Predicting Blast-induced Ground Vibration in Quarries Using Adaptive Fuzzy Inference Neural Network and Moth–Flame Optimization

Abstract

Blasting is a first preparatory stage that plays a fundamental role in the subsequent operations of an open pit mine. However, its adverse effects can seriously affect the environment and surrounding structures, especially ground vibration, which is measured by peak particle velocity (PPV). The present study proposes a robust model for predicting PPV in open pit mines. An adaptive fuzzy inference neural network (ANFIS) was used as the primary model. The moth–flame optimization (MFO), a swarm-based meta-heuristic algorithm, was integrated to ANFIS, leading to a MFO–ANFIS model, to improve its accuracy. Other intelligent models, such as XGBoost (extreme gradient boosting machine), ANN (artificial neural network), SVM (support vector machine), and two empirical equations (linear and non-linear), were also considered to compare with the proposed MFO–ANFIS model. The findings indicate that the proposed hybrid intelligent MFO–ANFIS model provided the best accuracy (i.e., 98.62%). Meanwhile, the other models provided accuracies of 50.55–96.96%. Among the other models, the artificial intelligence models (i.e., MFO–ANFIS, ANN, XGBoost, and SVM) were recommended to be better in predicting PPV compared to the empirical models. Besides, a sensitivity analysis was also adopted and discussed in this study to understand the role of the input variables in predicting PPV. The results revealed that explosive charge per borehole is more critical than total explosive used per blast; in addition, burden and distance from blast sites are still essential parameters in predicting PPV.