Detecting the development of explosive methane–air mixtures on a longwall face remains a difficult task. Even when atmospheric monitoring systems and computational fluid dynamics modeling are used to inspect methane concentrations, they are insufficient as a real-time warning system in crucial areas, such as near cutting drums. In our previous work, the long short-term memory algorithm has been adopted in order to predict and manage explosive gas zones in longwall mining operations prior to explosions. It was found that faster and more prominent coverage of accurate real-time explosive gas accumulation predictions is possible using the proposed algorithm. However, the data that was used for training and testing was more than 12TBs. In this work, we evaluated the prediction accuracy of the proposed methodology with a smaller data set. The data was developed and extracted based on CFD model outputs with similar locations and operational characteristics. The results indicate that the algorithm can forecast explosive gas zones in 3D with an overall accuracy of 82.3 percent for a small size dataset (n=1000). |
