Machine learning model for predicting blast fume dilution time in underground working areas

In underground mines, accurate prediction of dilution time is crucial to protect miners from toxic fumes as well as minimize production delays and ventilation costs. Previous researchers have used empirical equations, mine ventilation software, and computational fluid dynamics to calculate dilution time. These traditional methods have high computational cost and design limitations. In this research, machine learning (ML) techniques and data analytics were used to: (i) develop a ML model to predict dilution time based on five on-site blasting and auxiliary ventilation parameters and (ii) determine the order of importance of these parameters. The advantages of using the ML model over the traditional methods are its ability to consider a large number of factors, detect complex nonlinear relationships, and predict the dilution time near-instantaneously. Results show that the normalized artificial neural network has the best prediction capability with an R2 of 0.987. Among the blasting and ventilation parameters measured, air volumetric flow rate significantly affected the dilution time, followed by duct distance from face and the amount of explosive.