Prediction of Blast-Induced Ground Vibration Using Principal Component Analysis–Based Classification and Logarithmic Regression Technique

Ground vibration is one of the major hazards produced by rock-blasting operation. The accurate prediction of vibration is necessary for designing controlled blasting parameters. The existing vibration predictors consider maximum explosive charge weight per delay and distance as the parameters responsible for ground vibration. These predictors are based on the assumption that the geometrical parameters of the blast will be constant for a site. However, the mining sites with bigger production targets have varying geometrical parameters to suit the excavator utility. Accordingly, the other blast design parameters will also have an impact on ground vibration intensity. A principal component analysis is a dimension reduction technique. This technique along with multivariate logarithmic regression has been used in this paper to predict the ground vibration. The technique has classified the blast design parameters into four principal components. The regression with the scores from these principal components has been carried out. The evaluation of the model performance of predictors along with the existing empirical predictors has been carried out using R2 and RMSE values. The evaluation suggests that the predictor with logarithmic regression followed by principal component analysis gives better performance with respect to the existing empirical predictors.