Green policy for managing blasting induced dust dispersion in open-pit mines using probability-based deep learning algorithm

Abstract

Many artificial intelligence techniques have been employed in forecasting dust pollution due to bench blasting in mining operations. Whereas considering the uncertainty of blasting outcomes is an essential issue and is required to overcome. Therefore, this study integrates Monte-Carlo simulations (MCs) and artificial neural networks (ANN). A probability-based advanced version of the artificial neural network, i.e., deep neural network (DNN), is developed for simultaneously predicting particle matter (PM) and total suspended particulate (TSP) based on gathering data from the Asgarabad2 limestone mine that has been explored in Iran. A model was first developed using a probability-based deep neural network (PDNN) to predict dust pollution. Based on the obtained results of PDNN [ i.e., R2 (0.999 and 0.991), RMSE (1.259 and 3.424), and VAF (99.941 and 99.033) for the PM10 predictive model and R2 (0.999 and 0.998), RMSE (0.772 and 0.939), and VAF (99.956 and 99.735) for TSP predictive model], a considerable improvement in accuracy of PM10 and TSP predictive model is obtained by developing this new predictive model. Then, sensitivity analysis of PM10 and TSP to effective parameters was performed. Results indicated that wind speed is the most influential parameter on both PM10 and TSP. Therefore, wind analysis has been performed to specify predominant wind direction and average wind speed as well as identify areas affected by dust particles.