Science and Technology, Trondheim, 1998.
[8] Nelson PP, Yousof A Al-Jalil, Laughton C. Improved strategies for TBM
performance prediction and project management. In: RETC proceedings, 1999.
p. 963–79.
[9] Cheema S. Development of a rock mass boreability index for the performance
of tunnel boring machines. PhD dissertation, Colorado School of Mines,
Golden, CO, 1999.
[10] Barton N. TBM tunnelling in jointed and faulted rock. Rotterdam: Balkema;
2000.
17
[11] Grima MA, Bruines PA, Verhoef PNW. Modeling tunnel boring machine
performance by Neuro-fuzzy methods. Tunnell Undergr Space Tech
2000;15:259–69.
[12] Innaurato N, Mancini R, Rondena E, Zaninetti A. Forcasting and effective
TBM performance in a rapid excavation of a tunnel in Italy. In: Wittke W,
editor. Proceedings of the 7th international cong rock mechanics, 1991.
p. 1009–14.
[13] Sundaram NM, Rafek AG, Komoo I. The influence of rock mass properties in
the assessment of TBM performance. In: Proceedings of the 8th international
IAEG congress. Rotterdam: Balkema; 1998. p. 3353–9.
[14] McFeat-Smith I. Mechanised tunnelling for Asia. Workshop manual, orga-
nized by IMS Tunnel Consultancy Ltd., 1999.
[15] Sapigni M, Berti M, Bethaz E, Busillo A, Cardone G. TBM performance
estimation using rock mass classifications. Int J Rock Mech Min Sci 2002;39:
771–88.
[16] Cook NGW, Hood M, Tsai F. Observations of crack growth in hard rock loaded
by an indenter. Int J Rock Mech Min Sci 1984;21:97–107.
[17] Mishnaevsky JR. Physical mechanisms of hard rock fragmentation under
mechanical loading: a review. Int J Rock Mech Sci 1995;32:763–6.
[18] Liu HY, Kou SQ, Lindqvist CA, Tang CA. Numerical simulation of the rock
fragmentation process induced by indenters. Int J Rock Mech 2002;39:
491–505.
[19] Chiaia B. Fracture mechanisms induced in a brittle material by a hard cutting
indenter. Int J Solids Struct 2001;38:7747–68.
[20] Gong QM, Jiao YY, Zhao J. Numerical modeling of the effects of joint spacing
on rock fragmentation by TBM cutters. Tunnell Undergr Space Tech 2006;21:
46–55.
[21] Pang SS, Goldsmith W, Hood M. A force-indentation model for brittle rocks.
Rock Mech Rock Eng 1989;22:127–48.
[22] Lundberg B. Penetration of rock by conical indenters. Int J Rock Mech Min Sci
1974;4:269–72.
[23] Snowdon RA, Ryley MD, Temporal J. Reply to the discussion by HK Kutter and
HP Sanio of the paper by RA Snowdon, MD Ryley and J Temporal, ‘‘A study of
disc cutting in selected British rocks’’. Int J Rock Mech Min Sci 1983;20:105.
[24] Gehring K. Experience with TBM-application under extreme rock condition in
a South-African project leads to development of high-performance disk
cutters. In: Proceedings of the 7th international IAEG congress. Rotterdam:
Balkema; 1994. p. 4243–52.
[25] Snowdon RA, Ryley MD, Temporal J. A study of disc cutting in selected British
rocks. Int J Rock Mech Min Sci 1982;19:107–21.
[26] Rostami J, Ozdemir L. A new model for performance perdiction of hard rock
TBMs. In: RETC proceedings, 1993. p. 793–809.
[27] Gong QM, Zhao J. Influence of rock brittleness on TBM penetration rate in
Singapore Granite. Tunnell Undergr Space Tech 2007;22:317–24.
[28] Buchi E. The influence of joints on TBM performance. Private communication,
2007.
[29] Howarth DF. The effect of jointed and fissured rock on the performance of
tunnel boring machines. In: Proceedings of the international symposium on
weak rock, Tokyo, 1981. p. 1069–74.
[30] Aeberli U, Wanner WJ. On the influence of discontinuities at the application of
tunnelling machines. In: Proceedings of the 3rd International Congress IAEG,
Madrid, 1978. p. 7–14.
[31] Thuro K, Plinninger RJ. Hard rock tunnel boring, cutting, drilling and blast-
ing: rock parameters for excavatability. In: Proceedings of the 10th inter-
national congress ISRM, South African Institute on Mineral Metallurgy, 2003.
p. 1–7.
[32] Sanio HP. Prediction of the performance of disc cutters in anisotropic rock. Int
J Rock Mech Min Sci 1985;22:153–61.
[33] Gong QM, Zhao J, Jiao YY. Numerical modeling of the effects of joint
orientation on rock fragmentation by TBM cutters. Tunnell Undergr Space
Tech 2005;20:183–91.
[34] Public Works Department of Singapore, Geology of the Republic of Singapore,
1976.
[35] Rostami J, Ozdemir L, Bruland A, Dahl Filip. Review of issues related to
Cerchar abrasivity testing and their implications on geotechnical investiga-
tions and cutter cost estimates. In: RETC Proceedings, 2005.
[36] Zhao J, Broms BB, Zhou Y, Choa V. A study of the weathering of the Bukit
Timah granite, Part A: review, field observations and geophysical survey. Bull
Int Ass Eng Geol 1994;49:97–106.
[37] International Society for Rock Mechanics. Basic geotechnical description of
rock masses. Int J Rock Mech Min Sci 1981;18:85–110.
[38] Palmstrom A. The volumetric joint count—a useful and simple measure of the
degree of rock mass jointing. In: Proceedings of the 4th international congress
on international association on engineering geology, India. Rotterdam:
Balkema; 1982. p. 221–8.
[39] Brown ET. Rock characterization testing and monitoring. ISRM suggested
methods, 1981.
[40] Hoek E, Brown ET. Practical estimates of rock mass strength. Int J Rock Mech
Min Sci 1997;34:1165–86.
[41] Zhao J. Construction and utilization of rock caverns in Singapore. Part A: the
Bukit Timah granite bedrock resource. Tunnell Underg Space Tech
1996;11:65–72.
[42] Gong QM, Zhao J, Jiang YS. In situ TBM penetration tests and rock mass
boreability analysis in hard rock tunnels. Tunnell Undergr Space Techn
2007;22:303–16.
[43] Sherrod HP. Nonlinear regression analysis program. NLERG manual, 1996.
 ARTICLE IN PRESS

18 Q.M. Gong, J. Zhao / International Journal of Rock Mechanics & Mining Sciences 46 (2009) 8–18

[44] US Army Corps of Engineers. Engineering and design—tunnels and shafts in
rock. CECW-EG engineer manual. Appendix C: tunnel boring machine
performance concepts and performance prediction, 1997.
[45] Myrvang A, Blindheim OT, Johansen ED. Rock stress problems in bored
tunnels. Norweg Soil Rock Eng Ass 1998;11:56–62.
[46] Boniface A. Tunnel boring machine performance in basalts of the Lesotho
formation. Tunnell Undergr Space Tech 2000;15:49–54.
[47] Phien-wej N, Cording EJ. Sheared shale response to deep TBM excavation. Eng
Geol 1990;30:371–91.