Use of plastic bottles in blasting

Use of plastic bottles in blasting

• 1.
  Productivity Improvement inAggregate Quarry vis-a-vis Fragmentation by Blasting Ramesh MurldiharBhatawdekar Centre of Tropical Geoengineering (GEOTROPIK), UniversitiTeknologi Malaysia, 81310, Johor, Malaysia EdyTonnizam Mohamad Centre of Tropical Geoengineering (GEOTROPIK), UniversitiTeknologi Malaysia, 81310, Johor, Malaysia DanialJahedArmaghani Department of Civil and Environmental Engineering, Amirkabir University of Technology, 15914, Tehran, Iran. SaksaridChangthan Siam City Concrete Co, Bangkok 10110, Thailand Gyandera Kumar Pradhan AKS University, Madhya Pradesh, India ABSTRACT: Blasting for primary fragmentation, needs of an aggregate quarry has a unique role as compared with normal blasting in mining. The percentage recovery of aggregates, dictate the overall mine economics. Blasting design program has been a complex in mine feasibility stage as well as throughout the life of the mine. Selection of drilling, explosives, initiation system over the years has attracted a lot of scientific inputs. While geology and GSI have an overriding effect, inputs like mean block size, RQD%, powder factor, MC, ratios of hole diameter, spacing, stemming with burden, hold key to the success for determining mean fragment size. R2 value with ANN is 0.845 as compared to 0.674 with MVRA.Excavator output improved from 200 TPH to 245 TPH with reduction in mean fragment size from 0.27 m to 0.13 m. Productivity improvement without compromising on HSE norms and regulatory framework in Thailand Quarry are presented. Keywords: Artificial neural network (ANN), Geotechnical strength Index (GSI), RQD%, Powder factor, Maximum charge per delay (MC) 1 INTRODUCTION In Thailand, world class infrastructure is developed in and around Bangkok. Number of construction aggregate quarries and limestone quarries are developed as shown in Figure 1. Construction aggregates consists of Permian Ordovician limestone, Cretaceous granite and Tertiary basalt. Typical large quarries utilize 200 mm diameter drill hole which carries powder factor from 0.4 to 0.66 kg/ cu m (Tangchawal, 2017). Maximum charge per delay varies from 64.31 to 160.85 kg per delay for a distance of 150 m and 534.91 to 643.40 kg per delay for a distance of 300m. For smaller quarries burden to spacing pattern of 1.5m X 1.8 m to 3.5 m X 3.5 m patterns are used. For larger quarries burden to spacing pattern from 5 m x 6 m to 6 m X 8 m are used. The aggregate quarry under this study has annual production capacity of 2.5 MTPA and has expanded its capacity to 5 MTPA. The existing quarry use 76 mm diameter DTH drills, 2.2 Cum back hoe excavators and 25 T dump trucks. Feed size to primary crusher is 800 mm. The objective of this study is listed as follows: a) to classify the rock mass based on Geological Strength Index (GSI) b) to investigate key parameters influencing blast fragmentation through multi variate regression analysis (MVRA) and artificial neural network (ANN) c) to evaluate the efficiency of excavator with fragmentation. Analysis by previous researchers (Ash, 1968; Hustrulid, 1999; Jimeno et al., 1995) shows that fragmentation depends upon rock mass properties and geological discontinuities, properties of explosives, blast hole diameter, burden, spacing, bench height, stemming length, blast hole alignment and deviation, type of blast hole pattern rectangular or staggered, blast hole sub drilling, Prediction of blast fragmentation depends upon three main factors – rock mass properties, instant release of explosive energy and blast design.
• 2.
  Figure 1: Smalland large aggregate quarries around Bangkok, (Tangchawal, 2017) In situ block size and distribution with discontinuities play an important role in blastability of rock mass (Wang et al.,1991. 1992; Lu and Latham 1996; Lu 1997). As per blast fragmentation study by Kulatilake et al. (2010, 2012) and Mehrdanesh et al. (2017), mean in situ block size plays crucial role in predicting mean blast fragment size. RQD is a measure of unweathered drill core longer than 10 cms (Deere, 1966). RQD being easy, quick and hence used to indicate joint density for core hole logs. For blasting face, to establish mean fragment size, RQD of blasting face is considered one of important parameter (Chakraborty et al., 2004; Saliu and Akande (2007); Mehrdanesh et al., 2017). A condition of blasted block size distribution from a condition of in situ block size distribution is through blasting process (Latham and Lu, 1999). Explosive energy which results in breakage of rock during blasting can be considered in two ways. Over all rock mass to explosives consumed which is represented by powder factor. Instantly gas pressure is created to break rock after hole is fired. Maximum charge per delay in each blast can be correlated with maximum release of energy instantaneously. Various studies carried out to predict mean blast fragment size, powder factor is considered as crucial parameter (Chakraborty et al., 2004; Morin and Ficarazzo, 2006; Saliu and Akande, 2007; Gheibie et al., 2009; Kulatilake et al., 2010, 2012; Singh et al., 2016; Sharma and Rai 2017; Mehrdanesh et al., 2017). Maximum charge per delay to predict blast fragmentation is demonstrated by Monjezi et al. (2009) and Faramarzi et al. (2013). Studies by NIRM and CIMFR in India recommended that minimum hole diameter should be one hundredth of the bench height.On the other hand, Adhikari (1999) proposed the maximum hole diameter to be 0.01666 of the bench height. Bhandari (1997) proposed burden to be 15 to 40 times of hole diameter for effective blasting in opencast mines. For good blast fragmentation, stiffness ratio should be between 2 to 4 as per Konya and Walter (1990). Research studies at limestone quarries show that with increase in stemming length higher boulders are generated mainly from stemming portion (Venkatesh et al., 1999 and Cevizci and Ozkahraman, 2012). Thus, from various studies it is observed that instead of single parameter of hole diameter, burden, spacing and bench height, various ratios of these parameters contribute to blast performance. To predict mean blast fragmentation size, various ratios of burden to blast hole diameter, bench height to burden, spacing to burden and stemming length to burden play an important role (Chakraborty et al., 2004; Kulatilake et al., 2010, 2012; Faramarzi et al., 2013; Singh et al., 2016; Sharma and Rai 2017; Mehrdanesh et al., 2017). Figure 2: Limestone quarry at Thailand
• 3.
  The selected limestonequarry at Thailand is well developed with number of benches as shown in Figure 2. There is a variation in block size, orientation of joints and spacing of joints. 2 METHODS Knowledge of rock mass classification based on GSI is useful for blast design. Every blasting face was classified based on four types of rocks identified based on GSI for jointed rock mass (Marinos and Hoek, 2000). For this study, rock mass is classified as blocy, very blocky, blocky/seamy and disintegrated. GSI depends upon fundamental geological processes consisting of blockiness of rockmass and condition of joints (Marinos et al., 2007). From visual examination of each blasting face, GSI is assessed based on lithology, structure, joint condition. Average in-situ block size and RQD% were measured at each blasting face and GSI recorded for each blast. Blast design parameters consisting of hole diameter, burden, spacing, bench height, stemming height were recorded from individual blast design to determine various ratios. Maximum charge per delay, powder factor were also recorded. For individual blast, photographs were taken by placing ball at blasted muck pile and further image analysis by software. Following input parameters are considered to predict mean fragment size: Mean block size (XB), RQD%, powder factor, maximum charge per delay, (BD) burden to hole diameter ratio, (SB) spacing to burden ratio, stiffness ratio (HB) consisting of ratio of bench height to burden, (TB) stemming height to burden ratio. Total 78 blasting data set were collected for analysis. Methodology of multivariate regression analysis (MRA) and Artificial Regression Analysis (ANN) was adopted to predict mean blast fragment size (XB). Data on excavator output in Tonnes per hour (TPH) for each blast was collected. Excavator output was compared with fragment size greater than 800 mm (Boulder) and mean fragment size (X50). Drilling output was recorded based drilling output in meters per meter for each size of drill. Table 1 : Summary of data collection: Parameter Source of data/ methodology GSI Visual examination of face RQD% Actual field measurement Insitu block size Burden, spacing Blast design of individual blast Hole diameter Bench height Stemming height Powder factor Maximum charge per delay Powder factor Mean fragment size Photograph of blasted face & analysis with software Excavator efficiency Blast wise loading rate by excavator in TPH based on a) Boulders >800mm per 100 T b) mean fragment size X50 Drill efficiency Drill hole diameter, drilling output in meters per hour 2.1 Multivariate Regression Analysis The objective of multivariate regression analysis (MRA) is to resolve equivalent factors of a function to contribute best fit based on a set of data observations. This technique results with a function which is linear (straight-line) equation. In circumstances where more than one independent variable occurs, MRA is utilized to form the best-fit equation. A least square fit is performed to resolve engineering problems through MRA. While applying this technique, some coefficients are proposed by mode of backlash operator. The MRA equation is given as follows (Tonnizam Mohamad et al., 2017): z = x+y1m1+ y2m2+ y3m3+….+ynmn (1) where m1, m2, m3 ….mn are independent variables, y1, y2, y3, …. yn are coefficients of independent variables, x is constant andz is the output of the system. In order to develop a MRA equation for prediction of fragment size, using the mentioned system inputs, Excel software was selected and utilized. All 78 datasets were divided to 2 sections; training (57 datasets) and testing (21 datasets) and then, considering only training datasets, a MRA equation was constructed as follows: BS is burden to spacing ratio. RQD is expressed in %. MC is maximum charge per delay in Kg. PF is powder factor in Kg per cum of blasted material. BD is burden to hole diameter ratio. HB is bench height to burden ratio. TB is stemming height to burden ratio. After developing the above equation, there is a need to evaluate it through 21 testing datasets. Results of training and testing datasets showed that an acceptable performance prediction can be achieved by the developed MRA equation. Several
• 4.
  performance indices consistingof coefficient of determination (R2 ), root mean square error (RMSE) and variance account for (VAF) were used to evaluated predictive models in this study where their equations can be found in other studies (Armaghani et al., 2017). More discussion regarding the developed MRA will be given later. 2.2 Artificial Neural Network By studying the the human-brain information process motivated to develop ANN as soft computation technique. Network architecture, learning rule, and transfer function are three major factors of ANN. Recurrent and feed-forward are two major groups of ANNs. The study by Shahin et al. (2002) proposes that if there is no time-vulnerable factor in the ANN, the feed-forward ANN (FF- ANN) can be carried out. The multi-layer perceptron (MLP) neural network is one of well-known FF- ANNs (Haykin 1999). This type of ANN is linked to each other by means of various weights to a number of neurons in different layers (input, hidden and output layers). Kalinli et al. (2011) expressed the great performance of MLP-ANNs in approximating different functions in high-dimensional spaces. After providing the data into ANNs and before finding the solution, still, the ANN requires training. Dreyfus (2005) reported that the back-propagation (BP) algorithm is the most popular algorithms among various category of learning algorithms for training the MLP feed-forward neural networks. Kuo et al. (2010) interpretedthat in a classical BP-ANN, the imported data in the input layer kicks off to generate to hidden nodes through connection weights. In the BP-ANN, the flexible connection or weight, Wij is multiplied by the input from each neuron in the preceding layer, li.The sum of the weighted input signals is calculated at each node and subsequently, this value is combined to obtain a threshold value recognized as the bias value, Bij, (see Eq. 2). The output of node is produced (see Eq.3) by combining input, Ji, and then passing through a non-linear transfer function f (Jj), for instance a sigmoidal function.The output of each neuron, generally, produces the input to the next layer neuron.This procedure is pursued until the output is obtained.Yet, to pick up the error, subsequently the generated output is analysed with the expected output.The major reason for BP training is to substitute the weights between the neurons in iterative manner that reduces the mean square error (MSE) to least of the system.In classic artificial intelligence book by Fausett (1994), the BP algorithm is illustrated. (2) (3) Khamesi et al. (2015) illustrated that as a part of ANN modelling procedure, the prepared database was normalized during first stage in order to simplify the design procedure and given below: Znorm = (Z – Zmin) / (Zmax- Zmin) (4) where Z and Znorm are the measured and normalized values, respectively. Zmax and Zmin are the maximum and minimum values of the Z. Researchers, in the past, have suggested different percentages for the testing datasets. Amounts of 20%, 25% and a range of 20%-30% of total datasets were recommended for testing dataset in the previous studies (Swingler, 1996). Hence, as mentioned earlier, in the present study, 80% and 20% of total datasets were practiced for model development and confirming the developed models, respectively. In ANN modelling, choosing of the ANN training algorithm and also decision of selecting the network architecture are the most challenging task. Levenberg-Marquardt (LM) was chosen among all ANN training algorithms and utilised to train the ANN systems. On the contrary, as stated by many scholars (e.g. Singh et al. 2006), an ANN network with only one hidden layer can evaluate almost all engineering problems, therefore, an ANN network with one hidden layer was applied. Another important issue in designing an ANN model is number of hidden nodes where the maximum number of hidden nodes were reported by Hecht- Nielsen (1987) as 2Ni + 1. In this study, with Ni = 8, the maximum number of hidden node can be equal or less than 17. Hence, many ANN models were built based on the different number of hidden nodes and finally, according to obtained performance indices, a model with 11 hidden nodes outperforms the other models. The results demonstrated that by developing new ANN model, accuracy of the developed model was increased significantly compared to created model by MRA. Evaluation of the selected ANN model is given later. 3 RESULTS AND DISCUSSION 3.1 GSI Geological study was undertaken, and rock mass was classified based on geological strength index. Rock mass at different locations was classified into four categories (a) Blocky limestone (b) Very blocky limestone (c) Blocky/ seamy limestone and (d) Dis- integrated limestone as shown in Figure 3.
• 5.
  From 78 blastingdata sets, it is observed that GSI based rock mass classification, mean in-situ block size range for blocky limestone is 0.8 to 1.2 m, very blocky limestone is 0.5 to 0.8 m, blocky / seamy limestone is 0.3 to 0.6 m and disintegrated limestone is 0.1 to 0.3 m. Figure 3: Rock mass classification based on GSI (Bhatawdekar et al., 2016) Figure 4: Photographs of blasted limestone at Thailand Quarry for image analysis Figure 4 shows photographs of blasted muck pile. Based on blast fragmentation as per Figure 4, blast fragmentation analysis is classified based on rock mass classification based on blocky, very blocky, blocky/ seamy and disintegrated. Figure 5 shows image analysis done with software and graphs are classified based on limestone as per GSI stating in situ mean block size (X) for a given blast. Four graphs consist image analysis weight % finer Vs mean fragment size. Figures 5(a), (b), (c) and (d) consist of image analysis for blocky limestone, very blocky limestone, blocky/ seamy limestone and disintegrated limestone respectively. From blast fragmentation distribution curves in Figure 5, it is observed that disintegrated limestone has 40% below 0.1 m and thus maximum fine generation. 0 10 20 30 40 50 60 70 80 90 100 0.1 1 10 Weight%Finer m (a) 0 10 20 30 40 50 60 70 80 90 100 0.1 1 10 Weight%Finer m (b)
• 6.
  Figure 5: Blastfragmentation distribution (a) Blocky limestone (b) Very blocky limestone (c) Blocky/ Seamy limestone (d) Dis- integrated limestone 3.2 MVRA and ANN As stated before, R2 , RMSE and VAF were used to evaluate the obtained results. These performance indices were calculated for both MRA and ANN models as shown in Table 5. R2 were achieved as 0.678 and 0.764 for training and testing of MRA, respectively, while these results were obtained as 0.825 and 0.845 for ANN model. Apart from R2 , considering RMSE and VAF results, it was found that ANN is capable to increase accuracy prediction level of fragment size. By focusing on VAF results, it can be observed that by developing an ANN model VAF values were deeply improved from about 65% to more than 80%. This shows ability of ANN model in solving problem of predicting fragment size. Generally, it can be concluded that ANN model compared to MRA is performed better and can predict fragment size with higher accuracy level. The predicted fragment size values by MRA and ANN models together with their measured values are shown in Figures 6 and 7, respectively. These figures proved that ANN predictive model can provide a higher performance capacity in predicting fragment size. Table 2: The obtained results of developed models Model Performance Index R2 RMSE VAF (%) Train Test Train Test Train Test MRA 0.678 0.764 0.026 0.038 66.684 64.768 ANN 0.825 0.845 0.182 0.214 82.001 80.286 Figure 6: The predicted fragment size values by MRA model together with their measured 0 10 20 30 40 50 60 70 80 90 100 0.1 1 10 Weight%Finer m (c) 0 10 20 30 40 50 60 70 80 90 100 0.1 1 10 Weight%Finer m (d)
• 7.
  Figure 7 Thepredicted fragment size values by ANN model together with their measured values 3.3 Excavator Performance Table 3 shows excavator output compared with fragment size greater than 800 mm (boulder) per 100 Tof blasted material. Table 4 shows excavator output based on mean blast fragment size in ‘m’. Table 3: Excavator output in TPH compared with fragment > 0.8 m (Boulder) per 100 T Fragment > 0.8m per 100 T Excavator output TPH 200- 205 206- 210 211- 215 216- 220 221- 225 226- 230 231- 235 236- 240 241- 245 246- 250 4 5 6 7 8 9 10 11 12 Table 4: Excavator output in TPH compared with mean fragment size (X50) in m Mean fragmentation m Excavator output TPH From To 200- 205 206- 210 211- 215 216- 220 221- 225 226- 230 231- 235 236- 240 241- 245 246- 250 0.131 0.140 0.141 0.150 0.151 0.160 0.161 0.170 0.171 0.180 0.181 0.190 0.191 0.200 0.201 0.210 0.211 0.220 0.221 0.230 0.231 0.240 0.241 0.250 0.251 0.260 0.261 0.270 Output of excavator varies from 200 to 250 TPH. For increase in fragment size greater than 0.8 m (boulder) from 4 to 12 per 100 T of blasted limestone, excavator output drops from 245 to 200 TPH. For increase in mean fragment size from 0.13 to 0.27 m, excavator output drops from 245 to 200 TPH. Based on 8 variable inputs, it is possible to predict mean fragment size with ANN model and improve productivity of excavator. 3.4 Drill performance: Comparison is made whether to have additional DTH(Down The Hole) drills or new Top Hammer
• 8.
  (TH) drills. New102 mm diameter TH drills are procured to meet enhanced drilling requirement. New TH drill output is 30 m/hour as compared to old DTH drills output of 12 m/hour.Top hammer drill has improved productivity from 12m/hour for DTH drill to 30 m/ hour for Top hammer drill. Hole diameter increased marginally from 76 mm to 102 mm without having adverse environment impact on fly rock, ground vibration and air over pressure for meeting local regulation. Spacing and burden was proportionately increased. Thus, drilling productivity has improved from 270 T/ hour to 1300 T/hour of drilling. 4 CONCLUSIONS • Rock mass classification is an effective tool for blast design. This has to be done on the basis of study of geology of the area and field observations of the blasting face. of limestone into four types as blocky, very blocky, blocky/ seamy and disintegrated limestone supports for blast design based on observation of blasting face. • Eighnput Input parameters consisting of mean block size (XB), RQD%, powder factor, maximum charge per delay (MC), (BD) burden to hole diameter ratio, (SB) spacing to burden ratio, stiffness ratio (HB) consisting of ratio of bench height to burden, (TB) stemming height to burden ratio play significant role are suitable for predicting mean fragment size are suitable for predicting mean fragment size (X50). • R2 achieved as 0.678 and 0.764 for training and testing of MVRA, respectively. • R2 achieved as 0.825 and 0.845 for training and testing of ANN, respectively. • Excavator output improves from 200 TPH to 245 TPH with reduction in mean fragment size from 0.27 m to 0.13 m and bigger fragment size greater than 0.8 m from 12% to 4% • Introduction of top hammer drill from DTH drill has improved output 270 to 1300 T/ hour Thus objectives of improving overall productivity of aggregate quarry through drilling and blasting were achieved. REFERENCES: Adhikari, G.R., 1999. Selection of blasthole diameter for a given bench height at surface mines. International Journal of Rock Mechanics and Mining Sciences, 36(6), pp.843- 847. Armaghani, D.J., Mohamad, E.T., Narayanasamy, M.S., Narita, N. and Yagiz, S., 2017. Development of hybrid intelligent models for predicting TBM penetration rate in hard rock condition. Tunnelling and Underground Space Technology, 63, pp.29-43. Ash, R.L., 1968. The design of blasting rounds. Surface mining, p.387. Bhandari, S., 1997. Engineering rock blasting operations. A. A. Balkema. 388, p.388. Chakraborty, A.K., Raina, A.K., Ramulu, M., Choudhury, P.B., Haldar, A., Sahu, P. and Bandopadhyay, C., 2004. Parametric study to develop guidelines for blast fragmentation improvement in jointed and massive formations. Engineering Geology, 73(1), pp.105-116. Cevizci, H. and Ozkahraman, H.T., 2012. The effect of blast hole stemming length to rockpile fragmentation at limestone quarries. International Journal of Rock Mechanics and Mining Sciences, 53, pp.32-35. Deere, D. U., A. J. Hendron, F. D. Patton, and E. J. Cording. "Design of surface and near-surface construction in rock." In The 8th US symposium on rock mechanics (USRMS). American Rock Mechanics Association, 1966. Dreyfus, G., 2005. Neural networks: methodology and applications. Springer Science & Business Media. Faramarzi, F., Mansouri, H. and Farsangi, M.E., 2013. A rock engineering systems based model to predict rock fragmentation by blasting. International Journal of Rock Mechanics and Mining Sciences, 60, pp.82-94. Fausett, L. and Fausett, L., 1994. Fundamentals of neural networks: architectures, algorithms, and applications (No. 006.3). Prentice-Hall, Geological report of limestone deposit, Thailand Quarry (2010) Gheibie, S., Aghababaei, H., Hoseinie, S.H. and Pourrahimian, Y., 2009. Modified Kuz—Ram fragmentation model and its use at the Sungun Copper Mine. International Journal of Rock Mechanics and Mining Sciences, 46(6), pp.967-973. Haykin, S., 1999. Adaptive filters. Signal Processing Magazine, 6, p.1. Hecht-Nielsen, R., 1989. Kolmogorov's mapping neural network existence theorem. In Proceedings of the International Joint Conference in Neural Networks (Vol. 3, pp. 11-14). Hudaverdi, T., Kulatilake, P.H.S.W. and Kuzu, C., 2011. Prediction of blast fragmentation using multivariate analysis procedures. International Journal for Numerical and Analytical Methods in Geomechanics, 35(12), pp.1318- 1333. Hustrulid, W.A., 1999. Blasting principles for open pit mining: general design concepts. Balkema. Kalinli, A., Acar, M.C. and Gündüz, Z., 2011. New approaches to determine the ultimate bearing capacity of shallow foundations based on artificial neural networks and ant colony optimization. Engineering Geology, 117(1), pp.29- 38. Khamesi, H., Torabi, S.R., Mirzaei-Nasirabad, H. and Ghadiri, Z., 2014. Improving the performance of intelligent back analysis for tunneling using optimized fuzzy systems: case study of the Karaj Subway Line 2 in Iran. Journal of Computing in Civil Engineering, 29(6), p.05014010. Konya, C.J. and Walter, E.J., 1990. Surface blast design. Prentice-Hall. Kuo, R.J., Wang, Y.C. and Tien, F.C., 2010. Integration of artificial neural network and MADA methods for green supplier selection. Journal of cleaner production, 18(12), pp.1161-1170. Kulatilake, P.H.S.W., Qiong, W., Hudaverdi, T. and Kuzu, C., 2010. Mean particle size prediction in rock blast
• 9.
  fragmentation using neuralnetworks. Engineering Geology, 114(3), pp.298-311. Kulatilake, P.H.S.W., Hudaverdi, T. and Wu, Q., 2012. New prediction models for mean particle size in rock blast fragmentation. Geotechnical and Geological Engineering, 30(3), pp.665-684. Latham, J.P., Van Meulen, J. and Dupray, S., 2006. Prediction of fragmentation and yield curves with reference to armourstone production. Engineering geology, 87(1), pp.60-74. Lu, P. and Latham, J.P., 1996, January. In-situ Block Size Distribution Prediction with Special Reference to Discontinuities With Fractal Spacing Distributions. In ISRM International Symposium-EUROCK 96. International Society for Rock Mechanics. Lu, P., 1997. The characterisation and analysis of in-situ and blasted block-size distributions and the blastability of rock masses (Doctoral dissertation). Lyana, K.N., Hareyani, Z., Shah, A.K. and Hazizan, M.M., 2016. Effect of Geological Condition on Degree of Fragmentation in a SimpangPulai Marble Quarry. Procedia Chemistry, 19, pp.694-701. Marinos, P., &Hoek, E. (2000, November). GSI: a geologically friendly tool for rock mass strength estimation. In ISRM international symposium. International Society for Rock Mechanics. Marinos, P., Marinos, V., &Hoek, E. (2007). Geological Strength Index (GSI). A characterization tool for assessing engineering properties for rock masses. Underground works under special conditions. Taylor and Francis, Lisbon, 13- 21. Mehrdanesh, A., Monjezi, M. and Sayadi, A.R., 2017. Evaluation of effect of rock mass properties on fragmentation using robust techniques. Engineering with Computers, pp.1-8. Monjezi, M., Rezaei, M. and Varjani, A.Y., 2009. Prediction of rock fragmentation due to blasting in Gol-E-Gohar iron mine using fuzzy logic. International Journal of Rock Mechanics and Mining Sciences, 46(8), pp.1273-1280. Morin, M.A. and Ficarazzo, F., 2006. Monte Carlo simulation as a tool to predict blasting fragmentation based on the Kuz–Ram model. Computers & geosciences, 32(3), pp.352- 359. Saliu, M.A. and Akande, J.M., 2007. Drilling and Blasting Pattern Selection for Fragmentation Optimization in Raycon Quarry Ore, Ondo State. J. Eng. Applied Sci, 2(12), pp.1768-1773. Shahin MA, Maier HR, Jaksa MB (2002) Predicting settlement of shallow foundations using neural networks. J GeotechGeoenvironEng 128(9):785–793 Singh TN, Singh VK, Sinha S (2006) Prediction of cadmium removal using an artificial neural network and a neuro- fuzzy technique. Mine Water Environ 25(4):214–219 Swingler K (1996) Applying neural networks: a practical guide. Academic Press, New York Tangchawal, S. (2006). Planning and evaluation for quarries: case histories in Thailand. Tangchawal, S. (2017). Quarry evaluation and control of bench blasting. ASEAN Journal on Science and Technology for Development, 23(1), 19-26. Tonnizam Mohamad, E, Armaghani, D. J., Mahdyar, A., Komoo, I., Kassim, K. A., Abdullah, A., &Majid, M. Z. A. (2017). Utilizing regression models to find functions for determining ripping production based on laboratory tests. Measurement 111 (2017) 216–225 Wang, H., Latham, J. P., & Poole, A. (1991). Blast design for armour stone production. Quarry Management, 18(7), 17- 21. Wang H, Latham J-P, Matheson GD. Design of fragmentation blasting in surface excavation. In: Hudson, editor. Rock Characterisation, Thomas Telford, London, 1992;233±238.