Utilization of a nonlinear support vector machine to predict blasting vibration characteristic parameters in opencast mine

Characteristic parameters of blasting vibration (BVCP) have great effects on its damage level. The prediction of BVCP is helpful to study blasting vibration effect. In this paper, an attempt has been made to predict blast-induced ground vibration using support vector machine (SVM) to avoid the limitation of the prediction with only one index and to improve the prediction precision. A Grid search method-based SVM prediction model for BVCP was established on the basis of nonlinear model-based SVM. To construct the model, nine factors affecting blasting vibration characteristic variables are taken as input parameters, whereas, peak particle velocity (PPV), dominant frequency (Df) and its time duration (Dt) are considered as output parameters. A database consisting of 108 datasets was collected from Tonglvshan copper mine in China. From the prepared database, 93 datasets were used for the training of the model, whereas 15 randomly selected datasets were used for the validation of the SVM model. To compare the performance of the developed SVM model with that of artificial neural network (ANN) model, the same database was applied. Superiority of the proposed SVM model over ANN model was examined by calculated coefficient of determination for predicted and measured values of PPV, Df and Dt. Concluded remark is that the prediction's BVCP can reliably be estimated from the indirect methods using SVM analysis. Streszczenie. Przy przewidywaniu efektow i szkod wibracji wybuchowych wazny jest parametr BVCP - blasting vibration characteristic parameter. W artykule przedstawiono model matematyczny do prognozowania efektow drgan wybuchowych z wykorzystaniem metody SVM. (Wykorzystanie metody SVM do prognozowania parametrow wibracji wybuchowych w kopalniach odkrywkowych)