You are on page 1of 16

Int. J. Mining and Mineral Engineering, Vol. 4, No.

3, 2013 249

Utilising of linear and non-linear prediction tools for


evaluation of penetration rate of Tunnel Boring
Machine in hard rock condition

Alireaza Salimi*
Department of Mining Engineering,
South Tehran Branch,
Islamic Azad University,
Tehran, 4435-11365, Iran
E-mail: Salimi.tunneling@gmail.com
*Corresponding author

Mohammad Esmaeili
Department of Mining Engineering,
Science and Research Branch,
Islamic Azad University,
Tehran, 14778-93855, Iran
E-mail: mohamad.esmaeily@gmail.com

Abstract: Predicting Tunnel Boring Machine (TBM) penetration rate is a


crucial issue for the successful fulfilment of a mechanical tunnel project.
Penetration rate depends on many factors such as intact rock properties, rock
mass conditions and machine specifications. In this paper, linear and non-linear
multiple regression as well as Artificial Neural Network (ANN) techniques
were applied to predict the penetration rate of TBM. In developing of the
proposed models, five parameters, which include Uniaxial Compressive
Strength (UCS), Brazilian Tensile Strength (BTS), peak slope index (punch
penetration), spacing of discontinuities (of weakness planes) and orientation of
discontinuities with respect to the tunnel axis (β angle), were incorporated.
For this study, 46 datasets were collected. Performance of these models was
assessed through the R2, RMSE and MAPE. As a result, these indices revealed
that the prediction performance of the ANN model is higher than that of the
non-linear and linear multiple regression models.

Keywords: penetration rate; TBM; tunnel boring machine; ANN; artificial


neural network; multiple regression analyses.

Reference to this paper should be made as follows: Salimi, A. and


Esmaeili, M. (2013) ‘Utilising of linear and non-linear prediction tools
for evaluation of penetration rate of Tunnel Boring Machine in hard rock
condition’, Int. J. Mining and Mineral Engineering, Vol. 4, No. 3, pp.249–264.

Biographical notes: Alireaza Salimi received his BSc and MSc from Mining
Engineering (Excavation) Department of Mining at the South Tehran Branch,
Islamic Azad University in 2006 and 2008, respectively. He is currently

Copyright © 2013 Inderscience Enterprises Ltd.


250 A. Salimi and M. Esmaeili

studying as PhD student at Institute of Geotechnical Engineering, University of


Stuttgart (IGS). His areas of interest are blasting, rock mass characterisation,
and rock mechanics especially tunnelling.

Mohammad Esmaeili obtained his Bachelor’s Degree in Mining Engineering


from Islamic Azad University (IAU), Savadkooh Branch in 2007 and MSc
from Islamic Azad University (IAU), Science and Research Branch (SRB) of
Iran in 2011. He works as a Drilling and Blasting Engineer at Jalalabad Iron
Ore Complex in Kerman Province. His research interests are in the areas of
drilling and blasting, tunnelling and mechanical excavation. He has published
research papers at international journals and conference proceedings.

1 Introduction

Since the first Tunnel Boring Machine (TBM) was built by Robbins in the late 1950s,
many researchers have been addressing for predicting the TBM penetration rate which is
an essential factor for time planning, cost control and choosing the excavation method
(Tarkoy, 1975; Ozdemir, 1977; Nelson and O’Rourke, 1983; Rostami and Ozdemir,
1993; Bruland, 1999; Barton, 2000; Yagiz, 2002, 2006; Yagiz et al., 2008). Uniaxial
Compressive Strength (UCS) and tensile strength such as Brazilian Tensile Strength
(BTS) have been widely used as input parameters for predicting TBM penetration rate
(Graham, 1976; Farmer and Glossop, 1980; Hughes, 1986; McFeat-Smith, 1977;
Ozdemir, 1977). Gong and Zhao (2007) studied the influence of rock brittleness on TBM
penetration rate in a Singapore granite deep tunnel sewerage system. The effects of joint
orientation and joint spacing on penetration rate have been studied by many researchers
(Gong et al., 2005; Bejari et al., 2011). Rostami and Ozdemir (1993) developed a model
for estimating the cutting force of disc cutters based on the BTS and UCS of rock.
Consequently, Yagiz (2002) modified this model adding rock mass properties such as
orientation of discontinuities as an input parameter into the model. Cassinelli et al. (1982)
used correlation between rock structure rating system and actual TBM performance to
estimate the penetration rate. Barton reviewed a wide range of TBM-bored tunnels to
establish databases required to estimate penetration rate, advance rate and utilisation of
TBM. In order to estimate the TBM penetration rate, Barton slightly modified the Q rock
classification system and produced a new equation, defined as QTBM (Barton, 1999,
2000). Ribacchi and Lembo-Fazio (2005) evaluated the relationship between Rock Mass
Rating (RMR) and performance of a double shield machine in the Varzo tunnel. Also
Sapigni et al. (2002) studied the empirical relation between RMR and penetration rate.
Yagiz (2008) has also performed statistical analysis on data obtained from Queens’s
tunnel (New York) and proposed an empirical model to predict TBM penetration rate.
He has related four rock mass parameters (UCS, punch test index, spacing and orientation
of joints) to penetration rate of machine. Gong and Zhao (2009), by performing a
non-linear regression analysis on data obtained from two tunnels excavated in granitic
rock masses in Singapore, developed an empirical equation to estimate boreability
of rock mass. They proposed a relationship between four rock mass parameters
(UCS, brittleness, joint count number and orientation of joints) and boreability index of
Utilising of linear and non-linear prediction tools for evaluation 251

the rock mass. In addition to theoretical, laboratory, empirical and statistical solutions,
utilising artificial intelligence methods has been recently investigated in many studies
(Lee and Lee, 1996; Alvarez Grima and Babuska, 1999; Alvarez Grima et al., 2000;
Gokceoglu, 2002; Okubo et al., 2003; Benardos and Kaliampakos, 2004; Gokceoglu and
Zorlu, 2004; Nefeslioglu et al., 2003, 2006; Sonmez et al., 2006; Kahraman et al., 2006;
Yoo and Kim, 2007; Tiryaki, 2008). In the following, Artificial Neural Network (ANN),
multiple linear and non-linear regressions were performed to predict penetration rate of
TBM in the current study.

2 Field study

The Karaj–Tehran water supply project, designed to transfer 16 m3/s of water from the
Karaj (Amir-Kabir dam) with length of 16 km, is under construction by the TBM. It is a
double shield machine type manufactured by Herrenknecht, with 4.56 m diameter.
The cutter head was installed with 31, each of 432 mm diameter, disc cutters. Table 1
shows the main characteristics of TBM. The main geological unit outcropped in the
project area is the Karaj formation. The formation is composed of a variety of Pyroclastic
rocks, often interbedded with sedimentary rocks. The characteristic rock type is a green
vitric to crystal lithic tuff, but other tuffs, such as tuff breccias and sandy and silty tuffs,
occur together with shales, siltstones and sandstones. The area around the tunnel is
moderately folded and intensely faulted. The bored section of the tunnel has been passed
through a wide syncline and anticline and some thrust faults with different length
especially at the first 3 km of tunnel resulting in many delays due to instabilities in
tunnel walls. The most important thrust faults in the area are the Poorkhan-Vardij and
North Tehran faults. According to geological characteristics of the region, the whole
excavation area has been divided into nine zones (Table 2). This paper is focused on
Gta (3, 4) and Sts (1, 2) with the total length of 9 km.

Table 1 Main specifications of TBM

Parameters Value
Machine diameter 4.65 m
Cutters diameter 432 mm
Numbers of disc cutters 31
Disc nominal spacing 90 mm
Maximum operating cutterhead thrust 16,913 KN
Cutterhead power 5 × 250 = 1.250 KW
Cutterhead speed 0–11 rpm
Cutterhead torque (nominal) 1723 KNm (6.58 rpm)
1029 KNm (11 rpm)
Thrust cylinder stroke 1400 mm
Conveyor capacity (approx.) 200 m3/h
TBM weight (approx.) 170 tons
252 A. Salimi and M. Esmaeili

Table 2 Geotechnical zones in the Karaj–Tehran water tunnel

Gta Sts
1 2 3 4 1 2 Tsh Mdg Crz
Zone length (m) 520 750 1410 3670 620 3280 1200 2300 2230
Rock type Tuff, Clay, schist, Conglomerate, Mudstone, tuff, Crushed
conglomerate tuff mudstone, tuff siltstone zone
Parameter
AVG. Density (t/m3) 2.6 2.6 2.6 2.6 2.6 2.55 2.6 2.8 –
Over burden (m) 200 350 350 500 150 250 150 100 300
RQD 25–50 30–60 50–75 90–100 50–75 75–100 50–70 70–85 <25
AVG. UCS (MPa) 30 75 100 120 120 120 60 90 <30
Q 10–20 10–20 10–20 10–20 20–30 20–30 20–30 20–30 0–5

3 Input parameters

Generally speaking, it is so difficult to choose all the relevant parameters that have major
effect on penetration rate of TBM. However, some parameters are independent and the
other parameters are strongly correlated with each other. Thus, using all variables as
input parameters was not necessary. In this regard, the Karaj–Tehran water conveyance
tunnel was investigated in both, field and laboratory, in order to establish a database to be
utilised for development of ANN, linear and non-linear multiple regression models. In the
Karaj–Tehran tunnel, the use of double-shield TBM with segmental lining prevented a
continuous surveying of excavated rocks. Due to the continuous segmental lining,
inspection and surveying of the rock mass were performed during the daily maintenance
of the boring machine, accessing the face from the TBM cutter head. Additional input
parameters were also obtained through surface investigations such as surveying and
sampling the outcrops and from the cores obtained from exploration drilling. During the
machine downtime, some rock samples were taken from the tunnel face. Also, additional
samples were taken from the muck on the belt conveyor for rock strength tests.

3.1 Intact rock parameters


UCS is one of the most important engineering properties of rocks. UCS is influenced by
many characteristics of rocks such as constitutive minerals and their spatial positions,
weathering or alteration rate, micro-cracks and internal fractures, density and porosity
(Osanloo, 1998). Also, the UCS test is very acceptable and available test that can be used
for studying the qualitative parameters such as matrix type and rock cementation. UCS
in comparison with other parameters has better capability to describe the rock’s
behaviour during the penetration into the rock (Bieniawski, 1973). Thus, the UCS can be
considered as representative of rock strength, density, weathering and matrix type.
BTS is another intact rock property which is intended to provide an indication of rock
toughness of crack propagation between adjacent cutters. In other words, the direction
of fracture propagation is controlled by tensile stresses at the crack tip. The Uniaxial
compressive and BTS tests were performed in accordance with the procedure
Utilising of linear and non-linear prediction tools for evaluation 253

recommended in the ISRM standard (Brown, 1981). The sampling method is regular and
each set of samples is gathered from an advancing period of TBM (each advancing period
was 3 m). Another important intact rock property which affects boreability is the
brittleness or the plasticity, which the rock exhibits when subjected to the mechanical
forces generated by the cutting action of an excavator. On the other word, the ease or
difficulty of crack propagation in rock is often referred to as brittleness. In this regard,
punch penetration test was developed in the late 1960s to provide direct laboratory
measurement of certain physical characteristics of rocks such as toughness and hardness
(Handewith, 1970; Szwedzicki, 1998). Szwedzicki (1998) employed the test for
measuring rock hardness and stated that this test could be used for predicting cutability of
rocks. Dollinger and Handewith (1998) indicated that the punch penetration test provides
qualitative data for investigating rock behaviour under the indentor. Ozdemir (2002)
stated that the test could be utilised for quantifying rock brittleness and toughness by
using test results. In this study, the punch penetration test was utilised to measure rock
brittleness. Increasing brittleness of rocks results in improving the rock cutting efficiency
of any mechanical tool.

3.2 Rock mass properties


Majority of developed models utilise intact rock properties such as UCS as input
parameters for estimating TBM penetration rate; however, mechanical cutting predictions
relying only on intact rock properties may provide inaccurate results (Cigla et al., 2001;
Yagiz, 2008). Because rock masses are composed of rock material and joints, the existing
joint conditions certainly affect the rock breakage process. On the basis of a large number
of case histories, Bruland in 1998 concluded that, with the decrease of joints spacing, the
TBM penetration increases distinctly. It should be noted that in the Karaj–Tehran water
conveyance tunnel the regional jointing pattern was not separated from shear zones and
faults; therefore, combined effects of faults, shear zones, fissures and joints on TBM
penetration rate were evaluated. In addition to joint spacing, in foliated/bedded rock,
foliation can play a significant role in rock fracture propagation between cuts, depending
on foliation direction with respect to the direction of machine advances. When the
machine advances parallel to foliation planes (Figure 1), crack propagation is forced to
occur across the foliation planes. This reduces machine penetration because of increased
difficulty of rock breakage. When the foliation is perpendicular to the direction of
machine advance, rock failure occurs along the foliation planes as shown in Figure 2.
This case generally represents the most favourable boreability as the foliation planes
assist crack initiation and growth between adjacent cuts. Thus, beta angle (β) (the angle
between the plane of weakness and the tunnel axis) was calculated as a follows (Lislerud,
1988; Bruland, 1999).
β = arcsin[sin β f .sin( β t − β s )] (1)

where βs is the strike and βf is the dip of the fracture; and βt is the bearing of tunnel axis.
Hence, it can be concluded that each parameter has an important effect on the TBM
penetration rate. Although machine characteristics such as thrust or torque and power are
very important for overall performance and scheduling, it was assumed here that these
characteristics remain unchanged. In other words, all possible effects were directly
caused by the geological conditions. Also, in the case of ground composed of highly
254 A. Salimi and M. Esmaeili

fractured rock condition or complex formation, the properties of the ground tend to be the
most effective properties. Therefore, all of the mentioned parameters were chosen as
input parameters for both ANN, linear and non-linear multiple regressions. Descriptive
statistical distribution of variables in the database for the generated models is summarised
in Table 3.

Figure 1 Cutting direction parallel to foliation

Figure 2 Cutting direction perpendicular to foliation

Table 3 Descriptive statistics of the generated database for this study

Variables Minimum Maximum Mean Std. deviation


UCS (MPa) 101.07 130.59 119.94 7.718
BTS (MPa) 1.81 18.16 9.36 3.242
DPW (m) 0.05 0.8 0.24 0.237
PSI (KN/mm) 1.48 6.55 2.96 1.375
Beta (degree) 2.23 83.19 35.39 23.588
ROP (m/h) 1.22 4.78 2.79 0.874
UCS = Uniaxial Compressive Strength; BTS = Brazilian Tensile Strength;
DPW = Distance between plan of weakness; PSI = Peak slope index;
Beta = The angle between plane of weakness and tunnel axis; ROP = Rate of Penetration.
Utilising of linear and non-linear prediction tools for evaluation 255

4 Modelling

4.1 Multiple linear regression analysis


Multiple regression, a time-honoured technique going back to Pearson’s 1908 use of it,
is employed to predict the variance in an interval dependency, based on linear
combinations of interval, dichotomous, or dummy independent variables. The general
purpose of MR is to learn more about the relationship between several independent or
predictor variables and a dependent or criterion variable. The MR equation takes the form
y = b1x1 + b2x2 + … + bnxn + c, where b1, b2, …, bn are the regression coefficients.
Parameter c is a constant, representing the value of y when all independent variables are
0. The standardised versions of the b coefficient are the beta weights, and the ratio of the
beta coefficient is the ratio of the relative predictive power of the independent variables.
Many researchers have applied this particular method in various fields of science and
technology (Khandelwal and Singh, 2006; Monjezi et al., 2009). Here, variations
in the penetration rate of the TBM are assumed to be related to the five mentioned
parameters. In order to obtain the correlations between rock properties and actual
measured Rate of Penetration (ROP), firstly, stepwise statistical analysis and the
influence of each engineering rock parameters on ROP were carried out using SPSS
version 17. Then, five engineering rock properties and measured ROP were set as
independent and dependent variables, respectively. The regression model for predicting
the penetration rate is as follow:
ROP = −4.316 + 0.05 UCS + 0.05 BTS + 0.183 PSI
(2)
− 0.005β − 0.79 DPW R 2 = 0.54.

Figure 3 compares the predicted penetration rates by linear regression equation with
measured values and indicates differences between them.

Figure 3 Comparison of measured and predicted penetration rate by linear regression model
(see online version for colours)
256 A. Salimi and M. Esmaeili

4.2 Multiple non-linear regression analyses


Recently, the non-linear multiple regression analyses have been utilised in rock
engineering and engineering geology to solve complex problems (Gokceoglu et al.,
2009a, 2009b; Nefeslioglu et al., 2008; Yagiz et al., 2009). Caused by most of the
problems in geology involve complex and interacting forces, considering the nature of
the problems for prediction of ROP, the non-linear multiple regression models, which is a
method of finding a non-linear model of the relationship between the dependent variable
and a set of independent variables, were performed for the prediction of penetration rate
of TBM using SPSS version 17. Unlike traditional linear regression, non-linear
regression allows considering changes in multiple properties simultaneously and can
estimate a model with arbitrary relationships between independent and dependent
variables. In this study, equations in Table 4 were used to calculate the multiple non-
linear regression equation in order to predict penetration rate.

Table 4 Nine non-linear regression equations are performed by SPSS

Logarithmic regression y = a + b1 ln x

b1
Inverse regression y=a+
x
Quadric regression y = a + b1 x + b2 x 2

Cubic regression y = a + b1 x + b2 x 2 + b3 x 3

Power regression y = axb1

Compound regression y = ab1x


b 
a+ 1 
S-curve regression y=e  x

Growth regression y = e a + b1 x

Exponential regression y = aeb1 x

The most relevant equations between penetration rate of the TBM and input parameters
have been determined separately, then assembled in order to correct the coefficients.
Afterwards the non-linear equation related to penetration rate was obtained as:
2.9
ROP = −3.318 + 2e −0.52DPW + 0.016 β − 0.0002 β 2 + 2.04 PSI −
BTS (3)
 268.3 
3.22 −  
+ 0.75e  UCS 
R = 0.78.
2

The graph of the scatter plot between measured and predicted penetration rate by
non-linear regression is illustrated in Figure 4.
Utilising of linear and non-linear prediction tools for evaluation 257

Figure 4 Comparison of measured and predicted penetration rate by non-linear regression model
(see online version for colours)

4.3 Artificial Neural Network


Neural networks function is similar to the biological structure of human brains. They are
layered structure networks with highly interconnected processing elements (neurons) that
exist in the network layers. ANN has the ability to learn from the pattern acquainted
before. Once the network has been trained, with sufficient number of sample datasets,
it can make predictions, on the basis of its previous learning, about the output related to
new input dataset of similar pattern. To do so, the connections or weights between
the elements are modified by extracting a generalised correlation available in the
inputs–outputs. During the learning phase, the experimental examples are used as signals
for input and output layers. After learning, in the recall phase, prediction can be made for
new inputs (Wasserman, 1989; Nielsen, 1998). In this method, the output signals from
one layer, which are adjusted by weighting factors, are transmitted to subsequent layers.
In this way, the net input to each element is the sum of the weighted output of the
elements in the former layer. An activation function such as sigmoidal logistic function is
used to calculate the output of the elements. Both the input and output have to be
normalised to maintain the values in the range 0–1. The number of hidden layers and
attributed elements depend on the complexity of the problem to be solved.
Till date, many learning algorithms have been developed for neural networks. It is a
well established fact that the Back-Propagation (BP) method is the most efficient
technique for learning in multi-layer neural networks (Neaupanea and Achet, 2004). This
type of the network consists of at least three layers: input layer, hidden layer and output
layer (Anderson, 1995; Chauvin and Rumelhart, 1995). In the BP algorithm, the learning
phase includes a forward pass and a reverse pass. In the forward pass, a set of
input–output pairs is introduced to the model and then the output related to the initiated
patterns is calculated by the model at the end of this pass. In the reverse pass, the
calculated output is compared with that of target pattern. If the obtained difference (error)
is lower than a predefined threshold, the learning phase is finished. Otherwise, the error is
back propagated through the network, which results in adjustment of connection weights
(Alsmadi et al., 2009).
258 A. Salimi and M. Esmaeili

In the present study, 46 datasets were collected. 33 of them were chosen for training
and 13 for testing. Since there is no precise way to determine the most appropriate
network architecture, a trial-and-error procedure is typically used to identify the best
network for a particular problem. Eventually, the optimum network architecture with
5-17-10-1 was selected. This network is shown in Figure 5. A graphic comparison of the
measured and predicted ROP is shown in Figure 6. As seen in this figure, a good
conformity exists between the measured and predicted ROP by the ANN method.

Figure 5 Suggested ANN for predicting ROP (see online version for colours)

Figure 6 Comparison of measured and predicted penetration rate by the ANN model (see online
version for colours)

5 Results and discussion

In this section, prediction performance of ANN, linear and non-linear models has been
compared. The performance of the models was assessed by means of correlation
coefficient (R2), Root Mean Square Error (RMSE) and Mean Absolute Percentage Error
Utilising of linear and non-linear prediction tools for evaluation 259

(MAPE) (Alvarez Grima and Babuska, 1999; Finol et al., 2001; Gokceoglu, 2002;
Yilmaz and Yuksek, 2008, 2009).
Root Mean Square Error (RMSE), a measure of the goodness-of-fit, best describes an
average measure of the error in predicting the dependent variable. However, it does not
provide any information on phase differences.

1 n
RMSE = ∑
n i =1
( Aimeas − Aipred ) 2 (4)

where Aimeas is the ith measured element, Aipred is the ith predicted element and N is the
number of samples.
MAPE, which is a measure of accuracy in a fitted series value in statistics, was also
used for comparison of the prediction performances of the models. MAPE usually
expresses accuracy as a percentage:

1 n Aimeas − Aipred
MAPE = ∑ A
n i =1
× 100. (5)
imeas

The results of applying these models are compared in Table 5. The table clearly shows
that the prediction performance of the ANN model is higher than that of linear and
non-linear regressions. Figure 7 shows the comparison between measured and predicted
penetration rate; the results clearly demonstrate the superiority of the ANN model over
regression models.
Linearity hypothesis applied may be the cause of poor efficiency of the linear
statistical method in comparison with non-linear and ANN models. So, applying
non-linear tools is necessary to predict penetration rate of the TBM in a hard rock
condition. The effects of input parameters to predict ROP were compared with results of
previously obtained by different researchers. It was seen that, as a result of regression
analyses (linear and non-linear), there was an agreement between this study and previous
studies. It is worth to note that, generally, penetration rate of TBM decreases with an
increase of UCS and BTS. If the brittleness of rock increases, TBM penetration increases
as well, due to easy chipping when the machine goes forward to the rock mass.
As the spacing between the fracture increases, then the TBM penetration increases as
well; however, if the fracture spacing is really close or less than 0.2 m, then the machine
performance decreases due to downtime increment for support and other effects. The
relationship between ROP and β angle is almost consistent with the results of the field
study by Bruland (1998), Yagiz (2008) and numerical simulations by Gong et al. (2005).
The ROP of TBM increases with increment of β angle from 0 to 45º and then decreases;
ROP reaches a maximum at 45º in this case.

Table 5 Performance indices of models

Predictor models R2 RMSE MAPE (%)


Linear regression 0.54 0.54 40.2
Non-linear regression 0.78 0.4 25
Artificial Neural Network 0.83 0.38 20.9
260 A. Salimi and M. Esmaeili

Figure 7 Comparison of measured and predicted penetration rate (see online version for colours)

6 Conclusions

In the current study, linear, non-linear regression and ANN were investigated. During the
analyses procedure, five possible variables (UCS, BTS, PSI, DPW and β angle) were
applied to predict TBM penetration rate. A new ANN model with the architecture
5-17-10-1 was established to predict penetration rate in the Karaj–Tehran tunnel. The
RMSE, R2 and MAPE indices have been calculated as 0.38, 0.83 and 20.9% for the ANN
model, 0.4, 0.78 and 25% for the non-linear regression model and 0.54, 0.54 and 40.2%
for the linear regression model, respectively. As a result, the ANN model seems to of
competence to predict penetration rate of the TBM. In this study, the input data for the
analyses are only the sedimentary rocks encountered along the Karaj–Tehran water
conveyance tunnel in Iran, thus the application of the suggested formulae is limited to
similar geological settings.

References
Alsmadi, M.K.S., Omar, K.B. and Noah, S.A. (2009) ‘Back propagation algorithm: the best
algorithm among the multi-layer perceptron algorithm’, Int. J. Comput. Sci. Netw. Secur.,
Vol. 9, No. 4, pp.378–383.
Alvarez Grima, M. and Babuska, R. (1999) ‘Fuzzy model for the prediction of unconfined
compressive strength of rock samples’, International Journal of Rock Mechanics and Mining
Sciences, Vol. 36, No. 3, pp.339–349.
Alvarez Grima, M., Brunies, P.A. and Verhoef, P.N.W. (2000) ‘Modeling tunnel boring machine
by neuro-fuzzy methods’, Tunneling and Underground Space Technology, Vol. 15, No. 3,
pp.259–269.
Anderson, J.A. (1995) An Introduction to Neural Networks, MIT Press, Cambridge.
Barton, N. (1999) ‘TBM performance estimation in rock using QTBM’, Tunnels Tunnelling Int.,
Vol. 31, No. 9, pp.41–48.
Utilising of linear and non-linear prediction tools for evaluation 261

Barton, N. (2000) TBM Tunneling in Jointed and Faulted Rock, Balkema, Rotterdam.
Bejari, H., Kakaie, R., Ataei, M. and Khademi Hamidi, J. (2011) ‘Simultaneous effects of joint
spacing and joint orientation on the penetration rate of a single disc cutter’, Mining Sci.
Technol. (China), Vol. 21, No. 4, pp.507–512.
Benardos, A.G. and Kaliampakos, D.C. (2004) ‘Modeling TBM performance with artificial neural
networks’, Tunneling and Underground Space Tech., Vol. 19, No. 6, pp.597–605.
Bieniawski, Z.T. (1973) ‘Engineering classification of jointed rock masses’, Trans. S. Afr. Inst. Civ.
Eng., Vol. 15, pp.335–344.
Brown, E. (1981) Rock Characterization Testing and Monitoring, ISRM Suggested Methods.
Bruland, A. (1998) Hard Rock Tunnel Boring, Doctoral Thesis, Norwegian University of Science
and Technology, Trondheim.
Bruland, A. (1999) Hard Rock Tunnel Boring: Advance Rate and Cutter Wear, Norwegian Institute
of Technology (NTNU), Trondheim, Norway.
Cassinelli, F., Cina, S., Innaurato, N., Mancini, R. and Sampaolo, A. (1982) ‘Power consumption
and metal wear in tunnel-boring machines: analysis of tunnel boring machine operation in
hard rock’, Tunneling, Vol. 82, pp.73–81.
Chauvin, Y. and Rumelhart, D.E. (1995) Backpropagation: Theory, Architectures and
Applications, Erlbaum, Mahwah, NJ.
Cigla, M., Yagiz, S. and Ozdemir, L. (2001) ‘Application of tunnel boring machines in
underground mining development’, 17th Int. Mining Congress & Exhibition of Turkey,
pp.155–164.
Dollinger, G.L. and Handewith, H.J. (1998) ‘Use of the punch test forestimating TBM
performance’, Tunnel. Underground Space Technol., Vol. 13, No. 4, pp.403–408.
Farmer, I.W. and Glossop, N.H. (1980) ‘Mechanics of disc cutter penetration’, Tunnels Tunnell.
Int., Vol. 12, No. 6, pp.22–25.
Finol, J., Guo, Y.K. and Jing, X.D. (2001) ‘A rule based fuzzy model for the prediction of
petrophysical rock parameters’, Journal of Petroleum Science and Engineering, Vol. 29,
No. 2, pp.97–113.
Gokceoglu, C. (2002) ‘A fuzzy triangular chart to predict the uniaxial compressive strength of the
Ankara agglomerates from their petrographic composition’, Engineering Geology, Vol. 66,
No. 1, pp.39–51.
Gokceoglu, C. and Zorlu, K. (2004) ‘A fuzzy model to predict the uniaxial compressive strength
and modulus of elasticity of a problematic rock’, Engineering Applications of Artificial
Intelligence, Vol. 17, No. 1, pp.61–72.
Gokceoglu, C., Sonmez, H. and Zorlu, K. (2009a) ‘Estimating the uniaxial compressive strength of
some clay bearing rocks selected from Turkey by nonlinear multivariable regression and
rule-based fuzzy models’, Expert Systems, Vol. 26, No. 2, pp.176–190.
Gokceoglu, C., Zorlu, K., Ceryan, S. and Nefeslioglu, H.A. (2009b) ‘A comparative study on
indirect determination of degree of weathering of granites from some physical and strength
parameters by two soft computing techniques’, Materials Characterization, Vol. 60, No. 11,
pp.1317–1327.
Gong, Q.M. and Zhao, J. (2007) ‘Influence of rock brittleness on TBM penetration rate
in Singapore granite’, Tunnelling and Underground Space Technol., Vol. 22, No. 3,
pp.317–324.
Gong, Q.M. and Zhao, J. (2009) ‘Development of a rock mass characteristics model for TBM
penetration rate prediction’, Int. J. Rock Mech. Min. Sci., Vol. 46, No. 1, pp.8–18.
Gong, Q.M., Zhao, J. and Jiao, Y.Y. (2005) ‘Numerical modeling of the effects of joint orientation
on rock fragmentation by TBM cutters’, Tunnell. Undergr. Space Technol., Vol. 20, No. 2,
pp.183–191.
262 A. Salimi and M. Esmaeili

Graham, P.C. (1976) ‘Rock exploration for machine manufacturers’, in Bieniawski, Z.T. (Ed.):
Exploration for Rock Engineering, Balkema, Johannesburg, pp.173–180.
Handewith, H.J. (1970) ‘Predicting the economic success of continuous tunnelingand hard rock’,
71st Annual General Meeting of the CIM, Vol. 63, pp.595–599.
Hughes, H.M. (1986) ‘The relative cuttability of coal measures rock’, Min. Sci. Technol., Vol. 3,
pp.95–109.
Kahraman, S., Altun, H., Tezekici, B.S. and Fener, M. (2006) ‘Sawability prediction of carbonate
rocks from shear strength parameters using artificial neural networks’, International Journal
of Rock Mechanics and Mining Sciences, Vol. 43, pp.157–164.
Khandelwal, M. and Singh, T.N. (2006) ‘Prediction of blast induced ground vibrations
and frequency in opencast mine – a neural network approach’, J. Sound Vib., Vol. 289,
pp.711–725.
Lee, I.M. and Lee, J.H. (1996) ‘Predicting of pile bearing capacity using artificial neural networks’,
Computer and Geotechnics, Vol. 18, No. 3, pp.189–200.
Lislerud, A. (1988) ‘Hard rock tunnel boring: prognosis and costs’, Tunneling and Underground
Space Technology, Vol. 3, No. 1, pp.9–17.
McFeat-Smith, I. (1977) ‘Rock property testing for the assessment of tunneling machine
performance’, Tunnels and Tunneling, pp.29–33.
Monjezi, M., Rezaei, M. and Yazdian, A. (2009) ‘Prediction of backbreak in open-pit blasting
using fuzzy set theory’, J. Expert Syst. Appl., Vol. 37, pp.2637–2643.
Neaupanea, K.M. and Achet, S.H. (2004) ‘Use of backpropagation neural network for landslide
monitoring: a case study in the higher Himalaya’, J. Eng. Geol., Vol. 74, pp.213–226.
Nefeslioglu, H.A., Gokceoglu, C. and Sonmez, H. (2003) ‘A Mamdani model to predict the
weighted joint density’, Lecture Notes in Artificial Intelligence, Vol. 2773, pp.1052–1057.
Nefeslioglu, H.A., Gokceoglu, C. and Sonmez, H. (2006) ‘Indirect determination of weighted joint
density (wJd) by empirical and fuzzy models: Supren (Eskisehir, Turkey) marbles’,
Engineering Geology, Vol. 85, pp.251–269.
Nefeslioglu, H.A., Gokceoglu, C. and Sonmez, H. (2008) ‘An assessment on the use of logistic
regression and artificial neural networks with different sampling strategies for the preparation
of landslide susceptibility maps’, Engineering Geology, Vol. 97, Nos. 3–4, pp.171–191.
Nelson, P.P. and O’Rourke, T.D. (1983) Tunnel Boring Machine Performance in Sedimentary
Rocks, Report to Goldberg-Zoino Associates of New York, PC, by School of Civil and
Environmental of Civil Engineering, Cornell University, Ithaca, NY.
Nielsen, R.H. (1998) ‘Neurocomputing’, Picking the Human Brain, IEEE, Vol. 3, pp.36–41.
Okubo, S., Kfukie, K. and Chen, W. (2003) ‘Expert systems for applicability of tunnel boring
machine in Japan’, Rock Mechanics and Rock Engineering, Vol. 36, No. 4, pp.305–322.
Osanloo, M. (1998) Drilling methods, Tehran, Sadra.
Ozdemir, L. (1977) Development of Theoretical Equations for Predicting Tunnel Borability,
PhD Thesis, T-1969, Colorado School of Mines, Golden, CO, USA.
Ozdemir, L. (2002) Personnel Communication, Earth Mechanics Institute of Colorado School of
Mines, Golden, CO, USA.
Ribacchi, R. and Lembo-Fazio, A. (2005) ‘Influence of rock mass parameters on the performance
of a TBM in a Gneissic formation (Varzo tunnel)’, Rock Mech. Rock Eng., Vol. 38, No. 2,
pp.105–127.
Rostami, J. and Ozdemir, L. (1993) ‘A new model for performance prediction of hard rock TBM’,
Proceedings of the Rapid Excavation and Tunnelling Conference, Society for Mining and
Metallurgy and Exploration Inc., Boston, pp.793–809.
Sapigni, M., Berti, M., Behtaz, E., Busillo, A. and Cardone, G. (2002) ‘TBM performance
estimation using rock mass classification’, Int. J. Rock Mech. Min. Sci. Geomech., Vol. 39,
pp.771–788.
Utilising of linear and non-linear prediction tools for evaluation 263

Sonmez, H., Gokceoglu, C., Nefeslioglu, H.A. and Kayabasi, A. (2006) ‘Estimation of rock
modulus: for intact rock with an artificial neural network and for rock masses with a new
empirical equation’, International Journal of Rock Mechanics and Mining Sciences, Vol. 43,
pp.224–235.
Szwedzicki, T. (1998) ‘Draft ISRM suggested method for determining theindentation hardness
index of rock materials’, Int. J. Rock Mech. Min. Sci., Vol. 35, No. 6, pp.831–835.
Tarkoy, P.J. (1975) Rock Hardness Index Properties and Geotechnical Parameters for
Predicting Tunnel Boring Machine Performance, PhD Thesis, University of Illinois
at Urbana-Champaign.
Tiryaki, B. (2008) ‘Application of artificial neural networks for predicting the cuttability of
rocks by drag tools’, Tunneling and Underground Space Technology, Vol. 23, No. 3,
pp.273–280.
Wasserman, P.D. (1989) Neural Computing Theory and Practice, Van Nostrand Reinhold,
New York.
Yagiz, S. (2002) Development of Rock Fracture and Brittleness Indices to Quantify the Effects of
Rock Mass Features and Toughness in the CSM Model Basic Penetration for Hard Rock
Tunneling Machines, PhD Thesis, Department of Mining and Earth Systems Engineering,
Colorado School of Mines, Golden, Colorado, USA.
Yagiz, S. (2006) ‘TBM performance prediction based on rock properties’, Proceedings of
Multiphysics Coupling and Long Term Behavior in Rock Mechanics, EUROCK’06,
pp.663–670.
Yagiz, S. (2008) ‘Utilizing rock mass properties for predicting TBM performance in hard rock
condition’, Tunneling and Underground Space Technology, Vol. 23, pp.326–339.
Yagiz, S., Gokceoglu, C., Sezer, E. and Iplikci, S. (2009) ‘Application of two non-linear prediction
tools to the estimation of tunnel boring machine performance’, Engineering Applications of
Artificial Intelligence, Vol. 22, pp.808–814.
Yagiz, S., Rostami, J. and Ozdemir, L. (2008) ‘Recommended rock testing methods for predicting
TBM performance: focus on the CSM and NTNU Models’, Proceedings of the ISRM
International Symposium, 5th ASRM, Tehran, Iran, pp.1523–1530.
Yilmaz, I. and Yuksek, A.G. (2008) ‘An example of artificial neural network application for
indirect estimation of rock parameters’, Rock Mechanics and Rock Engineering, Vol. 41,
No. 5, pp.781–795.
Yilmaz, I. and Yuksek, A.G. (2009) ‘Prediction of the strength and elasticity modulus of gypsum
using multiple regression, ANN, ANFIS models and their comparison’, International Journal
of Rock Mechanics and Mining Sciences, Vol. 46, No. 4, pp.803–810.
Yoo, C. and Kim, J. (2007) ‘Tunneling performance prediction using an integrated GIS and neural
network’, Computer and Geotechnics, Vol. 34, pp.19–30.

Bibliography
Aeberli, U. and Wanner, W.J. (1978) ‘On the influence of discontinuities at the application
of tunneling machines’, Proceedings of the3rd International Congress IAEG, Madrid,
pp.7–14.
Buchi, E. (2007) The Influence of Joints on TBM Performance, Private Communication.
Chandok, J.S., Kar, I.N. and Tuli, S. (2008) ‘Estimation of furnace exit gas temperature (FEGT)
using optimized radial basis and backpropagation neural networks’, J. Rock Energy Convers.
Manag., Vol. 49, pp.1989–1998.
Cheema, S (1999) Development of a Rock Mass Boreability Index for the Performance of Tunnel
Boring Machines, PhD Dissertation, Colorado School of Mines, Golden, CO.
264 A. Salimi and M. Esmaeili

Howarth, D.F. (1981) ‘The effect of jointed and fissured rock on the performance of tunnel
boring machines’, Proceedings of the International Symposium on Weak Rock, Tokyo,
pp.1069–1074.
O’Rourke, J.E., Spring, J.E. and Coudray, S.V. (1994) ‘Geotechnical parameters and tunnel boring
machine performance at Good will Tunnel, California’, in Nelson, P.P. and Laubach, S.E.
(Eds.): Proceedings of the 1st North American Rock Mechanics Symposium, Rotterdam,
Balkema, Austin, Texas.
Thuro, K. and Plinninger, R.J. (2003) ‘Hard rock tunnel boring, cutting, drilling and blasting:
rock parameters for excavatability’, Proceedings of the 10th International Congress ISRM,
South African Institute on Mineral Metallurgy, pp.1–7.

You might also like