Comparison Between Neural Networks and Multiple Regression Analysis to Predict Rock Fragmentation in Open-Pit Mines

Abstract

Keywords Rock fragmentation Neural networks Multiple regression Gol-e-Gohar iron mine Open-pit mines1 IntroductionBlasting is the most frequently used means of quarryingand mining rock excavation, and the quality of the frag-mentation of rock mass is a major concern of any blastingoperation (Latham and Lu 1999).The prediction and assessment of the rock size distri-bution produced by blasting are important concerns inunderstanding the blasting process. The rock fragmentationdistribution influences downstream processes such as loadand haul rates, crushing and grinding performance, andalso ore recovery in beneficiation processes (Michaud et al.1997). In the studies at Que´bec Cartier Mines, MacKenzie(1966) found that the efficiency of all the subsystems ofmining is dependent on the fragmentation. Additionally,uniform particle size distribution also eliminates the needfor secondary blasting of large boulders.It should be noted that many controllable and uncon-trollable factors influence rock fragmentation. Effectivefactors influencing fragmentation are classified into threecategories: blast design parameters, explosive properties,and rock mass properties.Burden, spacing between boreholes, bench height, drill-hole diameter, hole length, charge depth, stem height,subdrilling, drilling pattern (square or staggered), holeinclination (vertical or inclined), blasting direction, andblasting sequence (instantaneous or delayed) are blastdesign parameters, which are controllable. The secondgroup consists of explosive properties. Explosive type(ANFO, water gel, emulsion, or dynamite), its density(varies between 0.80 and 1.60 g/cm