Quarry blast evaluation system for rock fragmentation

Abstract

Blasting produces energy to fragment the rock mass in mining, quarry and civil engineering projects. In mining and quarrying operation, blasting aims to extract the largest possible quantity of rock at minimum cost in the safest manner with minimum side effects such as ground vibration, flyrock and noise. Hence, blast design plays a vital role. Poor blast design is harmful to the surrounding and the desired rock fragmentation cannot be obtained. It affects the drilling and blasting cost as well as the efficiency of all the subsystems such as loading, hauling and crushing in mining operations. Therefore, this research aims to evaluate the significant parameters related to the blasting operation and establish a blast design model for better prediction of particle size of rock fragmentation. The study will focus on the granite quarry operation. Terrestrial and aerial survey technology namely Terrestrial Laser Scanning (TLS) and Unmanned Aerial Vehicle (UAV) respectively, are carried out during pre and post of the blasting for discontinuity mapping. Then the engineering properties of the rock are determined through the laboratory work. These properties are then utilised in Discrete Element Method (DEM) numerical simulation using Bonded Particle Method (BPM) and Particle Blast Method (PBM) to predict the blasting performance. Once the model is verified, the influencing parameters are further investigated through a series of parametric study on rock fragmentation. The parameters involved are burden, spacing, stemming, hole diameter, bench height and powder factor. The relationship between the spacing to burden (S/B) ratio, stemming to burden (T/B) ratio, burden to hole diameter (B/D) ratio, bench height to burden (BH/B) ratio and powder factor against the predict mean particle size (d50) and uniformity index (n) is studied. Furthermore, a machine learning algorithm is utilized to predict the d50, sieve size at 80% material passing (d80) and parameters n as the output product. MATLAB and RapidMiner software of machine learning algorithms with four different learnings, which are Linear Regression, Decision Tree, Random Forest and Support Vector Machine (SVM), are utilised in this study. Comparisons of the output predictions between the learning algorithms are conducted and the influential parameters for the predictions are identified. The results show that Random Forest learning is chosen as the best machine learning, since the results obtained show the highest R-squared value, with the lowest Root Mean Square Error (RMSE) value. The best R-squared and RMSE results for prediction of mean particle size are 0.85 and 0.046, respectively. In addition, the best R-Squared and RMSE results for prediction of uniformity index are 0.75 and 0.324, respectively. A quarry blast evaluation system for prediction of rock fragmentation was developed. The blast evaluation system and prediction for rock fragmentation developed is focused on open pit quarry but this also may be applicable to rock slope. The blasting evaluation system established in this study will be very beneficial to policymakers, practitioners and designers associated with quarry blasting for a safe quarry blasting operation. Hence this will help the engineer to make crucial decisions during the planning, design and operational stages of a quarry.