Prediction and Assessment of Back Break by Multivariate Regression Analysis, and Random Forest algorithm in hot strata/fiery seam of open-pit coal mine

Abstract

Coal seam in mines sometimes catches fire due to its property of spontaneous heating. Drilling and blasting techniques are the most economical operation for the removal of overburden and extraction of coal in an open-pit coal mine. Blasting in a fiery seam is one the most dangerous and risky operations and various environmental and technical problems are associated with it. While blasting in a fiery coal seam, it is essential to use the minimum amount of explosive in a hole and the blast should be taken quickly. As the amount of explosive to be used must be comparatively lesser; therefore, it becomes important to perform the blasting in a manner to maximize the utilization of explosive energy. Improper utilization of energy causes back break, which is one of the technical concerns for management due to the in-stability of high wall. This study is a unique attempt to predict back break, especially in fiery coal seam through the experimental blasts and their analysis by Multivariate Regression Analysis (MVRA), and Random Forest algorithm (RFA). Total 26 blasting were conducted at open pit mine. To obtain the most optimum blast design parameter, blast design parameters were varied. The observed back break during trial blasts varies from 1.4 to 10 m. The Sensitivity analysis performed with the collected data and it has been found that stiffness ratio and stemming length are the most influential parameters on generation of back break. The MVRA and RFA analysis have been adopted for accurate prediction of back break. The results show that the RFA analysis predicts the backbreak and achieved the RMSE of ± 0.59 m while in MVRA, RMSE achieved was ± 0.99 m, making RFA more suitable method to be used for backbreak prediction in fiery seam blasting.