Application of machine learning algorithms for prediction of Blast-induced Ground Vibration in View of Stiffness Ratio, Energy Coverage and Scaled Distance

The purpose of this research work is to predict blast induced ground vibration in surface mine by using classical and machine learning algorithms. For the purpose of minimizing blast-induced ground vibration to acceptable levels, the level of vibration must be predicted. Blast-induced ground vibration is defined peak particle velocity (ppv) in the ground. All data used to estimation were obtained by observing real blasting operations. After the measuring of the peak particle velocity, models of the prediction were created using independent site parameters. Most of the data is used to train the model, while remaining part is used for testing. The models were created using independent blasting parameters proportionally. Thus, more parameters are included in the models without complicating the models. A thorough validation process was conducted utilizing a diverse set of nine error criteria. Artificial intelligence models have been found to outperform traditional methods in predicting ground vibration. The mean absolute error values were found to be 1.42, 1.54, and 1.78 for ANFIS, GPR, and SVM, respectively. A similar situation is observed for other error criteria as well. ANFIS appears to be the most effective model for predicting ground vibration.