Predicting Ground Vibrations Due to Mine Blasting Using a Novel Artificial Neural Network-Based Cuckoo Search Optimization

Blasting plays a fundamental role in rock fragmentation, and it is the first preparatory stage in the mining extraction process. However, its undesirable effects, mostly ground vibration, can cause severe damages to the surroundings, such as cracks/collapses of buildings, instability of slopes, deformation of underground space, affect underground water, to name a few. Therefore, the primary purpose of this study was to predict the intensity of ground vibration induced by mine blasting operations with high accuracy, aiming to reduce the severe damages to the surroundings. A novel artificial neural network (ANN)-based cuckoo search optimization (CSO), named as CSO-ANN model, was proposed for this aim based on 118 blasting events that were collected at a quarry mine in Vietnam. Besides, stand-alone models, such as ANN, support vector machine (SVM), tree-based ensembles, and two empirical equations (i.e., USBM and Ambraseys), were considered and developed for comparative evaluation of the performance of the proposed CSO-ANN model. Afterwards, they were tested and validated based on three blasting events in practical engineering. The results revealed that the CSO algorithm significantly improved the performance of the ANN model. In addition, the comparative results showed that the accuracy of the proposed hybrid CSO-ANN model was superior to the other models with MAE (mean absolute error) of 0.178, RMSE (root-mean-squared error) of 0.246, R-2 (square of the correlation coefficient) of 0.990, VAF (variance accounted for) of 98.668, and a20-index of 1.0. Meanwhile, the other models only yielded performances in the range of 0.257-0.652 for RMSE, 0.932-0.987 for R-2, 20.942-98.542 for VAF and 0.227-0.955 for a20-index. The findings also indicated that explosive charge per borehole has a special relationship with ground vibration intensity. It should be considered and used instead of total explosive charge per blast in some cases, especially for the empirical models.

Automated summary

The study aimed to predict the intensity of ground vibration induced by mine blasting operations to reduce severe damages to the surroundings. A novel CSO-ANN model was proposed and validated based on 118 blasting events, showing that the CSO algorithm significantly improved the performance of the ANN model.