Applying machine learning techniques to mine ventilation control systems

Abstract

The purpose of the research is determination of mine ventilation system regulators positions providing required airflow on ventilated directions. Currently regulators positions are set iteratively that causes hunting. It is proposed to use historical data of the system for defining regulators functional dependencies on required airflow values with consideration of temporal variability of a ventilation network. The problem is solved by a regression model based on neural networks. Consequently, a set of model parameters is defined and the control algorithm of the system is modified for using a historical data set.