In the feld of civil and mining engineering, blasting operations are widely and frequently used for rock excavation, However, some undesirable environmental problems induced by blasting operations cannot be ignored. Blast-induced fyrock is one important issue induced by blasting operation, which needs to be well predicted to identify the blasting zone’s safety zone. This study introduces an adaptive weighted multi-kernel learning model (AW-MKL) to provide an accurate prediction of blast-induced fyrock distance in Sungun Copper Mine site. The proposed model uses a combination of multi-kernel learning (MKL) approach and adaptive weighting strategy based on weighted Euclidean distance and modifed local outlier factor (MLOF) to maximally improve the predictive ability of kernel ridge regression (KRR). To demonstrate the superiority of the proposed approach, six machine learning models were developed as comparisons, i.e., KRR, RF, GBDT, SVM, M5 Tree, MARS and AdaBoost. The outcomes of the proposed method achieved the highest accuracy in testing phase, with RMSE of 2.05, MAE of 0.98 and VAF of 99.92, which confrmed the strong predictive capability of the proposed AW-MKL in predicting blast-induced fyrock distance.
