Forecasting of Methane Hazard State in the Exploitation Wall Using Neural-Fuzzy System

In hard coal mines the series of natural hazards take place, one of the most dangerous are gaseous hazards. However, the most commonly occurring gaseous hazard is methane hazard. Methane is odorless and colorless, lighter than air organic gas, of which large quantities are emitted during an exploitation of hard coal. It is inflammable gas which with appropriate air concentration becomes a very dangerous explosive gas. In order to ensure safety of the staff, the critical concentration of this gas cannot be allowed to occur in mining atmosphere. Therefore, it is reasonable to apply mechatronic systems which task is to register current methane concentrations in mining atmosphere, also to develop analytical systems in order to forecast these concentrations. Application of advanced artificial intelligence systems provides opportunities for effective forecasting of methane concentrations in underground headings. This forecast is based on the registered measurement data and determined ventilated parameters. In the paper, an example of application of artificial intelligence methods to build an analyzing methane hazard model in the region of exploited wall was presented. The way of formation of neural-fuzzy system based on numerical data was described. Hybrid system was built to determine current level of methane hazard and forecast its future concentration in the region of longwall during normal mining operations. For determination of system parameters automatically, methods of knowledge recovery (based on fuzzy clustering of measured data) and adaptive learning algorithm of neural-fuzzy system were used. Developed tool gives a lot of opportunities for practical application.