The document proposes using machine learning techniques to improve control of mine ventilation systems. Specifically, it suggests using neural networks to define the functional dependencies between regulator positions and required airflow values over time. This would account for changes in the ventilation network and reduce issues like hunting. The researchers trained networks on historical data to predict regulator angles from required airflow. Adding a temporal parameter reflecting previous system states improved long-term predictions. The proposed control algorithm aims to efficiently and safely meet airflow requirements using the neural network models.