Prediction of Water Inrush Risk from Coal Seam Floor Based on Deep Learning Theory

Water inrush from coal seam floor is the result of the interaction between the underlying confined aquifer and the aquiclude rock mass, and its risk degree is affected by the characteristics of aquifer, aquiclude and structural plane. The comprehensive evaluation system is constructed by selecting 7 index factors, i.e. aquifer water pressure, unit water inflow, aquiclude thickness, aquiclude rock mass quality, number of bedding planes, proportion of mudstone and quantitative value of structural complexity. The actual risk level of water inrush in the mined area is characterized by combining fuzzy variable set theory and actual water inflow; The data set is constructed based on the mined area evaluation results. The convolutional neural network (CNN) model is trained and compared with BP neural network model to verify the reliability and accuracy of CNN model; Finally, the CNN model is used to predict the risk of water inrush from the coal seam floor in the unexploited area. The results show that CNN model has higher accuracy than BP model. The area of safety zone, low threat zone, medium threat zone and high threat zone in the unexploited area accounts for 21.48%, 45.78%, 24.11% and 8.63% respectively.