Prediction of an environmental impact of tunnel blasting using ordinary artificial neural network, particle swarm and Dragonfly optimized artificial neural networks

Abstract

Blasting is an intrinsic component of surface and underground excavation but it is associated with adverse environmental effects that can threaten the safety of lives and property. This study, therefore, proposed artificial intelligence (AI) based models for predicting the air overpressure (Aop) in the tunnel blasting. Three AI models which are ordinary artificial neural network (ANN), particle swarm optimized artificial neural network (PSO-ANN) and Dragonfly optimized artificial neural network (DA-ANN) are proposed. The input parameters into the models are the charge per delay (Cd), the number of holes (Nh), distance from the measuring station to the blasting point (Dm), and the rock mass rating (RMR) while the Aop is targeted output. The model parameters were obtained through field measurements and laboratory experiment. The performance of the models was evaluated using the coefficient of determination (R2), mean-squared error (MSE), mean absolute percentage error (MAPE), and the variance accounted for (VAF). Out of the different model simulations, the PSO-ANN model with 4-15-15-1 architecture performed best with R2 of 1, 0.984, 1, 0.9985, MSE of 0.0004, 0.125, 5.5E-0.6, 0.018, MAPE of 0.004, 0.152, 0.002, 0.025, and VAF of 99.996, 98.29, 1, 98.85 for the respective training, testing, validation and overall datasets. The selected model was compared with the MLR model and empirical model predictions. The proposed model outperformed them. Hence, the proposed model can predict Aop with a high degree of accuracy.