TROPICAL GEOGRAPHY ›› 2015, Vol. 35 ›› Issue (5): 770-776.doi: 10.13284/j.cnki.rddl.002756

Previous Articles     Next Articles

Susceptibility Prediction of Underground Mining Collapse Based on GIS and BP Neural Network

ZHANG Guolia,YANG Baolinb,ZHANG Zhia,WANG Shaojuna   

  1. (a.School of Public Management;b.School of Earth Sciences,China University of Geosciences,Wuhan 430074,China)
  • Online:2015-09-30 Published:2015-09-30


By using GIS, spatial analyses, including extraction, classification, valuation, statistics and normalization, were made with the data of the study areas in Chengchao Iron Mine and Daye Iron Mine of Hubei province. An index data set, including elevation, slope, strata, the recent distance from underground mines,the distribution density of underground mines, the ratio of mining thickness and depth, the buffer of alteration contact zone, groundwater depth, and the surface feature types, was constructed to assess the susceptibility of collapse in mining area. With IDL language to call Matlab neural network toolbox, the index data set of the study area in 2011and 2012 was used as input data, and the susceptibility was used as expected output. So the model based on BP neural network for predicting the susceptibility of underground mining collapse was constructed. By selecting and optimizing the training sample, this model realized the prediction for the susceptibility of collapse in 2013. The results indicated that the underground mining collapse area accounted for 89.91% of the total area highly susceptible to collapse; with the increase of the susceptibility level, the ratio of the underground mining collapse area to the susceptible area increased; the distribution of underground mining collapse was of obvious zonality, and the high susceptibility area distributed basically along the contact zone between the rock mass and the surrounding rock. The model solved the problem of nonlinear mapping in collapse prediction, as the predicted results were in accord with the actual survey. The results show that BP neural network model and GIS technology for evaluating the susceptibility of underground mining collapse would have a certain feasibility.

Key words: GIS, BP neural network, underground mining collapse, susceptibility prediction