The Prediction Of Mining Subsidence In Mountain Area Based On BP Neural Network

The research of mining subsidence in mountain area is the focus part of the subsidence research. For the mountainous and rich coal resources region—ShanXi, it has especially significant meaning in this part.Due to the influence factors as topography and the complexity of the rock mass underground mining, land sinking law in a mountainous area is different from that on flat ground. Thus, the prediction result from traditional model is differing from the practical situation.Neutral network was developed in late 80's last century as an emerging discipline, which has higher fitting capacity and prediction accuracy in non-linear system. But So far, for less application in mountainous mining subsidence prediction, neutral network is still at the exploratory stage.Dongshan Coal Mine is the key coal mine enterprise in ShanXi province, and the region there has complex terrain, which is a typical mountainous mining. So, the paper selects 71505 working face as the research region, and adopts a mature BP algorithm with strong features and unique advantages in solving the non-linear issue, then establishes prediction BP neural network model. Finally introduces BP neural network into mountain mining subsidence, and tries to use it in researching Dongshan coal mining subsidence problem.In this paper, measured data from observation stations is arranged firstly, and then law of surface movement and deformation is analyzed. After that, twelve influence factors are chosen and quantified as slope, aspect, mechanical properties of overlying strata, coal bed pitch, ratio of depth and thickness, coal-mining method and roof management, size of goaf, face advances rate, angle position of observation points, distance between face center and measuring points, internal and external position of observation points, length between observation point. Then the corresponding input and output layer are determined and BP natural network model is built for predicting mountainous mining subsidence in MATLAB. Finally, surface movement and deformation in Dongshan mine is predicted by this model, and examines with measured data. The results show that the prediction value basically reflects the trend of surface movement and deformation, and the prediction accuracy is satisfactory.Finally, the paper tentatively applies the established BP model to Haoyu mine in order to test the universality of this model.The research shows it is possible to use BP neural network to predict mountain mining subsidence theoretically and practically.