Abstract:
In order to improve the efficiency and simplicity of mining area collapse prediction, it is proposed to use neural network algorithm for analysis and prediction. Eight hazard factors including hydrological characteristics, geological structure, final mining time, overburden strata strength, roof span, mining depth, coal mining height, and number of spatially superposed layers are selected as the evaluation indicators of collapse susceptibility, and the mining area is divided into 4 classes of collapse susceptibility. The spatial analysis function of GIS is used to raster the predicted data. The central and southern units of the mining area are used as samples for training to construct a BP neural network model, and the remaining northern units are used to test the prediction effect. Combine the GIS system to image the output results of the model to obtain the zoning map of the collapse susceptibility of the mining area. The results show that the neural network model achieves a rapid convergence effect during the training process, and the prediction result is basically consistent with the collapse that has occurred, and it is suitable for initial prediction of mining area collapse.