Scholarly article on topic 'Prediction of blast-induced flyrock in Indian limestone mines using neural networks'

Prediction of blast-induced flyrock in Indian limestone mines using neural networks Academic research paper on "Earth and related environmental sciences"

CC BY-NC-ND
0
0
Share paper
Keywords
{"Artificial neural network (ANN)" / Blasting / "Opencast mining" / Burden / Stemming / "Specific charge" / Flyrock}

Abstract of research paper on Earth and related environmental sciences, author of scientific article — R. Trivedi, T.N. Singh, A.K. Raina

Abstract Frequency and scale of the blasting events are increasing to boost limestone production. Mines are approaching close to inhabited areas due to growing population and limited availability of land resources which has challenged the management to go for safe blasts with special reference to opencast mining. The study aims to predict the distance covered by the flyrock induced by blasting using artificial neural network (ANN) and multi-variate regression analysis (MVRA) for better assessment. Blast design and geotechnical parameters, such as linear charge concentration, burden, stemming length, specific charge, unconfined compressive strength (UCS), and rock quality designation (RQD), have been selected as input parameters and flyrock distance used as output parameter. ANN has been trained using 95 datasets of experimental blasts conducted in 4 opencast limestone mines in India. Thirty datasets have been used for testing and validation of trained neural network. Flyrock distances have been predicted by ANN, MVRA, as well as further calculated using motion analysis of flyrock projectiles and compared with the observed data. Back propagation neural network (BPNN) has been proven to be a superior predictive tool when compared with MVRA.

Academic research paper on topic "Prediction of blast-induced flyrock in Indian limestone mines using neural networks"

Contents lists available at ScienceDirect

Journal of Rock Mechanics and Geotechnical Engineering

Full length article

Prediction of blast-induced flyrock in Indian limestone mines using neural networks

CrossMark

R. Trivedia, T.N. Singh b*, A.K. Raina

a Central Institute of Mining and Fuel Research, Council of Scientific and Industrial Research (CSIR), Dhanbad, India b Department of Earth Sciences, Indian Institute of Technology Bombay, Powai, Mumbai 400076, India

c Central Institute of Mining and Fuel Research, Council of Scientific and Industrial Research (CSIR), Regional Centre, Nagpur, India

ARTICLE INFO

Article history: Received 16 May 2014 Received in revised form 30 June 2014 Accepted 1 July 2014 Available online 8 August 2014

Keywords:

Artificial neural network (ANN) Blasting

Opencast mining Burden Stemming Specific charge Flyrock

ABSTRACT

Frequency and scale of the blasting events are increasing to boost limestone production. Mines are approaching close to inhabited areas due to growing population and limited availability of land resources which has challenged the management to go for safe blasts with special reference to opencast mining. The study aims to predict the distance covered by the flyrock induced by blasting using artificial neural network (ANN) and multi-variate regression analysis (MVRA) for better assessment. Blast design and geotechnical parameters, such as linear charge concentration, burden, stemming length, specific charge, unconfined compressive strength (UCS), and rock quality designation (RQD), have been selected as input parameters and flyrock distance used as output parameter. ANN has been trained using 95 datasets of experimental blasts conducted in 4 opencast limestone mines in India. Thirty datasets have been used for testing and validation of trained neural network. Flyrock distances have been predicted by ANN, MVRA, as well as further calculated using motion analysis of flyrock projectiles and compared with the observed data. Back propagation neural network (BPNN) has been proven to be a superior predictive tool when compared with MVRA.

© 2014 Institute of Rock and Soil Mechanics, Chinese Academy of Sciences. Production and hosting by

Elsevier B.V. All rights reserved.

1. Introduction

Due to the explosive force, rock fragments are propelled and thrust high into the air and beyond the safety limit of blast area, thus termed as "flyrock". This is mainly due to the flaws presented in the blast design and also due to the mis-interpretation of rock mass behavior. The phenomena of flyrock are always uncontrolled and can never be brought down to zero. But beyond safe permissible limits, flyrock can cause some serious damage to the property and can inflict serious to fatal injury to the personnel, thus making it be one of the main causes of accidents and deaths in opencast mines. Number of accidents as well as number of deaths and injuries due to blasting in Indian mines has been shown in Figs. 1 and 2 (DGMS, 2013, 2014). Flyrock is one of the intriguing problems in opencast mining.

By blasting, we obtain a fixed and/or constant size of rock due to fragmentation, which can help in optimizing the further

* Corresponding author. Tel.: +91 22 25767271. E-mail addresses: tnsingh@iitb.ac.in, tnsiitb@gmail.com (T.N. Singh). Peer review under responsibility of Institute of Rock and Soil Mechanics, Chinese Academy of Sciences.

1674-7755 © 2014 Institute of Rock and Soil Mechanics, Chinese Academy of Sciences. Production and hosting by Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.jrmge.2014.07.003

production economics. During explosion, a part of energy is used in breaking and displacing the rock, while the rest of energy (a major portion) is used up in ground vibrations, air blasts, noises, back breaks, flyrocks, dusts, etc., thus posing a thrust to the nearby life and property (Pal Roy, 1995). Flyrock occurrence can be explained with the help of three basic mechanisms, i.e. cratering, rifling and face burst, as described in Fig. 3 (Moore and Richards, 2005).

Inadequate burden, inadequate stemming length, faulty drilling, back breaks, loose rock on top of the bench due to poor previous blast, very high explosive concentration, inappropriate delay timing, and their sequence, and inaccuracy of delays are the prominent blast design parameters responsible for flyrock problems (Workman and Calder, 1994; Siskind and Kopp, 1995; Adhikari, 1999; Rehak et al., 2001). Unfavorable geological conditions, such as open joints, weak seams, and cavities, have been identified as the major causes of flyrock hazards in opencast mines (Persson et al., 1984; Fletcher and D'Andrea, 1986; Bhandari, 1997; Shea and Clark, 1998).

Blast performance is basically governed by geological and geotechnical data, such as rock quality designation (RQD), un-confined compressive strength (UCS), and joint setting. Joints have an important role to play in any blasting operation as they determine both the safety and performance. Joints are the natural planes of weakness that offer practically no resistance to split. Joints are the zones of discontinuity and weakness and thus during blasting they get affected first, rather than the stable homogenous

Fig. 1. Number of fatal and serious accidents due to blasting in Indian mines (DGMS, 2013).

regions. Therefore, they control the rock breakage process by determining which area gets affected first. Rock fragmentation and over break are influenced by the joint sets. The blast design can be improved by review and analysis of past data of the blasts conducted in the mines (Bhandari, 2011; Parihar and Bhandari, 2012).

Flyrock due to blasting in opencast mines is complex in nature as it is a random phenomenon. Raina et al. (2007, 2011) attempted to devise a criterion for prediction of blast-induced flyrock distances and focused on the factors on which the phenomenon of flyrock depends. An effective prediction of a blast in opencast mines can address the flyrock problem (Ladegaard-Pedersen and Holmberg, 1973; Lundborg, 1974; Roth, 1979; Richards and Moore, 2004; Monjezi et al., 2010).

In this paper, an attempt has been made to calculate initial velocity and angle of ejection of flyrock projectile, and maximum horizontal throw of flying fragments. Flyrock distances have been predicted with the help of feed forward back propagation neural network (BPNN), because it is the most versatile and robust technique (Sawmliana et al., 2007; Singh et al., 2008). Predictive problems, such as ground vibrations, air blast, flyrock, fragmentation, and back break, can be solved using back propagation algorithms (Huang and Wfinstedt, 1998; Singh and Singh, 2005; Tawadrous and Katsabanis, 2005; Remennikov and Mendis, 2006; Singh and Verma, 2010; Khandelwal and Monjezi, 2013; Trivedi et al., 2014).

Experimental blasts have been conducted in 4 limestone mines in India. Out of datasets of 125 experimental blasts, 95 datasets have been used to train the neural network; remaining datasets are used for testing and validation of neural network. Flyrock distances for 30 experimental blasts have been predicted by artificial neural network (ANN), multi-variate regression analysis (MVRA), calculated using motion analysis of videos of flyrock projectiles and compared with the observed data.

Fig. 2. Number of deaths and injuries due to blasting in Indian mines (DGMS, 2014).

Fig. 3. Mechanisms of blast-induced flyrock in opencast mines.

2. Materials and methods

2.1. Field study

Four limestone mines have been selected in this study, i.e. (1) Bamangaon and Mehgaon mines, ACC, Katni, (2) Sheopura-Kesarpura (S-K) mine, shree cement, Beawar, (3) Aditya mine, UltraTech, and (4) Chittorgarh and Kotputli mines, UltraTech, Jaipur. The mining lease of Bamangaon and Mehgaon mines is located at Kymore Village, Vijayraghavgarh Block, Katni District, Madhya Pradesh State of India. It lies between longitude E80°29' and E80°57'E and latitudes N23°48' and N24°8', survey of India top-osheet No. 64A/53. The lease area is 51.2022 km2. The true dip of limestone bed varies between 10° and 20° from west to northwest. However, at some places it is gentle in nature.

The S-K mine is located near Beawar City, Ajmer District, Rajasthan. The mining lease of 856.8 ha lies between longitude E74°22' and E74°26' and latitudes N26°1' and N26°5', on toposheet No. 45J/8 of survey of India. The planned capacity of the mine is 2.0 million tons of limestone peryear. The major portion of area is rocky and there is no vegetation. There occur two parallel ridges of limestone extending along length of area. Between two ridges, there is alluvium/soil. Structurally the area represents a tight isoclinal synclinorium fold where limestone constitutes two limbs of fold, which are separated by a shallow valley. The general strike direction of limestone beds in the area is N30° E. Generally beds are dipping in WNW direction with dip of 45° to as high as 60°.

The Aditya limestone mine is located in Tehsil—Chittorgarh and Nimbahera, Chittorgarh District (Raj). The site is located 18 km Southwest of Chittorgarh Town. The mining lease area of the Aditya limestone mines forms a part of survey of India toposheet Nos. 45L/ 9 and 45L/10 between latitudes 24°43' and 24°45' north and longitudes 74°35' and 74°37' east. The leasehold area of mine is 760.692 ha with the planned production capacity of 6.6 million tons of limestone annually with a stripping ratio of 1:0.33. Structurally the area represents a syncline fold. In spite of above folds, study of dip and strike readings indicates N—S trend with maximum of 10° deviation on either side. The dip varies between narrow ranges of 0° and 20°. Dip direction changes from east to west due to folding.

The Kotputli limestone mine (Grasim Cement) is located at Mohapura Jodhapura near Kotputli Town, Jaipur District, Rajasthan. It is situated at a distance of 165 km south from Delhi and 106 km north from Jaipur. The lease area of the mine is 5.4878 km2. The mine lies between latitude N27°39' and N27°42' and longitude E76°6' and E76°9', survey of India toposheet No. 54A/2. The production capacity of the captive limestone mine is 6 million tons per year. The limestone formations of this area is light to dark gray in color, low to medium grained crystalline, hard and massive in nature. Color banding was observed at some places. The general strike direction of the rock formation of this area is NE—SW with variable dip ranging from 38° to 80° due east.

Fig. 4. Motion analysis of flyrock projectile using 'ProAnalyst' in a blast.

(a) Face before blasting. (b) Face after blasting.

Fig. 5. Photographs of bench face before and after blasting operation.

Limestone mining is being implemented by fully mechanized opencast mining methods in all the mines under the study. The working pit has been excavated by developing working benches of 9—10 m in height. Each bench is connected with a ramp. Necessary haul roads have been developed all around the benches for approaching the working faces. Shovels carry out the excavation and loading operations. Transportation of limestone and waste rock is carried out by dumpers. Before crushing, the limestone from crusher hopper is passed through grizzly screen for screening out intrusive clay. The clay-free limestone is crushed to required size and transported to stock pile located inside plant through belt conveyor. Overburden, mainly clay and soil, is very thin in nature, and it is scrapped by dozer and then lifted with the help of shovel and dumper combination and disposed off to dump yards.

2.2. Methodology used in data generation

The problem of blast-induced flyrock is prominent as these mines are surrounded by the villages, and management of blasting events in these mines has been a challenging task.

Blast design data, such as burden, spacing, stemming, average depth of blast holes, blast hole diameter, charge per hole, linear charge concentration, specific charge (i.e. ratio of explosive consumed to tonnage of rock broken per blast, kg/t), maximum

throw of flyrock or flyrock distance from the blasting face, are generated before blasting operation. Geotechnical data, such as volumetric joint count (Jv), joint spacing, dip and strike of major joint set, and joint condition, have been generated at blasting face and at exposed rock in vicinity of blasting face before blasting

Table 1

Symbols used for blast design and geotechnical data.

Parameters used Unit Symbol

Charge per hole kg Q

Linear charge concentration kg/m ql

Depth of holes m lb

Burden m B

Spacing m Sb

Stemming m ls

Specific charge kg/ton q

Blast hole diameter mm d

Unconfined compressive strength MPa Sc

Rock quality designation % RQD

Launching velocity of flyrock projectile m/s V0

Launching angle of flyrock projectile Degree «0

Gravitational acceleration m/s2 g

Maximum distance or throw of flyrock, observed m Rf

Maximum distance of flyrock projectile, calculated m Rfc

Flyrock distance predicted by ANN m Rfa

Flyrock distance predicted by MVRA m Rfm

Table 2

Blast design and geotechnical parameters at various mines.

Blast No. Name of the mine d (mm) Q (kg) ql (kg/m) lb (m) B (m) Sb(m) ls(m) q (kg/t) sc (MPa) RQD (%) v0 (m/s) »<>(°) Rf(m

bl Mehgaon 115 47.26 8.7 8.7 3 5 3.1 0.14 61 64 26 64 45

b2 Mehgaon 115 50.04 8.7 8.9 3 5.2 3 0.14 62 62 26 66 45

b3 Mehgaon 115 44.88 8.7 8.5 3.1 5.2 3.2 0.13 62 63 25 66 43

b4 Mehgaon 115 47.25 8.7 8.8 3.2 5.5 3.3 0.12 64 65 25 68 38

b5 Bamangaon 115 47.26 8.5 9.1 3.5 5.5 3.6 0.11 67 70 23 68 30

b6 Bamangaon 115 50 8.8 8.8 3.1 5.5 3.1 0.13 62 62 26 67 44

b7 S-K 165 65 16.4 8.3 4.3 6 3.4 0.12 66 62 23 65 37

b8 S-K 165 67.5 16.4 8.5 4.3 6 3.4 0.12 65 62 24 66 38

b9 S-K 165 85 16.7 9 4 5.8 3 0.16 60 61 28 67 51

b10 S-K 165 31.2 16.3 5.5 4.5 5.7 3.7 0.09 67 70 27 75 28

b11 S-K 165 86.11 16.4 9.6 4.4 6.5 3.5 0.12 66 65 25 69 37

b12 S-K 165 65.38 16.6 8 4.1 6.1 3.2 0.13 62 60 27 68 46

b13 S-K 165 92.07 16.7 9.5 4 6 3 0.16 60 60 28 67 49

b14 S-K 165 85.14 16.5 9.5 4.2 6.4 3.2 0.13 65 65 27 70 41

b15 S-K 165 88.76 16.6 9.5 4.1 6.3 3.3 0.14 63 62 28 70 43

b16 S-K 165 89.89 16.6 9.4 4.2 6.3 3.2 0.14 64 62 27 68 45

b17 S-K 165 69.29 16.5 8.5 4.2 6.3 3.4 0.12 64 64 27 68 40

b18 S-K 165 81.47 16.6 9 4.2 6.5 3.2 0.13 65 61 26 66 45

b19 Aditya 115 34.09 8.7 6.9 3.9 5.8 3.1 0.1 56 59 25 71 35

b20 Aditya 115 40.7 8.6 8 4.1 6.2 3.4 0.08 59 63 25 73 30

b21 Aditya 115 42.5 8.6 9 4.2 6 3.5 0.08 60 64 26 74 29

b22 Aditya 115 56.14 8.5 10.3 4.5 6.5 3.7 0.07 63 67 21 71 24

b23 Aditya 115 32.75 8.5 7 4.5 6.3 3.5 0.07 63 67 24 77 22

b24 Aditya 115 51.81 8.6 9.3 4.1 6.3 3.2 0.09 57 63 24 70 31

b25 Aditya 115 53 8.6 9.7 4.1 6.4 3.2 0.09 59 62 24 70 31

b26 Aditya 115 31 8.5 6.9 4.2 5.8 3.3 0.08 60 65 27 77 27

b27 Kotputli 115 51.04 9 9.5 3 4.2 3.1 0.17 62 62 26 64 50

b28 Kotputli 115 51.36 9 9.5 3 4.2 3.1 0.17 62 63 27 65 49

b29 Kotputli 115 51.61 8.7 10 3.3 4.5 3.6 0.14 65 69 23 66 33

b30 Kotputli 115 51.75 8.7 10 3.5 4.4 3.4 0.13 67 73 25 73 28

operation takes place. The RQD has been estimated using volumetric joint count method (Palmstrom, 1982). Rock samples from blasting face have been collected from the mines and tested in the laboratory in Central Institute of Mining and Fuel Research, Council of Scientific and Industrial Research (CSIR), Dhanbad, India. The UCS of 1 inch cube samples has been tested and density and UCS of rock samples have been known. To identify the maximum throw of blast-induced flyrock, the distance of fragment of 10 cm or more, in size, from the face blasted has been considered. The portable GPS (global positioning system) was used to measure the distance of flyrock from the face blasted.

The experimental blasts were monitored with high-resolution video camera of 24 frames per second. The interval between two consecutive frames is 42 ms. The view of the blasting events was shot from safe distance in view safety of the equipments and personnel. The flying fragment motions were captured in videos of the blasting events. The videos of the blasts have been analyzed to know launching velocity and launching angle of the flyrock projectiles. The AVI files obtained through unloading the images from the camera to the computer for slow motion analysis using 'ProAnalyst 1.5.6.7' software of XCITEX, USA (Figs. 4 and 5). A pre-calibration technique is used in the calculation. The calibration in the present study was made using three red flags separated horizontally and vertically to known distance. For verification, the other known parameters like bench height, cut length, etc. were used.

Launching angle and launching velocity of flyrock projectile for each blast of all the limestone mines under the study with appropriate symbols are shown in Tables 1 and 2. The flyrock distances have been calculated using general trajectory formula (Richards and Moore, 2004):

vg sin(2#0)

2.3. ANN approach to predict blast-induced flyrock

The pattern of the result is predicted by ANN on the basis of preceding learning. Once the neural network has been established and trained, any similarities in the new pattern will be detected and the result pattern will change accordingly, thus providing the technique interpolation capability. ANN is trained using back propagation algorithm. The feed forward BPNN comprises 3 layers, i.e. input layer, hidden layer and output layer. Layers are made up of neurons which are the basic processing units. These neurons connect the layers using appropriate weight. The output of the neurons in the input layer serves as input for the neurons in the hidden layer and the same applies to the connection between hidden and output layers. The problem defines the number of hidden layers

Table 3

List of input and output parameters used in ANN.

Input parameters Output parameter

ql (kg/m) B (m) ls (m) q (kg/t) sc (MPa) RQD (%) Rf(m)

8.5—16.7 3—4.6 3—3.6 (3.3) 0.07—0.18 (0.12) 58—68 (62) 55—79 (65) 20—56 (34)

(8.7 for d = 115 mm, 16.5 for d = 165 mm) (3.3 for d = 115 mm, 4.2 for d = 165 mm)

Note: Values in the brackets are mean values of input and output parameters.

and the neurons in them. In the present case, the BPNN with 'Levenberg-Marquardt' algorithm and 'log-sigmoid' transfer function has been undertaken. After trial of a number of different combinations, two hidden layer and ten neurons in each hidden layer have been found as the best predictive model for the case under the study. The input and output parameters used in ANN model are shown in Table 3. Ninety five datasets have been used to train the neural network. Thirty datasets have been used for testing and validations of neural network. Architecture of neural network and its performance during training are shown in Figs. 6 and 7. The regression plots of ANN during training, testing and validation indicate the excellence of the selected network as shown in Fig. 8.

2.4. Prediction of flyrock by MVRA

The relation among the variables can be calculated using MVRA and the least squares fit method to solve the dataset. Regression matrix solves the simultaneous equations thus created. MVRA has been conducted by the same datasets and the same input parameters used in ANN. The equation for prediction of flyrock by MVRA is as follows:

105'1q°'51q0'14 BaM/aMspsRQpaiö

Fig. 6. Training of back propagation neural network.

3. Results and discussion

3.1. Performance of neural network model

Flyrock distance has been predicted by neural network and multiple regressions. Relationships between flyrock distance predicted by ANN and MVRA have been compared with the observed

Best Validation Performance is 7.7938 at epoch 9

Validation

0 5 10 15

15 Epochs

Fig. 7. Performance of neural network during training.

Fig. 8. Regression plots of ANN during training, testing and validation.

ones as shown in Figs. 9 and 10, respectively. Also, a line of 1:1 slope intersecting origin has been drawn in Figs. 9 and 10. It is evident that the predicted values are closer to observed values in ANN when compared with MVRA. Mean absolute error (MAE) is 2.5 m in case of MVRA, whereas it is 0.92 m in case of ANN. Similarly, root mean square error (RMSE) is 3.1 m in case of MVRA and 0.99 m in case of ANN. Coefficient of determination (R2) is 0.815 in case of MVRA whereas it is 0.983 in case of ANN. Maximum flyrock distances have been calculated by projectile motion (Eq. (1)) as explained earlier and compared with the observed values as well as values predicted by ANN and MVRA, as

shown in Fig 11. MAE and RMSE in terms of flyrock distances calculated by projectile motion theory are as high as 5.2 m and 6.1 m, respectively. Higher MAE may be due to the fact that projectile motion theory does not take into account the air resistance offered to flying fragments during its travel. These indices of model performance indicate an excellent correlation between flyrock distances predicted by ANN and observed values. Thus, the proposed ANN model has exceptionally high capability in predicting complicated blasting problems.

Fig. 9. Relation between flyrock distances predicted by ANN and observed values.

Fig. 10. Relation between flyrock distance predicted by MVRA and observed values (Solid line indicates best fit line and dotted line is 1:1 slope).

Fig. 11. Comparison between predicted, calculated and observed flyrock distances.

3.2. Sensitivity analysis

In order to assess the impact of an independent variable on maximum throw flyrock, sensitivity analysis has been performed in 'Matlab 7.11.0 (R2010b)' using 'simulation tool'. While performing sensitivity analysis of an independent variable, all other independent variables are set at their mean values as specified in Table 3. Sensitivity analyses for all the six independent variables have been performed, as shown in Fig. 12 and the results have been described in Table 4. The linear charge concentration and specific charge have a positive correlation with the flyrock distance, whereas burden, stemming, UCS and RQD have a negative correlation with the flyrock distance. Impact of rate of change in burden and stemming are more pronounced at blast hole of d = 165 mm, whereas impact of rate of change in UCS and RQD are more pronounced in case of blast hole of d = 115 mm, and impact of rate of change in linear charge concentration and specific charge are more or less same in both cases.

Fig. 12. Effects of the six independent variables on maximum throw of flyrock.

Table 4

Results of sensitivity analysis.

Independent Rate of change variable

For blast hole d = 115 mm For blast hole d = 165 mm

Linear charge ARf/Aq = 0.41 m per 0.1 kg/m ARf/Aq1 = 0.4 m per 0.1 kg/m concentration

Burden ARf/AB = (-)0.55 m per 0.1 m ARf/AB = (-)1.8 m per 0.1 m

Stemming ARf/Als = (-)0.2 m per 0.1 m ARf/Als = (-)0.8 m per 0.1 m

Specific charge Rf/Aq = 0.75 m per 0.1 kg/ton Rf/Aq = 0.7 m per 0.1 kg/ton

UCS ARf/Asc = (-)0.3 m per MPa ARf/Asc = (-)0.1 m per MPa

RQD ARf/ARQD = (-)0.65 m per 1% ARf/ARQD = (-)0.5mper1%

4. Conclusions

The ANN with two hidden layers and log-sigmoid transfer function has been found as the best predictive combination in back propagation algorithm. While solving the blast-induced flyrock predictive problem, six independent variables have been selected, i.e. linear charge concentration, burden, stemming length, specific charge, UCS, and RQD. Among the various flyrock models namely ANN, multiple regression and 'projectile motion model', ANN has been proved to be an excellent model due to its highest coefficient of determination and lowest MAE as well as RMSE. Sensitivity analysis has been performed and we found a positive correlation of linear charge concentration and specific charge with the flyrock distance, whereas burden, stemming length, UCS and RQD bear a negative correlation with the flyrock distance. Impact of burden and stemming length are more pronounced at blast hole of d = 165 mm whereas impact of UCS and RQD are more pronounced in case of blast hole of d = 115 mm.

Conflict of interest

The authors wish to confirm that there are no known conflicts of interest associated with this publication and there has been no significant financial support for this work that could have influenced its outcome.

References

Pal Roy P. Breakage assessment through cluster analysis of joint set orientations of exposed benches of opencast mines. Geotechnical and Geological Engineering 1995;13(2):79—92.

Palmstrom A. The volumetric joint count — a useful and simple measure of the degree of jointing. In: Proceedings of the 4th international congress of international association of engineering geology; 1982. p. 221—8. New Delhi.

Parihar CP, Bhandari S. Improving blasting operations using data management and analysis. In: Proceedings of internationalconference on rock fragmentation by blasting, Fragblast 10; 2012. p. 405—9. Delhi, India.

Persson P, Holmberg R, Lee J. Rock blasting and explosives engineering. New York: CRC Press; 1984.

Raina AK, Chakraborty AK, More R, Choudhury PB. Design of factor of safety based criterion for control of flyrock/throw and optimum fragmentation. Mining Engineering 2007;87:13—7.

Raina AK, Chakraborty AK, Choudhury PB, Sinha A. Flyrock danger zone demarcation in opencast mines: a risk based approach. Bulletin of Engineering Geology and the Environment 2011;70(1):163—72.

Rehak TR, Bajpayee TS, Mowrey GL, Ingram DK. Flyrock issues in blasting. In: Proceedings of the 27th annual conference on explosives and blasting technique, vol. I. Cleveland: International Society of Explosives Engineers; 2001. p. 165—75.

Remennikov AM, Mendis PA. Prediction of airblast loads in complex environments using artificial neural networks. WIT Transaction on the Built Environment 2006;87:269—78.

Richards AB, Moore AJ. Flyrock control by chance or design. In: Proceedings of 30th annual conference on explosives and blasting technique; 2004. New Orleans.

Roth JA. A model for the determination of flyrock range as a function of shotcon-dition. US Department of Commerce; 1979. p. 61. NTIS Report No. PB81222358.

Sawmliana C, Pal Roy P, Singh R, Singh TN. Blast induced air overpressure and its prediction using artificial neural network. Mining Technology 2007;116(2): 41—8.

Shea CW, Clark D. Avoiding tragedy: lessons to be learned from a flyrock fatality. Coal Age 1998;103(2):51—4.

Singh TN, Singh V. An intelligent approach to predict and control ground vibration in mines. Geotechnical and Geological Engineering 2005;23(3):249—62.

Singh TN, Dontha LK, Bharadwaj V. A study into blast vibration and frequency using ANFIS and MVRA. Mining Technology 2008;117(3):116—21.

Singh TN, Verma AK. Sensitivity of total charge and maximum charge per delay on ground vibration. Geomatics, Natural Hazards and Risk 2010;1(3):259—72.

Siskind DE, Kopp JW. Blasting accidents in mines: a 16 year summary. In: Proceedings of the 21st annual conference on explosives and blasting technique. Cleveland: International Society of Explosive Engineering; 1995. p. 224—39.

Tawadrous AS, Katsabanis PD. Prediction of surface blast patterns in limestone quarries using artificial neural networks. International Journal for Blasting and Fragmentation 2005;9(4):233—42.

Trivedi R, Singh TN, Mudgal K. Impact of geotechnical parameters on blast induced flyrock using artificial neural network — a case study. In: Proceedings of the 2nd international conference on advanced technology in exploration and exploitation of minerals; 2014. p. 128—34. Jodhpur, India.

Workman JL, Calder PN. Flyrock prediction and control in surface mine blasting. In: Proceedings of the 20th annual conference on explosives and blasting technique; 1994. p. 59—74. http://dx.doi.org/10.1007/s12517-010-0133. Austin.

Adhikari GR. Studies on flyrock at limestone quarries. Rock Mechanics and Rock Engineering 1999;32(4):291-301.

Bhandari S. Engineering rock blasting operations. Rotterdam: A.A. Balkema; 1997. p. 315-22.

Bhandari S. Information management for improved blasting operations and environmental control. In: Proceedings of the 3rd Asia-Pacific symposium on blasting techniques. China: Xiamen; 2011.

Directorate general of mines safety (DGMS). Monthly review of accidents in India (up to December 2013); 2013.

Directorate general of mines safety (DGMS). Statistics of mines in India (metal & oil); 2014.

Fletcher LR, D'Andrea DV. Control of flyrock in blasting. In: Proceedings of the 12th conference on explosives and blasting technique; 1986. p. 167-77. Atlanta, Georgia.

Huang Y, Wfinstedt S. The introduction of neural network system and its applications in rock engineering. Engineering Geology 1998;49(3/4):253-60.

Khandelwal M, Monjezi M. Prediction of flyrock in open pit blasting operation using machine, learning method. International Journal of Mining Science and Technology 2013;23(3):313-6.

Ladegaard-Pedersen A, Holmberg R. The dependence of charge geometry on flyrock caused by crater effects in bench blasting. Swedish Detonic Research Foundation Report DS 1973:1-38.

Lundborg N. The hazards of fly rock in rock blasting. Swedish Detonic Research Foundation Report DS 1974:12.

Monjezi M, Bahrami A, Yazdian V. Simultaneous prediction of fragmentation and flyrock in blasting operation using artificial neural networks. International Journal of Rock Mechanics and Mining Sciences 2010;47:476-80.

Moore AJ, Richards AB. Golden pike cut-back flyrock control and calibration of predictive model. Prepared for Kalgoorlie consolidated gold mines; 2005.

Dr. T.N. Singh is working as professor in the Department of Earth Sciences, Indian Institute of Technology — Bombay, Mumbai, India. His research interests are mainly rock sciences and rock engineering, slope stability, rock blasting, natural disaster prediction and management, soft computing and numerical application in Applied Geo-sciences, etc. Till now he has published more than 250 research papers in different referred journals of repute. He is editor of a dozen of National and International Journals related to his subject.

Dr. Ratnesh Trivedi completed his Bachelors in Mining Engineering from Mugneeram Bangur Memorial Engineering College, MBM Jodhpur, Rajasthan and then went on to finish his Masters of Technology in Mining Engineering from Indian Institute of Technology, Banaras Hindu University, Varanasi. Currently he is working with Central Institute of Mining and Fuel Research, Council of Scientific and Industrial Research (CSIR), Dhanbad, India as a scientist and actively involved in research related to rock blasting, mine planning, slope stability and geo-environmental technology.