Prediction of blast induced ground vibration and its associated dominating frequency using a comprehensive support vector machine model

Prediction of ground vibration (peak particle velocity, PPV) and frequency is an important task in geoscience. Till date, many empirical equations are derived from conducting trial blasts in the mines, to arrive at attenuation predictors. These predictors are usually site dependant and hold good for one particular type of rock/strata. These predictors cannot be used for geologically different rocks/strata. This paper presents the application of support vector machine (SVM) model, for predicting blast induced ground vibration and frequency, using a one comprehensive model for three different types of rock. Inputs to the SVM model formulation involve twenty one parameters consisting of geotechnical properties of the rock mass, blast design geometry and explosive properties, and the output of the formulated model is peak particle velocity and its associated frequency. The data has been acquired from eight limestone mine, five coal mine and two manganese mines. The comprehensive SVM model is formulated using 400 datasets and tested with 143 datasets. The performance indicators to the model not only justify its formulation but also the advantages it has over site specific predictor models. The SVM model not only predicts PPV, with higher correlation coefficients as compared to predictor equations, but also the associated frequency.