DEVELOPMENT OF MODELS FOR OPTIMUM ROCK FRAGMENTATION PREDICTION USING ROCK MASS PROPERTIES AND BLASTING PARAMETERS

Abstract

Prediction of size distribution in rock fragmentation is the key to projection and evaluation of a successful mining investment, as it determines the efficiency of every other unit operation as well as the overall economics of mining operation. However, the heterogeneous nature of rocks and inability to factor all parameters that influence fragmentation into existing models made them site-specific with diverse degree of accuracy. Consequently, there is no existing model for prediction of muck-pile size distribution that fit into all condition. This research is centred to the development of models that will establish suitable geometric parameters and effectively predict the distribution of muck-pile based on rock engineering and geological properties. These were achieved by evaluating fresh rock and discontinuity properties (uniaxial compressive strength (UCS), porosity (n), specific gravity (GS), rebound hardness number (RN) rock mass rating (RMR) and blastability index (BI)) in accordance with standard procedures as well as blast geometric parameters (powder factor (PF), energy factor (EF), quantity of charge (Q), burden (B), hole diameter (D), drilled-hole pattern, spacing to burden ratio, stemming to burden ratio, stiffness factor and the burden to hole diameter ratio) and estimating average mean fragment size using Split Desktop (version 3.1) software for various image analysis in thirty different blasting locations in Ondo, Ogun, Oyo, Edo and FCT, Nigeria. These led to the development two predictive models using Regression Statistics and Rock Engineering System of modelling method. Also, fragmentation risk factor for the selected Quarries were evaluated using the vulnerability index method. The results show that the strength of rock mass ranged from moderate to strong rocks, while the fragmentation risk assessment revealed that the risk factor varied from low to moderate and moderate to high. It was also observed that the vulnerability index increases with the mean fragment size of muck-pile. The results of the interaction matrix of the rock engineering system shows that all the sixteen selected parameters have a reasonable interaction factor. More so, burden has the highest interaction with a weighty factor of 10.05% while the ratio of stemming to burden has the least factor of 3.271%. The Rock Engineering System and the Regression models proposed in this research show high level of accuracy than the existing Kuz-Ram and Modified Kuz-Ram models when the predicted mean fragment size for each of the locations was compared to the actual measured mean size. This was ascertained with the correlation coefficient value (R2) of the Rock Engineering System, Regression, Kuz-Ram, and Modified Kuz-Ram model of 0.9212, 0.8554, 0.2379 and 0.4132 respectively. Likewise, the Variance Accounted For (VAF) and Root Mean Square Error (RMSE) for both the testing and training datasets shows that the Rock Engineering System (RES) model performed better in the prediction of mean fragment size when compared to other models. The RES model has the RMSE of 3.946 and 1.46 for both training and testing datasets respectively. Hence, it can predict rock fragmentation values with high accuracy in comparison to Regression, Kuz-Ram, and Modified Kuz-Ram models having values of 7.205 and 2.46, 3694.16 and 52.34, and 342.11 and 14.44 for training and testing datasets respectively. These models had successfully bridged the gap of interpreting the relationship between uncontrollable and controllable rock parameters, which are necessary for the prediction of muck-pile fragmentation size. On the whole it was recommended that for effective blasting output, mining engineers should make use of models (RES and Regression) developed in this research.