PCA-GWO-SVR machine learning applied to prediction of peak vibration velocity of slope blasting

Aiming at the low accuracy of traditional empirical formulas in complex site environment， a predictive model for peak blasting vibration velocity based on grey wolf optimization support vector regression （PCA-GWO-SVR） with principal component analysis （PCA） feature selection is proposed. Based on the monitoring data of blasting excavation of dam abutment trough on the right bank of Baihetan Hydropower Station， the blasting center distance， maximum single-shot charge quantity， elevation differ? ence， longitudinal wave velocity， bore spacing and bore row distance are selected as input parameters， and the characteristic values are selected by data dimension reduction of PCA， and the six selected features are dimensionally reduced to four characteristics with higher correlation. Support vector regression （SVR） is improved by grey wolf optimization algorithm （GWO） to obtain the optimal parameters. Parameters are input into the SVR model for evaluation. The research results show that the PCA-GWO-SVR algorithm has better agreement with the predicted values and the measured values of Sadowski formula， improved Sadowski formu? la， SVR， PCA-SVR， GWO-SVR. The predicted results are more accurate and can predict the peak value of blasting vibration of slope more effectively， which provides help for safety control of blasting construction of slope.