An Inventive Approach for Simultaneous Prediction of Mean Fragmentation Size and Peak Particle Velocity Using Futuristic Datasets Through Improved Techniques of Genetic XG Boost Algorithm

Abstract

In the current study, two algorithms, custom XG Boost (CXGBA) and improved genetic XG Boost algorithm (IGXGBA), have been chosen to create an empirical formula for the simultaneous prediction of the mean fragmentation size (MFS) and the peak particle velocity (PPV) with sourced datasets of geo-blast parameters such as spacing burden ratio (S/B), stemming length (T), decking length (DL), firing pattern (FP), total quantity of explosive (TE), maximum charge per delay (MCD), measuring distance (MD), joint angle (JA), joint spanning height (JSP), joint set number (Jn), and rock compressive strength. Advanced technical combinations like K-10 cross-validation, and grid search executed along genetic algorithm processes with a high mutation rate to XGBoost algorithm. All algorithms were executed using Python programming in the Google Colab platform. The results unveiled that IGXGBA is superior and effective in-terms of metric R², RMSE, and MAPE in predicting MFS and PPV. A WEB APP called Bhanwar Blasting Formula (BBF) was created utilizing Google Cloud Platform (GCP) and FLASK APP to benefit practicing mining engineers to predict blasting results easily from the site itself and identify optimization.