Prediction of blasting-induced flyrock using M5P tree technique

Summary: Based on statistics, flyrock is the main reason of 20 to 40% of blasting induced accidents in mines. Therefore, the accurate prediction of flyrock has a remarkable role on reducing its detrimental effects. In this paper, a model tree was developed for flyrock prediction using M5P technique. This model was trained and tested by the blasting database of Sungun copper mine. The results showed that the proposed model can estimate the flyrock with an acceptable error. Flyrock is one of the most challenging safety issues of blasting operation in open pit mines, which always threatens the safety of personnel and equipment. Thus, the accurate prediction of flyrock seems necessary for determination of safe blasting zone. During past years, many empirical models have been developed using artificial intelligence techniques for flyrock prediction. Most of these models do not indicate the relationship between input and output parameters, so they are opaque and vague. The main purpose of this paper is to construct a transparent and understandable model for flyrock prediction. Methodology and Approache: In this paper, a model is developed for flyrock prediction using M5P technique. This technique presents a tree structure which contains linear equations in its leaves and using these equations the flyrock can be predicted easily. In this model, the flyrock distance is estimated using the most important controllable blasting parameters (i.e. burden, spacing, stemming, blasthole length, blasthole diameter, powder factor and mean charge per blasthole). Results and Conclusions: The accuracy and efficiency of proposed model was evaluated using various statistical indices such as coefficient and determination (R2), variance account for (VAF) and root mean square error (RMSE). These indices were obtained 92.1%, 92%, and 3.9, respectively. Therefore, it can be concluded that M5P technique is a suitable and efficient means that provides an acceptable prediction for flyrock. Furthermore, the results indicated that the burden and blasthole diameter are the most and the least effective parameters on flyrock prediction, respectively. The output of this model can be considered as a preliminary estimation of flyrock, based on which the risk of hazards and potential accidents can be reduced to a desirable extent.