A novel hybrid MARS model based on grey wolf optimizer to improve tunnel blasting vibration prediction

Abstract

Abstract Among several adverse effects of tunnel blasting, ground vibration is the most concerned by blasting engineer. Nevertheless, estimation of ground vibration is fiendishly difficult in comparison with other adverse effects that the internal relationship of explosive, blasting design and rock property are complicated. Traditional empirical-based prediction model has been certain constraints in engineering practice. In this study, a novel hybrid machine learning model is developed that using multivariate adaptive regression splines(MARS) technique and meta-heuristic algorithm termed grey wolf optimizer(GWO). To validate the developed hybrid GWO-MARS model, a standalone MARS, multiple linear regression(MLR) and empirical model were also build for comparison. The results indicate that the satisfactory accuracy of the hybrid GWO-MARS in predicting blasting vibration. The standalone MARS and empirical models are slightly worse than GWO-MARS model. Furthermore, MLR is unsuitable in current investigation.