Download as pdf or txt
Download as pdf or txt
You are on page 1of 20

Fragblast 2002, Vol. 6, No. 2, pp.

169188

1385-514x/02/0602-169$16.00 # Swets & Zeitlinger

Blast Design Variation on Blast Fragmentation


D. THORNTON1 , S.S. KANCHIBOTLA2 AND I. BRUNTON3 ABSTRACT
Most blast fragmentation models assume the rock mass properties, explosive properties and blast design variables to be constants and uniformly distributed within a blast. However, in reality all these input variables vary within a blast resulting in variation in the resulting fragmentation size distribution. A stochastic modelling approach is introduced in this paper to quantify this variation. This technique takes the input variables as statistical distributions rather than constants and through several thousand iterations, generates a statistical representation of the expected fragmentation resulting from a production blast. A case study of three production blasts from a large open pit mine are presented and the modelled fragmentation `envelope' shows good agreement with the fragmentation `envelope' estimated from Split image analysis. The various blast-related parameters inuence different parts of the fragmentation distribution, e.g., rock strength and explosive velocity of detonation have most impact on the nes. The technique is used to identify the parameters that have the greatest inuence on various size fractions. Such an analysis will be useful to direct resources to efciently minimise the variation.

Modelling the Impact of Rockmass and

1. INTRODUCTION The Kuz-Ram [1, 2] model or variants of it are widely used to predict blast fragmentation in the mining industry. The JKMRC's blast fragmentation model uses a modied version of the Kuz-Ram model to predict the coarse end of the distribution while the ne end is estimated from theoretical equations. The model has been published on several occasions [3, 4] and a summary of it is provided in the Appendix. The model uses rockmass, blast pattern and explosive properties to predict the entire size distribution of a blast and like the Kuz-Ram model, is a single-hole model, assuming the same parameters for the entire blast volume. However, in reality, each

Australia, 2831 October, 2001. Some have been slightly modied for re-publication in Fragblast. 1 Address correspondence to: D. Thornton, Julius Kruttschnitt Mineral Research Centre, Isles Road, Indooroopilly, Qld 4069, Australia. E-mail: d.thornton@mailbox.uq.edu.au 2 DynoConsult, 15th Floor, Hong Kong Bank Building, 300 Queen Street, Brisbane, Qld 4000, Australia. 3 Julius Kruttschnitt Mineral Research Centre, Isles Road, Indooroopilly, Qld 4069, Australia.

This contribution was presented at the Explo 2001 International Conference in the Hunter Valley, NSW,

170

D. THORNTON ET AL.

input parameter will have some variation associated with it. For example, rock mass properties such as joint spacing and strength can vary considerably within a pattern and burden and spacing could vary by several metres due to drilling inaccuracy and the requirement for inll holes. This paper describes a method of modelling the variation that will result from the inherent variation that exists in large-scale production blasting. The variation associated with the pattern and explosive parameters is controllable to some extent (e.g., burden, spacing, hole depth, stemming length and velocity of detonation (VOD)) while the variation in the rockmass is uncontrollable. Blasts can be designed to react to various rockmass characteristics but there will always exist some random variation by virtue of its heterogeneous nature. In practice, the Kuz-Ram model can be used in one of two ways to predict the size distribution, keeping in mind that it is a single-hole model and as such, there is no interaction between holes. 1. It can predict the `average' size distribution for an entire blast from the `average' values of the input parameters; and 2. It can predict the size distribution around a single hole from the actual values of the parameters associated with that hole. Different parameters inuence different parts of the size distribution [5]. For example, the VOD of the explosive and the intact rock strength predominately affect the amount of nes while the rockmass structure (jointing) and pattern dimensions control the coarse end. As the various parameters change throughout the pattern, the resulting fragmentation will also change. Therefore, the traditional method of producing a single size distribution for an entire blast has no way of gauging the variation that is inherent in the muckpile. There are several benets of being able to predict and understand variation in a blasted muckpile, three are listed below. 1. Drill and blast improvement The quality of blast design implementation will inuence the average fragmentation as well as the variability in fragmentation throughout the muckpile. Tighter patterns with high specic explosive energy (especially shock energy) blasts produce more nes compared to larger patterns and low specic explosive energy. Similarly, if two blasts have the same pattern dimensions, the blast with the least variation (or standard deviation) will produce the more consistent fragmentation. Since different parameters inuence different parts of the size distribution by different amounts, it will be more cost effective to concentrate resources to improving certain aspects of the drilling and blasting. 2. Process control From the perspective of controlling downstream processes, it is important to know how much variation is inherent in the system so that one knows when to intervene and make a change. For example, if the variability in the feed to

ROCKMASS AND BLAST DESIGN VARIATION ON BLAST FRAGMENTATION

171

a SAG mill is solely responsible for throughput randomly uctuating by say, 100 t/h, then while the throughput remains within these limits, changes to the characteristics of the milling circuit could actually make the system more unstable and reduce throughput. 3. Blast design If the rock mass varies signicantly within a pattern, the pattern can be designed to react to this variation. For example, if we know that there is 30% very hard and massive rock in a predominantly soft rockmass, then does one design the entire pattern for a medium rock or for the hardest? Alternatively, if the different domains were in separate denable zones, then it may be better to vary the pattern so that it is optimised for each zone. Stochastic or Monte Carlo simulations can be used to facilitate these process improvement and process control by quantifying the sensitivity of each parameter and quantifying the variability of the resulting fragmentation. 2. STOCHASTIC MODELLING OF FRAGMENTATION VARIATION 2.1. Data Collection Data collection for blast fragmentation modelling can be subdivided into three categories: 1. rock mass characterisation, 2. blast geometry, and 3. explosive properties. The input parameters associated with each category have some degree of variation associated with them. This variation is taken into account by measuring the parameters in the eld and assigning probability distribution functions to these parameters. The three categories listed above require varying degrees of effort to obtain quality data representing input model parameters. For example, rock mass data such as in situ block size distribution is harder to quantify than blasthole depth or spacing. This is because of the limited exposures of the rock mass that can be measured, compared to the ease at which hole depth or spacing can be recorded. Table 1 summarises the parameters required for fragmentation modelling, followed by a more detailed description of each. 2.1.1. Rock Mass Characterisation The simplest possible description of the rock mass for blasting purposes must say something about the intact rock strength and the structure of the rock mass. Blasting will loosen existing rock mass structure to liberate in situ blocks as well as creating new fractures within the intact material. Sources of geological data include

172

D. THORNTON ET AL.

Table 1. Summary of parameters required for fragmentation modelling. Category Rock mass characterisation Parameter In situ block size Unconned compressive strength Young's modulus Density Blast geometry Hole depth Burden Spacing Hole diameter Sub-drill Stemming height Velocity of detonation Relative weight strength Density Method obtained Scanline mapping, core logging, photogrammetry Laboratory testing, drill monitoring Laboratory testing Laboratory testing Tape measure, drill monitoring Survey Survey Blast design, tape measure Tape measure, drill monitoring Tape measure Point to point or continuous VOD monitoring Calculated from detonation code Cup density

Explosive properties

exploration core logging and testing, surface and borehole geophysics, bench mapping and sampling, and equipment performance monitoring [6]. Due to the heterogenous nature of the rock mass, it is important to quantify the variability of these parameters. The estimate of in situ block size is not a routine or simple task and is the subject of active research by a number of groups throughout the world. Problems arise from the limited access available to observe jointing, the complex nature of joint geometry and the statistics of their distribution. There are a number of models that attempt to simulate and predict the size and shape of in situ blocks and their corresponding distributions. These models include JOINTS [7], SIMBLOCK [8] and STEREOBLOCK [9] and those being developed in the JKMRC-ITASCA International Caving Study [10]. All require detailed information on jointing parameters such as joint set spacing, mean orientation, termination statistics and the mean trace length of the set. UCS, Young's Modulus, and density testing of intact rock material are routinely conducted in laboratories. Due to the cost and relative slowness of the testing procedures (sample preparation and transport off-site) in relation to open cut production cycles, point load testing, which is commonly used in geomechanics, is now commonly used by the JKMRC to estimate UCS and Young's Modulus. 2.1.2. Blast Geometry Blast geometry parameters are relatively easy to obtain through surveying methods, and are routinely conducted at some mine sites. These parameters impact on both the

ROCKMASS AND BLAST DESIGN VARIATION ON BLAST FRAGMENTATION

173

ne and coarse ends of the modelled fragmentation distribution. By controlling variation in these parameters a more consistent and uniform fragmentation can be achieved. 2.1.3. Explosive Properties The selection of the most appropriate explosive is an integral part of the blast design and optimisation process. The selected explosive must not only deliver the amount of energy required to fragment and loosen the target rock, it must also be suited to the prevailing conditions [11]. Variations in the explosive properties directly impact on the degree of nes produced by the blast. The nes produced is directly related to the borehole pressure generate by the explosive, which in turn is dependent upon the velocity of detonation (VOD) and density of the explosive, and rock mass strength [5, 12]. 2.2. Methodology There are two types of variation; systematic (e.g., cap rock and faults) and random. Stochastic modelling is not appropriate for modelling systematic variation and therefore is not within the scope of this paper. Instead, the paper will investigate, by example, stochastic modelling as a means of predicting the results of random variation. The term stochastic is used when there is a random component to a model, e.g., values for the input variables are sampled from a statistical population. The most common sampling technique is termed the Monte Carlo method. The name was introduced during World War II as a code name for simulation of problems associated with the development of the atomic bomb. The technique uses random or pseudorandom numbers to sample from a probability distribution. Any given sample may fall anywhere within the range of the input distribution but of course, samples are more likely to be selected from regions of the distribution that have higher probabilities of occurrence. With enough iterations, Monte Carlo sampling `recreates' the input distributions and with the aid of a computer, it is very efcient to conduct hundreds or thousands of `what-if' scenarios, modelling most combinations of input parameters and quantifying the statistical distribution of outcomes. Figure 1 is an illustration of the stochastic modelling methodology in the context of blast fragmentation modelling with the following comments about each step. 1. Data is collected for rock mass characterisation, blast geometry, and explosive properties. 2. This data is analysed to assign a probability distribution to each parameter. Probability distribution functions are dened for each parameter, which represent the range of values that would be expected in the eld.

174

D. THORNTON ET AL.

Fig. 1. Diagram of the stochastic modelling methodology.

3. Monte Carlo simulations are conducted to determine a statistical representation of all possible fragmentation results. 4. The mean and condence intervals (range) of the modelled and measured results are then compared. 3. CASE STUDY A stochastic modelling approach was taken to estimate the fragmentation distribution for three blasts at Porgera gold mine in Papua New Guinea [13]. The blasts were conducted in the same blast domain with powder factors ranging from 0.65 kg/m3 (Blast 1) to 1.12 kg/m3 (Blast 3). The study was conducted as part of a Mine to Mill audit and aimed to increase the crushing and grinding circuit throughput by optimising the ROM size distribution. 3.1. Fragmentation Assessment There are a number of methods that can be used to assess fragmentation for a given blast, including [11]: sieving a representative sample to assess the size distribution of fragments; monitoring production statistics related to fragmentation, such as secondary breakage, load and haul production rates, secondary breakage, and crusher throughput; and  image analysis to give a quantitative measure of fragmentation based on images of the muckpile, run of mine (ROM) and crusher product.
 

The advantage of image analysis techniques over other methods is that it provides quantitative fragmentation data relatively quickly and inexpensively, with little

ROCKMASS AND BLAST DESIGN VARIATION ON BLAST FRAGMENTATION

175

impact on the mine production cycle. The Split image analysis system was used to quantify the fragmentation distribution for each blast. Previous eldwork has demonstrated that Split is a good practical alternative to full-scale screening to quantify the ROM size distribution [3, 14]. The Split program was initially developed at the University of Arizona and has been commercialised by Split Engineering Inc via a licensing agreement with the university. The ore from the blasts was videoed as it was being dumped into the primary crusher and later analysed with Split. The estimation of the ROM size distribution is improved by obtaining many images over a long period of time. This is done to provide a representative sample of the muckpile since images taken from the surface of the muckpile are not a true representation of the entire muckpile due to segregation, e.g., the stemming zone will be coarser. Generally any fragmentation estimate made from image analysis systems will underestimate the nes because they can't be resolved. However, in this study, the size distribution estimated by Split was corrected for nes using the technique described by Atasoy et al. [14]. 3.2. Stochastic Blast Fragmentation Modelling Fragmentation modelling using the JKMRC blast fragmentation model was conducted for three blasts. The stochastic approach described above was used to take into account the variability of model parameters. These parameters are summarised in Tables 24. All of these parameter values were obtained from eld audits or laboratory measurements, except the relative weight strength (RWS) of the explosive, which was obtained from product data sheets. Each parameter was tted with a probability distribution function that best described it (i.e., highest correlation). Normal, Logistic, Log-Normal and Weibull distributions were used. An example of the Logistic distribution applied to hole depth data for Blast 2 is shown in Figure 2 (a log-normal distribution was used for the mean in situ block size).

Table 2. Blast geometry parameters for all blasts. Parameter Blast 1 Hole depth (m) Burden (m) Spacing (m) Hole diameter (mm) Sub-drill (m) Powder factor (kg/m3 ) 10.4 5.5 6.2 200 1.0 0.65 Mean Blast 2 9.1 4.7 5.4 229 0.5 0.87 Blast 3 9.9 4.8 5.5 229 0.5 1.12 0.6 0.4 0.4 5.0 Standard deviation Blast 1 Blast 2 0.7 0.9 0.7 5.0 Blast 3 0.6 0.2 0.3 5.0

176

D. THORNTON ET AL.

Table 3. Explosive parameters (emulsion) for all blasts. Parameter Blast 1 Explosive column length (m) Density (kg/m3 ) Velocity of detonation (m/s) Relative weight strength (% ANFO) 5.3 1250 4500 80 Mean Blast 2 3.7 1250 4500 80 Blast 3 5.4 1270 4500 80 Standard deviation Blast 1 1.6 200 Blast 2 0.8 200 Blast 3 1.3 200

Table 4. Rock mass parameters for all blasts. Parameter Density (m3 ) Youngs modulus (GPa) Unconned compressive strength (MPa) Mean in situ block size (m3 ) Mean 2725 58 138 0.8 Standard deviation 82 8 20 0.3

Fig. 2. Logistic distribution applied to hole depth data for Blast 2.

Five thousand Monte Carlo simulations were conducted for each blast to calculate the mean and 90% condence interval for the modelled fragmentation distribution. The correlation between the modelled and `measured' envelopes for Blast 2 are very good (Fig. 3). The `measured' size distribution is estimated from Split analysis at the

ROCKMASS AND BLAST DESIGN VARIATION ON BLAST FRAGMENTATION

Fig. 3. ROM size distribution for Blast 2 predicted by Split and JKMRC blast fragmentation model.

177

178

D. THORNTON ET AL.

primary crusher with the nes correction obtained from a belt cut of the crusher product. The mean size distributions are almost identical and the 90% envelope matches very closely, deviating only slightly at the ne end. For completeness, the other two blasts are included but the correlation between the model and Split is not as good, particularly in the mid-size fractions. Figure 4 and Figure 5 are the comparison for Blast 1 and Blast 3 respectively. There are several reasons for the discrepancy but it is beyond the scope of this paper to discuss them in detail. The following points provide a brief explanation: The image quality of Blast 1 was poor because all images were obtained at night with inadequate lighting and the camera was too oblique to the trucks;  The fragmentation from Blast 3 was so ne that, at the resolution used, individual particles were not delineated very well; and  The JKMRC Crush Zone blast fragmentation model (see Appendix) often underestimates the amount of mid-size fractions [15]. This is in part an artefact of joining the separate ne and coarse estimations and because is has no physical mechanism for estimating the mid-size fractions. An improvement is the subject of ongoing research.


3.3. Sensitivity Analysis The variation in fragmentation that exists in a production blast is clear from these analyses. For example, in Figure 3 the top size varies from about 700 mm to 3 m and at the other end of the scale, the percent passing 10 mm varies from about 515%. The top size has a direct impact on the primary crusher productivity as well as digging and hauling productivity while the percentage of nes (say 10 mm) has a direct impact on a typical SAG mill circuit. From a control perspective, it would be advantageous to reduce this variation and stochastic modelling provides a mechanism for determining the impact or sensitivity that each variable has on the result. A sensitivity analysis was conducted on the data from Blast 1 to identify the variables that have the greatest inuence on various parts of the modelled fragment size distribution. The analysis is carried out with two different analytical techniques: 1. regression analysis where sampled input variable values are regressed against output values, leading to a measurement of sensitivity by input variable. 2. rank correlation calculation where correlation coefcients are calculated between the output values and each set of sampled input values. These methods were used to determine the top ve input parameters (in order of importance) that effect fragmentation results for selected size fractions (Table 5). For the ne end of the fragmentation distribution, the UCS, explosive length, and explosive VOD are critical input parameters. This is expected as the fragmentation

ROCKMASS AND BLAST DESIGN VARIATION ON BLAST FRAGMENTATION

Fig. 4. ROM size distribution for Blast 1 predicted by Split and JKMRC blast fragmentation model.

179

180
D. THORNTON ET AL.

Fig. 5. ROM size distribution for Blast 3 predicted by Split and JKMRC blast fragmentation model.

ROCKMASS AND BLAST DESIGN VARIATION ON BLAST FRAGMENTATION Table 5. Sensitivity of fragmentation results to input parameters for Blast 1. Size fraction (mm) 1 1 5 10 50 100 250 500 1000 UCS UCS UCS UCS Block size Block size Expl. length Expl. length Top 5 input parameters effecting fragmentation results 2 Expl. length Expl. length Expl. length Expl. length UCS Expl. length Block size Bench height 3 VOD VOD Spacing Block size Expl. length Bench height Bench height Block size 4 Spacing Spacing VOD Spacing Bench height Burden Burden Burden 5

181

Burden Burden Burden Burden Spacing Spacing Spacing Spacing

model uses the UCS and explosive VOD to calculate the radius of crushing around the borehole (see Appendix). The critical parameters for the coarse end of the fragmentation distribution are the in situ block size, explosive length and bench height (stemming length). The form of the model's underlying equations (e.g., a cubed expression is more signicant than a squared one) and the mean and standard deviation of the parameters combine to determine the relative rankings. Similarly, if one wants to control the variation of the results, either the model or the data can be changed. The model can be ignored since changing it has far wider implications than just reducing the variation. From a perspective of providing more consistent fragmentation, it is evident from this analysis that a tight quality control must be exercised on explosive length and VOD, bench height, spacing, and burden. The rockmass parameters of UCS and mean in situ block size are not included since they cannot be controlled. Stochastic modelling can be used to determine the extent of uncontrollable fragmentation variation that is inherent due to rock mass variation, i.e., the variation cannot be reduced beyond this level. 4. CONCLUSIONS Stochastic modelling is a technique where an existing model of a system is used to quantify the variation that is expected in the system. In this paper, the authors have used it to model the variation in fragmentation resulting from production blasts in a large open pit mine. Variations in fragmentation distribution can be modelled by considering the statistical distribution of input parameters. The correlation between the modelled variation and the actual variation (the width of the envelope) is generally good and in one blast where all conditions were favourable, not only the variation but

182

D. THORNTON ET AL.

the values also show excellent agreement (Blast 1). Neither the JKMRC blast fragmentation model nor Split image analysis are perfect and various things can affect the correlation between the mean fragmentation distribution predicted by the two methods as highlighted in the case study. The over-prediction of the mid-size fractions is a known limitation of the model [15] and Split is insensitive to nes [3, 14] and so is prone to error (variation) in this region. Despite the inherent issues associated with fragmentation modelling and assessment, the stochastic modelling approach adds a new `dimension' to the analysis. It provides a mechanism for determining the most inuential parameters, of a particular model, on various size fractions of the size distribution. This information can be used to focus quality control resources to reduce fragmentation variation as a direct result of blast pattern implementation and to determine the minimum amount of variation that can be expected as a result of uncontrollable rock mass variation. 4.1. Limitations The stochastic modelling technique described in this paper certainly has limitations. It is only as good as the underlying model and in this case, the JKMRC blast fragmentation model often lacks accuracy in the mid-size fractions. For the Monte Carlo sampling to be representative of the actual variable populations, one must collect good data and then represent it by an appropriate statistical distribution. Inappropriate statistical distributions can result in sampled values that are impossible such as negative burden or spacing (it is not necessarily appropriate to simply `clip' a normal distribution). Not all variation is random, there will often be a systematic component to the variation. For example, it may be possible to dene and account for such things as hard cap rock, fault zones and perimeter holes when analysing the data but stochastic techniques are not appropriate in these circumstances since the systematic variation is more signicant than the random variation. REFERENCES
1. Kuznetsov, V.M.: The Mean Diameter of Fragments Formed by Blasting Rock. Soviet Mining Science 9(2) (1973), pp. 144148. 2. Cunningham, C.V.B.: Fragmentation Estimations and the Kuz-Ram Model Four Years on. In: Proceedings Second International Symposium on Rock Fragmentation by Blasting. Keystone, Colorado, 1987, pp. 475487. 3. Kanchibotla, S.S., Valery, W. and Morrell, S.: Modelling Fines in Blast Fragmentation and its Impact on Crushing and Grinding. In: Proceedings EXPLO '99. The Australasian Institute of Mining and Metallurgy, Melbourne, 1999, pp. 137144. 4. Thornton, D.M., Kanchibotla, S.S. and Esterle, J.S.: A Fragmentation Model to Estimate ROM Size Distribution of Soft Rock Types. In: Proceedings Twentyseventh Annual Conference on Explosives and Blasting Technique. International Society of Explosives Engineers, Cleveland, 2001, pp. 4153.

ROCKMASS AND BLAST DESIGN VARIATION ON BLAST FRAGMENTATION

183

5. Bergmann, O.R.: Effect of Explosive Properties, Rock Type and Delays on Fragmentation in Large Model Blasts. In: Proceedings First International Symposium on Rock Fragmentation by Blasting. , Sweden, 1983, pp. 7178. Lulea 6. Lizotte, Y.C. and Scoble, M.J.: Geological Control Over Blast Fragmentation. CIM Bulletin (1994, Sept.), pp. 5771. 7. Villaescusa, E.: A Three Dimensional Model of Rock Jointing. PhD thesis, University of Queensland, Brisbane. 8. Peaker, S.M.: Development of a Simple Block Size Distribution for the Classication of Rock Masses. MSc thesis, University of Toronto, Canada, 1990. 9. Hadjigeorgiou, J., Grenon, M. and Lessard, J.F.: Dening in situ Block Size. CIM Bull 91(1020) (1998), pp. 9194. 10. Harries, N. and Eadie, B.: A Model to Predict Primary Fragmentation. Condential Report to Sponsors, International Caving Study, JKMRC, Brisbane. 11. Scott, A. (ed.): Open Pit Blast Design Analysis and Optimisation. JKMRC Monograph No. 1, JKMRC, Brisbane, 1996. 12. Sarma, K.S.: Models for Assessing the Blasting Performance of Explosives. PhD thesis, The University of Queensland (JKMRC), Brisbane, 1994. 13. Grundstrom, C., Kanchibotla, S., Jankovich, A. and Thornton, D.: Blast Fragmentation for Maximising the SAG Mill Throughput at Porgera Gold Mine. In: Proceedings 27th Annual Conference on Explosives and Blasting Technique. International Society of Explosives Engineers, Cleveland, 2001, pp. 383399. 14. Atasoy, Y., Brunton, I., Tapia-Vergara, F. and Kanchibotla, S.S.: Implementation of Split to Estimate the Size Distribution of Rocks in Mining and Milling Operations. In: Proceedings Mine to Mill. The Australasian Institute of Mining and Metallurgy, Melbourne, 1998, pp. 227233. 15. Hall, J. and Brunton, I.: Critical Comparison of JKMRC Blast Fragmentation Models. In: Proceedings Explo 2001. The Australasian Institute of Mining and Metallurgy, Melbourne, 2001. 16. Lownds, C.M.: Computer Modelling of Fragmentation From an Array of Shotholes. In: Proceedings , Sweden, 1983, pp. 455468. First International Symposium on Rock Fragmentation by Blasting. Lulea 17. Kojovic, T., Michaux, S. and McKenzie, C.: Impact of Blast Fragmentation on Crushing and Screening Operations in Quarrying. In: Proceedings EXPLO '95. The Australasian Institute of Mining and Metallurgy, Melbourne, 1995, pp. 427436. 18. Comeau, W.: Explosive Energy Partitioning and Fragment Size Measurement Importance of Correct Evaluation of Fines in Blasted Rock. In: Proceedings Fragblast-5 Workshop on Measurement of Blast Fragmentation, Montreal, 1996, pp. 237240. 19. Aler, J., Du Mouza, J. and Arnold, M.: Evaluation of Blast Fragmentation Efciency and its Prediction by Multivariate Analysis Procedures. Int. J. Rock Mech. Min. Sci and Geomech Abstr. 33(2) (1996), pp. 189196. 20. Lilly, P.A.: An Empirical Method of Assessing Rock Mass Blastability. In: Proceedings Large Open Pit Mining Conference. The Australasian Institute of Mining and Metallurgy and The Institute of Engineers Australia, Newman Combined Group, Melbourne, 1986, pp. 8992. 21. Dukino, R.D.: Prediction of Iron Ore Product Size Using Fracture Mechanics. In: Proceedings National Conference on Ironmaking Resources and Reserves Estimation. The Australasian Institute of Mining and Metallurgy, Melbourne, 1997, pp. 185190. 22. Jaeger, J.G. and Cook, N.G.W.: Fundamentals of Rock Mechanics, 3rd ed. Chapman & Hall, London, 1979.

184

D. THORNTON ET AL.

APPENDIX Blast Fragmentation Modelling Kuz-Ram Model The mean fragment size in the Kuz-Ram model can be calculated by Equation (1) [1].  0:8   V 115 0:633 0:167 x50 A Q 1 Q E where x50 mean or 50% passing size (cm); A rock factor (empirical constant determined by the rock density, strength and jointing); V volume of the blasted rock per blasthole (m3); Q mass of explosive per blasthole (kg); and E relative weight strength of explosive (ANFO 100). An estimate of the entire fragment size distribution is given by the RosinRammler equation: Rx 1 e0:693x50
x n

where R x x x50 n proportion of the material passing screen of size x; screen size; mean size; and uniformity index.

The uniformity index is determined from the blast design through Equation (3) [2].   r   14 B R1 W L 3 n 2:2 1 1 D 2 B H where D B W R H L Charge diameter (mm); Burden (m); Standard deviation of drilling accuracy (m); Spacing/burden ratio; Bench height (m); and Charge length (m).

ROCKMASS AND BLAST DESIGN VARIATION ON BLAST FRAGMENTATION

185

The uniformity parameter in the Kuz-Ram model, which controls the shape of the fragment size distribution function, is not inuenced by the mechanical properties of the rock or the characteristics of the explosive. A possible reason for this is in the nature of the numerical model used for development of the formula for calculation of the uniformity coefcient [16]. Underlying assumptions exclude the possibility for shear failure or plastic deformation of the rock, and fragmentation occurs due to creation of radial cracks under the inuence of the tensile stress. Despite this, the experience of JKMRC research demonstrates that the conventional Kuz-Ram model predicts the coarse end of the distribution with reasonable accuracy but signicantly underestimates the percentage of nes [17, 18]. Fine fragments are the result of fragmentation of competent rock within the in situ rock matrix. Fine rock fragmentation is predominantly controlled by the interaction of the explosive and the rock matrix. Aler, Du Mouza, and Arnold [19], investigated blast fragmentation efciency as a function of blast geometrical parameters and explosive energy parameters using a multivariate statistical method. They concluded that the inuence of geometrical parameters (burden and spacing) could be considered independent of those associated with the explosive energy. In many mining applications this may not be signicant however, in softer ore types and in operations where the nes percentage affects the economics of the operation, it will be important to estimate the nes with reasonable accuracy. The following model improves the prediction of the nes end of the fragmentation distribution. The JKMRC Crush Zone Model Kuz-Ram Model Modications Although the Kuz-Ram model has found favour for predicting ROM distributions, particularly in hard rock applications, it has limitations. The original rock quality factors are based on a subjective description, though Lilly [20] greatly improved on this.  The energy factor is based on explosive energies derived from the ideal detonation codes. However, the energy released by an explosive in rock blasting and its relative partition, is a function of the rock mass properties, explosive properties and connement provided by the blast. In some blasting applications, the effective energy released by an explosive during blasting can be much different to the theoretical energy estimated by the ideal codes [12].


In the JKMRC's modication, the rock factor is calculated by using a blastability index similar to that proposed by Lilly [20]. The rock factor is a rating system to dene the blastability of the rock mass. Lilly [20] describes it as `the ease with which the rock can be fragmented with explosive energy.' It is based on physical-mechanical and

186

D. THORNTON ET AL.

structural properties of the in situ rock mass. Therefore, a rock with a high rock factor value requires a high energy explosive to sufciently fragment the rock and visa versa. The rock factor used in the JKMRC model is similar to that proposed by Lilly [20] with some minor modications to account for measured rockmass properties [3, 4]. Fines Estimation One of the reasons for under estimation of nes in the Kuz-Ram model may be that the nes in a blast are generated by a different breakage mechanism compared to coarse fragments. Dukino [21] illustrated the effect of different loading mechanism on the failure of a specimen (Fig. 6). Under uniaxial tension the specimen will fail due to the extension of the largest aw most favourably oriented to the stress eld. Thus the fracture products are relatively few and large. However under compression, the failure products are smaller and more numerous. Fines in a blast are generated predominantly by the crushing of rock around the blast hole due to compressive failure while the coarse fragments are generated predominantly by tensile failure (Fig. 7). A cylinder of rock around each hole may therefore be dened within which crushing of this nature takes place. The volume of the crushed zone is determined by calculating the point at which the radial stress around the blasthole exceeds the dynamic compressive strength of the rock. The stress around the blasthole can be estimated by assuming that explosion gases apply pressure on the blasthole walls quasi-statically. The following equation is used to estimate the radial stress around the blasthole [22]: r 2  x pb 4 x

Fig. 6. Schematic representation of fracture mechanisms under different loading mechanisms [21].

ROCKMASS AND BLAST DESIGN VARIATION ON BLAST FRAGMENTATION

187

Fig. 7. Photograph of the different damage zones as a result of blasting in concrete.

where x r pb radial stress at a distance `x' from the centre of the blasthole; radius of the blasthole; and borehole pressure.
2 e C d 4

The borehole pressure is estimated by the equation: pb where Cd e velocity of detonation of the explosive; and density of the explosive. 5

Currently a size of 1 mm is used to dene the coarsest particle that results from crushing. This has been chosen on the basis of results from a number of mines where ROM sizings were available. The volume of crushed material can be calculated from the radius of the crush zone. The volume of rock blasted by each hole is calculated from the pattern and the percent of blasted rock smaller than 1 mm is simply the ratio of these two volumes. The uniformity index for the ne end of the distribution (nne) is then calculated by rearranging the RosinRammler equation (Equation (2)).

188

D. THORNTON ET AL.

Generating the Complete Size Distribution The estimated coarse and ne parts of the size distribution are nally joined to produce the entire size distribution. The proportion of 1 mm material determines the point where they join, i.e., as the percentage of ne material increases, the merge point increases (Fig. 8).

Fig. 8. Illustration of how the entire modelled size distribution is made up by separate ne and coarse segments that join at a point determined by the amount of nes.

You might also like