In te Name of God Comparison of Empirical Fragmentation Models at te Gol-Goar Iron Ore Mine Dr. Moammad Taji 24-26 August 2015 Sydney, Australia
Contents 2
Introduction Fragmentation surely is one of te most important and sensitive result of blasting. 3
1- Introduction Empirical models are te first series wic ave been presented. In recent years, due to proceeding computer science, ypotetical models suc as fuzzy logic systems, artificial neural network based systems and etc. were improved in order to increase accuracy and decreasing errors. Despite tis, empirical models still are used because tey ave ig speed in acieving te result and tey do not need to ave special software or algoritms. 4
Review of Blast Fragmentation Models Empirical fragmentation models predict size distribution by making connections between effective parameters of fragmentation. Denis & Gama (1970) SveDeFo (1974) Modified Kuz-Ram (1987) CZM TCM (1999) KCO Cunningam (2005) Larson Kuznetsov (1973) Kuz-Ram (1983) Kou & Rustan (1993) CK (2003) Geibi et al (2009) 5
3- Case Study on Gol-Goar Iron Ore Mine Te Gol-Goar iron mine is located some 55 km soutwest of Sirjan in te province of Kerman. 6
Case Study on Gol-Goar Iron Ore Mine Te total ore reserve: 1135 million ton Te Gol-Goar deposit forms in six separate anomalies at a confinement of about 10 km lengt and 4 km widt. 7
Case Study on Gol-Goar Iron Ore Mine Used initiation system is Nonel. Te used explosive is ANFO or Emulan (Mixed of Bulk Emulite and ANFO). Delay time between different rows is 50 ms. Number of rows and oles per blast is 2-7 and 50-120, respectively. Benc eigts vary 15 m. Blast oles of 251 mm are vertically drilled. Te drilling operation is performed carefully witout any deviation. 8
Measuring Fragmentation Te 20 blast blocks containing Magnetite and Hematite was cosen and examined. On te average for eac blast 14 images in tree different pases of fragmented muckpile (over, middle, end) were taken. Analyzed by: split-desktop software 9
Measuring Fragmentation Blast No. Rock Type 11-234 Magnetite 20.58 44.01 13-76 Magnetite 17.91 60.89 14-13 Magnetite 18.66 42.63 12-155 Magnetite 25.40 60.61 14-14 Magnetite 22.81 45.44 13-78 Magnetite 20.60 51.12 12-156 Magnetite 32.09 59.96 12-162 Magnetite 25.25 56.01 12-164 Magnetite 33.80 91.13 12-167 Hematite 15.99 32.42 13-85 Magnetite 47.07 153.49 9-320 Hematite 14.20 33.94 10-321 Hematite 08.98 14.89 11-239 Magnetite 18.54 35.99 11-240 Hematite 07.67 18.49 13-87 Magnetite 18.74 49.87 11-241 Hematite 15.85 90.30 14-18 Magnetite 22.56 57.45 11-242 Magnetite 27.14 72.27 12-171 Hematite 22.60 44.71 10
Measuring Fragmentation o Split-Desktop as a better tan GoldSize. o As by doing sieve analyze, te level of relative inaccuracy for Split-Desktop was 15% and for GoldSize was 55.1%. o Also, Split-Desktop software could analyze fines. 11
Discussion 12
13
14
In modified Kuz-Ram model (2005) assigned timing is taken into account. Also, some canges ave been done on scoring rock factor. Considering delays of blasts in te model, it did not present acceptable result and just made te model complex. 15
16
17
SveDeFo model is based on te Larson model in wic ole dept and stemming are considered as small flexibility; as remarkable canges in entrance values does not affect on output values. Te process of its curve in compare wit oters is like a flat line wic sould not be like tis. To sum up, tis is a cautious model wic anticipates values far smaller tan real values. 18
19
5-2- Studying size of distribution functions in models Rosin-Rammler function Swebrec function 20
5-2- Studying size of distribution functions in models Kuz-Ram model did not present good values for uniformity index (n). Tese values in some blasting patterns ave eavy differences; te reason can be te land. Differences in BI of blasting patterns can be one of te most effective factor on uniformity index. 21
5-2- Studying size of distribution functions in models 22
5-2- Studying size of distribution functions in models 23
5-2- Studying size of distribution functions in models Values of real uniformity index, calculating by some models for 20 blast. 24
5-2- Studying size of distribution functions in models SveDeFo model as recommended a stable value (n=1.35) for all blasting patterns. It means tat tis model for all blasting patterns wit different type and pattern geometry, present a specific slope for teir size distribution curve. n= 1.35 (fixed) 25
5-2- Studying size of distribution functions in models 26
5-2- Studying size of distribution functions in models 27
5-2- Studying size of distribution functions in models Comparison actual curve and eac predictive fragmentation curves 28
Conclusions After studying 20 blasts in Gol-Goariron ore mine, it was observed tat any of empirical prediction fragmentation models did not present an acceptable performance, except Kuz-Ram model wic presented closer result to reality for Magnetite in some blasting patterns. Te comparison between Rosin-Rammler function and Swebrec function indicates tat te performance of Swebrec in fines was better tan Rosin-Rammler. Tis result for te upper part of fragment size distribution curve, coarse, is te reverse. Presented blastability index in 2002 gets large number in some blasting patterns in quantitative to comparison wit blastability index in 1986 acieved by Lilly wic is good. But in some patterns it does not work well and makes te conditions more complex. 29
Tanks for your attention taji@ymail.com 30