See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/259164272

A comparative study on the application of various artificial neural networks to

simultaneous prediction of rock fragmentation and backbreak

Article  in  Journal of Rock Mechanics and Geotechnical Engineering · August 2013

DOI: 10.1016/j.jrmge.2013.05.007

CITATIONS READS

34 1,719

4 authors:

Ahmad Reza Sayadi Masoud Monjezi

Tarbiat Modares University Tarbiat Modares University
37 PUBLICATIONS   260 CITATIONS    79 PUBLICATIONS   2,226 CITATIONS   

SEE PROFILE SEE PROFILE

Nemat Talebi Manoj Khandelwal

Islamic Azad University, South Tehran Branch Federation University Australia
21 PUBLICATIONS   169 CITATIONS    86 PUBLICATIONS   2,159 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Desgin and implementation of productivity cycle in a mining firm View project

Application of Artificial Intelligence (AI) Techniques in Blasting Operations in Mines View project

All content following this page was uploaded by Nemat Talebi on 07 September 2014.

The user has requested enhancement of the downloaded file.

 Journal of Rock Mechanics and Geotechnical Engineering 5 (2013) 318–324

Journal of Rock Mechanics and GeotechnicalEngineering

Journal of Rock Mechanics and Geotechnical

Engineering
journal homepage: www.rockgeotech.org

A comparative study on the application of various artificial neural networks to

simultaneous prediction of rock fragmentation and backbreak
A. Sayadi a , M. Monjezi b , N. Talebi a , Manoj Khandelwal c,∗
a
Islamic Azad University, Tehran South Branch, Tehran, Iran
b
Faculty of Engineering, Tarbiat Modares University, Tehran, Iran
c
Maharana Pratap University of Agriculture and Technology, Udaipur, India

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

Article history: In blasting operation, the aim is to achieve proper fragmentation and to avoid undesirable events such
Received 27 March 2012 as backbreak. Therefore, predicting rock fragmentation and backbreak is very important to arrive at a
Received in revised form 5 July 2012 technically and economically successful outcome. Since many parameters affect the blasting results in a
Accepted 25 July 2012
complicated mechanism, employment of robust methods such as artificial neural network may be very
useful. In this regard, this paper attends to simultaneous prediction of rock fragmentation and backbreak
Keywords:
in the blasting operation of Tehran Cement Company limestone mines in Iran. Back propagation neural
Rock fragmentation
network (BPNN) and radial basis function neural network (RBFNN) are adopted for the simulation. Also,
Backbreak
Artificial neural network
regression analysis is performed between independent and dependent variables. For the BPNN modeling,
Back propagation a network with architecture 6-10-2 is found to be optimum whereas for the RBFNN, architecture 6-
Radial basis function 36-2 with spread factor of 0.79 provides maximum prediction aptitude. Performance comparison of the
developed models is fulfilled using value account for (VAF), root mean square error (RMSE), determination
coefficient (R2 ) and maximum relative error (MRE). As such, it is observed that the BPNN model is the
most preferable model providing maximum accuracy and minimum error. Also, sensitivity analysis shows
that inputs burden and stemming are the most effective parameters on the outputs fragmentation and
backbreak, respectively. On the other hand, for both of the outputs, specific charge is the least effective
parameter.
© 2013 Institute of Rock and Soil Mechanics, Chinese Academy of Sciences. Production and hosting by
Elsevier B.V. All rights reserved.

1. Introduction Djordjevic, 2005; Monjezi et al., 2009). On the other hand, the
blasting operation usually is accompanied by various unwanted
Backbreak is one of the undesirable phenomena in the blasting phenomena such as backbreak. Backbreak is the fractured zone
operation. In other words, a blast without any unwanted effects beyond the last blasting row (Jimeno et al., 1995). Occurrence
can be evaluated as a successful activity, and in such activity, a of this phenomenon is an indication of wasting potential explo-
large proportion of the available energy has been consumed in sive energy. Moreover, it has some other hazardous effects such
the right direction, i.e. rock fragmentation. Rock fragmentation as slope instability. Therefore, remedial measures should be pre-
can be considered as the main objective of the blasting opera- sented for diminishing and/or omitting backbreak. The effective
tion. Size distribution of the rock fragments is very important on blast design parameters are (1) blasting pattern components, (2)
the overall mining and processing plant economics (Michaux and rock mass geomechanical properties, and (3) explosive specifica-
tions (Thornton et al., 2002; Zhu et al., 2007, 2008). Implementation
of a suitable blasting pattern, as a controllable parameter, is very
∗ Corresponding author. Tel.: +91 294 2471379. important in preventing backbreak and achieving proper fragmen-
E-mail address: mkhandelwal1@gmail.com (M. Khandelwal). tation (Monjezi and Dehghani, 2008). Gates et al. (2005) pointed
Peer review under responsibility of Institute of Rock and Soil Mechanics, Chinese out that the backbreak is increased when inappropriate delay tim-
Academy of Sciences. ing is applied. Many researchers believe that excessive burden
is the main cause of the backbreak and producing oversize rock
fragments (Konya and Walter, 1991; Konya, 2003). To date, sev-
eral empirical models have been developed to predict the blasting
results. However, complicated nature of the problem due to mul-
1674-7755 © 2013 Institute of Rock and Soil Mechanics, Chinese Academy of tiplicity of the effective parameters has caused development of
Sciences. Production and hosting by Elsevier B.V. All rights reserved.
simplified prediction models with limited number of independent
http://dx.doi.org/10.1016/j.jrmge.2013.05.007
 A. Sayadi et al. / Journal of Rock Mechanics and Geotechnical Engineering 5 (2013) 318–324 319

fragmentation as the model outputs. Fig. 2 shows the undesirable

backbreak after blasting in mines.
It is noted that, for determining fragmentation quality, image
processing method is employed. As such, 80% passing size (D80)
is considered as the fragmentation evaluation index. Variations of
the input and output parameters are given in Table 2. In this study,
103 datasets are collected from practical blasting operations of the
mines. The available datasets are grouped into training and testing
datasets. For this, using sorting mechanism, 10% of the datasets are
kept apart for testing and evaluating of the simulations.

3. Statistical analysis

Multivariate regression analysis (MVRA) is an extension of

regression analysis, which was firstly employed by Pearson in
1908 (Yilmaz and Yuksek, 2009). This method can easily be used
for determining the linear and/or nonlinear relationship between
dependent predictive and independent criterion variables. The
Fig. 1. Tehran cement company limestone mines.
main form of MVRA is

variables. The simplification assumptions are the main cause of Y = ˇ0 + ˇ1 x1 + ˇ2 x2 + · · · + ˇn xn (1)

poor performance of the empirical models. Moreover, simulta-
where ˇ1 , ˇ2 ,. . ., ˇn are the coefficients of regression model; ˇ0 is
neous prediction of backbreak and fragmentation is not possible
a constant value; Y is the dependent variable; and x1 , x2 ,. . ., xn are
using previously developed models. In order to overcome short-
the independent variables.
comings of the empirical models, artificial intelligence (AI) based
Two MVRA models are developed to predict backbreak and frag-
methods can effectively be applied to solving complicated prob-
mentation considering input parameters given in Table 2.
lems. Some of the most popular AI paradigms are artificial neural
Eqs. (2) and (3) show mathematical formulations of the
network (ANN), fuzzy inference system (FIS) and genetic algorithm
developed models for predicting backbreak and fragmentation,
(GA).
respectively. Also, statistical details of the MVRA models are sum-
ANN has capability of learning, evoking and generalizing from
marized in Table 3.
the given patterns (Cheng and Ko, 2006). Its high performance
in solving complicated problems has made this technique so BB = 0.494B + 1.082S + 0.015H + 1.203T − 0.056SC + 23.576SD − 8.501 (2)
applicable. Various applications of the ANN method in rock engi- Fr = 0.371B + 0.215S − 0.012H + 0.182T − 0.025SC + 6.45SD − 1.959 (3)
neering have been reported in the literature (Cai and Zhao, 1997;
Yang and Zhang, 1997a, 1997b; Maulenkamp and Grima, 1999; 4. Basis of artificial neural network
Benadros and Kaliampakos, 2004; Ermini et al., 2005). Also, sev-
eral researchers have implemented the method in the field of ANN is a subsystem of AI. This computational system is a sim-
mine blasting (Khandelwal and Singh, 2005, 2006, 2007, 2009; ulation of human brain (Maulenkamp and Grima, 1999). Original
Bakhshandeh et al., 2010; Kulatilake et al., 2010; Khandelwal, 2010, ANN was introduced by McCulloch and Pitts (1943), and since then
2012; Monjezi et al., 2010). it was popular and applicable to various fields of science and tech-
In this paper, an attempt has been made to simultaneously pre- nology to solve complicated problems. Capabilities of the technique
dict backbreak and fragmentation due to blasting operation in the are calculating arithmetic and logical functions, generalizing and
Tehran Cement Company limestone mines using ANN method. transforming independent variables to the dependent variables,
parallel computations, nonlinearity processing, handling imprecise
2. Case study or fuzzy information, function approximation and pattern recogni-
tion.
Tehran Cement Company limestone mines, i.e. Bibishahrbanoo, ANN is trained using a set of real inputs and their corresponding
Nesari and Safaie, are located at the southeast of Tehran. These outputs. For a better approximation, sufficient number of datasets
mines are under development and have total proved limestone is required. Performance of the trained model is checked with part
deposits of 41.3 million tons. From the geological point of view, of the available data known as testing datasets. To find out the best
these mines are situated in the sedimentary rocks of Cretaceous possible network, various topologies are constructed and tested.
period. The limestone layers with an eastwest extension have 75◦ The process of model training-testing has to be continued until the
dip to the north. Limestone is the main exposure layer in the optimum model with minimum error and maximum accuracy is
area while in some parts black shale and cream marl are also achieved. ANN training-testing (Monjezi and Dehghani, 2008) is
observed. The Nesari mine is located 10 km northeast of Tehran illustrated in Fig. 3.
Cement Company. Layers of dolomite and dolomitic limestone are A neural network has a layered structure, and each layer con-
observed in this mine in a narrow strip formation. Safaie Moun- tains processing units or neurons. Problem effective variables
tain is also located in the northwest of Bibishahrbanoo Mountain are placed in the input layer, whereas objectives or dependent
(Fig. 1). variables are put in the last (output) layer. The computation compo-
The blasting pattern specifications of limestone mines are pre- nents (black box) of the system are the neurons of hidden layers. All
sented in Table 1. Mean fragment size of 45 cm is suitable for the of the layers are connected to each other by weighted connections.
mine primary crusher. Fig. 4 shows a typical ANN structure. Each neuron is connected to
The controllable parameters of burden, spacing, stemming, the neurons in the subsequent layer. However, there is no connec-
bench height, specific charge and specific drilling are considered as tion between the neurons of the same layer (Demuth and Beale,
inputs to develop an ANN model for predicting backbreak and rock 1994).
320 A. Sayadi et al. / Journal of Rock Mechanics and Geotechnical Engineering 5 (2013) 318–324

Table 1
Blasting pattern specifications of limestone mines.

Main explosive type Secondary explosive type Blasting hole pattern Bench height (m) Hole diameter (mm) Rows per blast Holes per row

ANFO Emulation V-cut (staggered) 15 76 1–4 10–15

Fig. 2. The undesirable backbreak after blasting in mines.

Table 2
Basic statistics of inputs and output parameters.

Burden B (m) Spacing S (m) Hole height H (m) Stemming T (m) Specific charge SC (kg/m3 ) Specific drilling SD (m/m3 ) Back break BB (m) Fragmentation Fr (m)

1.8–4.5 (3.02) 1.7–4.3 (3.35) 2.5–28.5 (17.8) 1.6–3 (2.8) 1.96–28.68 (8.303) 0.063–0.226 (0.109) 1–4 (2.41) 0.37–1.76 (0.82)

Note: Numbers in parentheses represent the average values.

In the training process, the interconnections among the neurons

are initially assigned specific weights. The network would be able
to perform a function by adjusting the initial weights.
A single neuron containing multiple inputs (x1 ,· · ·, xn ) and a sin-
gle output (y) is shown in Fig. 5. In the process of ANN training, an
initial arbitrary value (weight) is assigned to the connections and
then to combine all of the weighted inputs and generate the neuron
output, and the following equation is applied:

O= xi wi + b (4)

where xi is the inputs, wi is the connection weights, and b is the

bias.
To map a neuron net output to its actual output, an activation
function f has to be selected. The transfer function can be expressed
as Fig. 4. Artificial neural network structure.

y = f (O) = f ( xi wi + b) (5)
Applying Eq. (5) to the neuron initial summation output result-
ing from Eq. (4), the neuron final output within a range of [0, 1] or
[−1, 1] is achieved depending on the type of applied transfer func-
tion. It is noted that a single activation function should be selected
for the neurons of a particular layer. Type of the activation func-
tion is fully dependent on nature of the problem to be solved. Also
their respective graphic presentations are shown in Table 4. During

Fig. 3. Artificial neural network training-testing process. Fig. 5. Neuron structure.

 A. Sayadi et al. / Journal of Rock Mechanics and Geotechnical Engineering 5 (2013) 318–324 321

Table 3
Linear regression coefficients for backbreak and fragmentation.

Independent variables Regression coefficient Standard error

Backbreak Fragmentation Backbreak Fragmentation

Constant −8.501 −1.959 7.746 1.882

B 0.494 0.371 0.804 0.194
S 1.082 0.215 0.868 0.213
H 0.015 −0.012 0.035 0.009
T 1.203 0.182 0.72 0.178
SC −0.056 −0.025 0.049 0.014
SD 23.576 6.45 19.33 4.756

the training process, network tries to decrease difference between 4.2. Radial basis function neural network
predicted and real values.
To do so, a specific algorithm is selected by which connection Radial basis function neural network (RBFNN) is one of the effi-
weights and biases are repeatedly updated until the minimum error cient artificial networks. These types of the networks are mostly
is provided. There are various types of training algorithms, such as used for function approximation. However, they can also be applied
back propagation and radial basis (Demuth and Beale, 1994). for pattern recognition and classification. Arriving in very small
errors during training process can be considered as the main
advantage of RBFNN over BPNN (Haykin, 1999; Christodoulou and
4.1. Back propagation neural network Georgiopoulos, 2001). Unlike BPNN, in the structure of RBFNN,
there is only one hidden layer that makes computation time very
Back propagation neural network (BPNN) normally has a multi- less. Moreover, transfer function ϕ of the hidden layer is always of
layer structure with one or more nonlinear hidden layer and a linear the Gaussian type:
output layer. It is widely used as a predicting tool in various fields of  
geo-sciences. Generally, in BPNN four transfer functions are used as 1
ϕ(P) = exp − ||P − Cj ||2 (7)
presented in Table 4. These networks can be used to make nonlinear 2j
and/or linear correlation between input(s) and output(s).
Various types of functions, such as Newton and gradient where P is the input vector; Cj and  j are the center and extension
descent, can be used for training BPNNs. In the simplest form, (spread factor) of Gaussian function, respectively.
weights and biases are frequently updated to decrease performance As illustrated in Fig. 6a, ϕ(P) reaches the maximum value (1.0)
function. Two different techniques (incremental method and batch when P is equal to 0.0. In this way, when difference between values
method) are implemented in the learning process of the ANN. In the of weights and inputs is lower, the neuron output will be greater. In
incremental method, weights and biases are upgraded after each fact, here the amount of output of hidden layer shows the absolute
input entrance to net but in the batch method upgrading process is difference between connection weights and inputs. In the RBFNN,
done after entrance of all inputs. Generally, performance function the jth network output dj (Demuth and Beale, 1994) can be calcu-
is considered as the mean square error (MSE), which is calculated lated by
by the following equation (Demuth and Beale, 1994; Benadros and
Kaliampakos, 2004):

N
dj = ϕj (P)wij (8)

1
N j=1

(yi − yi )
2
MSE = (6) where ϕj is the jth neuron output, and wij is the output layer weight.
N
i=1 During the training process, parameters Cj , ␴j and wij are deter-
mined by the network to provide the best approximation function.
where N is the number of input–output datasets. In this process, optimum number of neurons required for the hid-
den layer is also determined by the network. The structure of a
RBFN is illustrated in Fig. 6b.
Table 4
BPNN most usual transfer functions.
5. Results and discussion
Transfer function Transfer function diagram
situation
To compare model performance of the regression analysis and
ANN method, value account for (VAF), root mean square error
(RMSE), determination coefficient (R2 ) and maximum relative error
(MRE) are utilized:
Hidden layers  var(y − y )

VAF = 100 1 − (9)
var(y)

 N
1
RMSE = (y − y )2 (10)
N
i=1
Output layer

N 2
2 i=1
(y − ȳ)(y − ȳ )
R = N N (11)
i=1
(y − ȳ)2 i=1
(y − ȳ )2
322 A. Sayadi et al. / Journal of Rock Mechanics and Geotechnical Engineering 5 (2013) 318–324

Fig. 7. The correlation of measured and predicted data with back propagation neural
network.

Fig. 6. (a) Radial basis transfer function (radbas) and (b) structure of a radial basis
function network.

|y − y |

MRE = max 100 (12)
y
where y and y’ are the measured and predicted values, respectively;
ȳ and ȳ are the average measured and average predicted values,
respectively; var(·) is the variance.
Table 5 shows the performance of some of the constructed
BPNN models. As it is observed from this table, BPNN model with
architecture 6-10-2 gives the best result with minimum errors
and maximum accuracy, and is considered as the optimum model
amongst the BPNN models. Also, Table 6 shows the performance of
some of the constructed RBFNN models with various spread factors.
As it is seen from Table 6, the model with spread factor of 0.79 pro-
vides the best results. Furthermore, performance of the regression
model is shown in Table 7. Figs. 7–9 show the correlation between
predicted and measured outputs for the three methods of model-
ing. In Figs. 7–9, dashed line shows 1:1 slope line, where measured
and predicted values will be same. From Tables 5–7 and Figs. 7–9, it Fig. 8. The correlation of measured and predicted data with radial basis function
neural network.
is noted that BPNN modeling shows better prediction capability as
compared to the other applied methods. Superiority of BPNN over
RBFNN was also reported by Monjezi et al. (2010). output parameters. In this method, all the data pairs are defined
as a specific point in m-dimensional space. In this way, each of the
parameters is directly connected to the outputs. The strength of
6. Sensitivity analysis
this relation Rij is calculated by
Cosine amplitude method (CAM) of sensitivity analysis was
m
x x
k=1 ik jk
first introduced by Yang and Zhang (1997a). This technique was Rij = m 2
 m (13)
x x2
employed to find out the most effective input parameters on k=1 ik k=1 jk
 A. Sayadi et al. / Journal of Rock Mechanics and Geotechnical Engineering 5 (2013) 318–324 323

Table 5
The calculated performance indices for back propagation neural network models.

No. Net structure R2 (%) RMSE VAF (%) MRE (%)

BB Fr BB Fr BB Fr BB Fr

1 6-10-2 87.1 84.8 0.2206 0.0673 90.21 95.28 16.58 16.23

2 6-14-2 90.2 86.7 0.3052 0.1294 82.73 86.34 20.2 27.45
3 6-18-2 74.9 65.3 0.4867 0.3581 77.35 74.81 43.5 62.7
4 6-5-5-2 75.6 66.1 0.3955 0.3024 81.47 63.23 36.11 59.89
5 6-10-6-2 88.5 79.5 0.5438 0.4147 70.31 47.36 52.74 68.55
6 6-10-10-2 86.4 73.8 0.367 0.2634 85.34 80.03 27.96 33.57

Table 6
The calculated performance indices for radial basis function neural network models.

No. Spread factor Net structure R2 (%) RMSE VAF (%) MRE (%)

BB Fr BB Fr BB Fr BB Fr

1 0.7 6-36-2 48 53.6 0.461 0.249 89.55 76.12 34.1 38.46

2 0.79 6-36-2 51.5 59.7 0.3108 0.112 88.82 86.53 25 29.63
3 0.8 6-36-2 50.3 58.1 0.354 0.158 93.4 89.6 29.03 32.52

Table 7
The calculated performance indices for multivariate regression analysis model.

R2 (%) RMSE VAF (%) MRE (%)

BB Fr BB Fr BB Fr BB Fr

75.1 30.1 0.672 0.291 34.77 21.73 45.32 60.22

Fig. 10. Sensitivity analysis for backbreak and fragmentation.

7. Conclusions
Fig. 9. The correlation of measured and predicted data with multivariate regression
analysis. Precise prediction of backbreak and fragmentation is very cru-
cial for success of a mining project. In this paper, an attempt is
made to utilize different types of ANNs for predicting simulta-
where xi and xj are inputs and outputs, respectively; and m is the neous fragmentation and backbreak in the blasting operation of
number of all datasets. The larger the Rij is, the higher the influence Tehran Cement Company limestone mines. The ANN models are
of relevant input is. trained using a database including 103 datasets. To achieve more
In Fig. 10, it can be inferred that the stemming and burden are reliable predictive models, parameters including burden, spacing,
the most influential input parameters on the backbreak and frag- stemming, bench height, specific charge and specific drilling are
mentation. It is noted that for both the outputs, specific charge is considered as the model inputs to predict outputs fragmenta-
the least effective parameter. tion and backbreak. BPNN and RBFNN are adopted for this study.
324 A. Sayadi et al. / Journal of Rock Mechanics and Geotechnical Engineering 5 (2013) 318–324

Also, regression analysis is performed between the same indepen- Khandelwal M, Singh TN. Evaluation of blast induced ground vibration predictors.
dent and dependent variables. For the BPNN and RBFNN modeling, Soil Dynamics and Earthquake Engineering 2007;27(2):116–25.
Khandelwal M, Singh TN. Prediction of blast induced ground vibration using artificial
networks with architectures 6-10-2 and 6-36-2 respectively are neural network. International Journal of Rock Mechanics and Mining Sciences
found to be optimum. Efficiency of the developed models is exam- 2009;46(7):1214–22.
ined using testing datasets. Indices of VAF, RMSE, R2 and MRE Khandelwal M. Evaluation and prediction of blast induced ground vibration using
support vector machine. International Journal of Rock Mechanics and Mining
are calculated for predicted outputs and compared with the real Sciences 2010;47(3):509–16.
outputs. It is found that performance of the BPNN model with Khandelwal M. Application of an expert system for the assessment of blast vibration.
maximum accuracy and minimum error is better than that of Geotechnical and Geological Engineering 2012;30(1):205–17.
Kulatilake PHSW, Wu Q, Hudaverdi T, Kuzu C. Mean particle size predic-
the RBFNN and statistical models. Also, it is observed that inputs tion in rock blast fragmentation using neural networks. Engineering Geology
burden and stemming are the most effective parameters on the 2010;114(3):298–311.
outputs, whereas specific charge is the least effective parameter Konya CJ. Rock blasting and overbreak control. 2nd ed. Washington, DC: US Depart-
ment of Transportation, Federal Highway Administration; 2003.
for both the outputs. At the end, it is recommended that hybrid
Konya CJ, Walter EJ. Rock blasting and overbreak control. 1st ed. Washington, DC:
models, combination of fuzzy logic and/or genetic algorithm with US Department of Transportation, Federal Highway Administration; 1991.
neural networks, could be applied for further research. Maulenkamp F, Grima MA. Application of neural networks for the prediction of the
unconfined compressive strength (UCS) from Equotip hardness. International
Journal of Rock Mechanics and Mining Sciences 1999;36(1):29–39.
References McCulloch WS, Pitts WH. A logical calculus of ideas immanent in nervous activity.
Bulletin of Mathematical Biophysics 1943;5(4):115–33.
Bakhshandeh H, Mozdianfard MR, Siamaki A. Predicting of blasting vibra- Michaux S, Djordjevic N. Influence of explosive energy on the strength of the rock
tions in Sarcheshmeh copper mine by neural network. Safety Science fragments and SAG mill throughput. Miner Engineering 2005;18(4):439–48.
2010;48(3):319–25. Monjezi M, Amiri H, Farrokhi A, Goshtasbi K. Prediction of rock fragmentation
Benadros AG, Kaliampakos DC. Modeling TBM performance with artificial neu- due to blasting in Sarcheshmeh copper mine using artificial neural networks.
ral networks. Tunneling and Underground Space Technology 2004;19(6): Geotechnique and Geology Engineering 2010;28(4):423–30.
597–605. Monjezi M, Dehghani H. Evaluation of effect of blasting pattern parameters on
Cai JG, Zhao J. Use of neural networks in rock tunneling. In: Proceedings of the 9th backbreak using neural networks. International Journal of Rock Mechanics and
international conference on computer methods and advances in geomechanics. Mining Sciences 2008;45(8):1446–53.
A.A. Balkema: Rotterdam; 1997. p. 629–34. Monjezi M, Rezaei M, Yazdian Varjani A. Prediction of rock fragmentation due to
Cheng MY, Ko CH. A genetic-fuzzy-neuro model encodes FNNs using SWRM and blasting in Gol-E-Gohar iron mine using fuzzy logic. International Journal of
BRM. Engineering Application of Artificial Intelligence 2006;19(8):891–903. Rock Mechanics and Mining Sciences 2009;46(8):1273–80.
Christodoulou C, Georgiopoulos M. Applications of neural networks in electromag- Thornton D, Kanchibolta SS, Brunton I. Modeling the impact and blast design vari-
netics. Norwood, MA, USA: Artech House Publishers; 2001. ation on blast fragmentation. International Journal of Fragmentation Blasting
Demuth H, Beale M. Neural network toolbox user’s guide. Natick, MA, USA: The 2002;6(2):169–88.
Math Work, Inc; 1994. Yang Y, Zhang Q. Analysis for the results of point load testing with artificial neu-
Ermini L, Catani F, Casagli N. Artificial neural networks applied to landslide suscep- ral network. In: Proceedings of the 9th international conference on computer
tibility assessment. Geomorphology 2005;66(1–4):327–43. methods and advances in geomechanics. Rotterdam: A.A. Balkema; 1997a. p.
Gates W, Ortiz LT, Florez RM. Analysis of rockfall and blasting backbreak problems. 607–12.
In: Proceedings of the 40th U.S. symposium on rock mechanics. Alexandria, VA: Yang Y, Zhang Q. A hierarchical analysis for rock engineering using artificial neural
American Rock Mechanics Association; 2005. p. 671–80. networks. Rock Mechanic and Rock Engineering 1997b;30(4):207–22.
Haykin S. Neural networks: a comprehensive foundation. Upper Saddle River, NJ: Yilmaz I, Yuksek G. Prediction of the strength and elasticity modulus of gypsum
Prentice-Hall; 1999. using multiple regression, ANN, and ANFIS models. International Journal of Rock
Jimeno CL, Jimeno EL, Carcedo FJA. Drilling and blasting of rocks. Rotterdam: A.A. Mechanics and Mining Sciences 2009;46(4):803–10.
Balkema; 1995. Zhu Z, Mohanty B, Xie H. Numerical investigation of blasting-induced crack ini-
Khandelwal M, Singh TN. Prediction of blast induced air overpressure in opencast tiation and propagation in rocks. International Journal of Rock Mechanics and
mine. Noise Vibration Worldwide 2005;36(2):7–16. Mining Sciences 2007;44(3):412–24.
Khandelwal M, Singh TN. Prediction of blast induced ground vibrations and fre- Zhu Z, Xie H, Mohanty B. Numerical investigation of blasting-induced damage in
quency in opencast mine: a neural network approach. Journal of Sound and cylindrical rocks. International Journal of Rock Mechanics and Mining Sciences
Vibration 2006;289(4):711–25. 2008;45(2):111–21.

View publication stats