Predicting blast-induced peak particle velocity using BGAMs, ANN and SVM: a case study at the Nui Beo open-pit coal mine in Vietnam

Abstract

One of the most adverse effects encountered during blasting in open-pit mines is ground vibration. The peak particle velocity (PPV) is a measure used for ground vibrations; however, accurate prediction of PPV is challenging for blasters as well as managers. Herein, boosted generalised additive models (BGAMs) were applied for estimating the effects of blast-induced PPV. An empirical equation, support vector machine (SVM) and artificial neural network (ANN) were also adapted and used to approximate the blast-induced PPV for comparison. Herein, a database covering 79 blasting cases at Nui Beo’s open-pit coal mine, Vietnam, were used as a case example. Several performance indicators such as the coefficient of determination (R2), root-mean-square error (RMSE) and mean absolute error (MAE) were used to evaluate the quality of each predictive model. According to the results, the proposed BGAM performed better than the other models, yielding the highest accuracy with an R2 of 0.990, RMSE of 0.582 and MAE of 0.430. ANN and SVM models exhibited only slightly lower performance, while the empirical technique had the worst performance. Two testing blasts were performed to validate the accuracy of the developed BGAMs in practical engineering and the results showed that the BGAMs provided high accuracy than other models. Results also revealed that the elevation difference between the blasting site and monitoring point is one of the predominant parameters governing the PPV predictive models.