Study On Dynamic Prediction Of Mining Subsidence

Subsidence prediction is one of the key contents in mining subsidence, which has great significance for both theoretical study and practice of mining subsidence. After studying different prediction methods for mining subsidence, the dynamic mining subsidence prediction methods have been studied with help of real data obtained from one mine area using MATLAB software. The main works and results are as follows:1. The minding subsidence theory, methods and contents are discussed in this paper, then the advantages and disadvantages of different prediction methods of mining subsidence are summarized and the corresponding formula of dynamic prediction method are derived.2. The different dynamic prediction methods of mining subsidence such as gray model, polynomial fitting estimation, time series analysis, neural networks and adaptive Collocation are analyzed and compared with the help of mining subsidence data form one mine area, and the prediction precision of the above methods are analyzed. The results show that nearly all the prediction methods of mining subsidence can hardly attain high prediction precision especially the AB model when the subsidence of mine area changed greatly and the time separate of observation is much long. However, the adaptive collocation is relatively good. When subsidence of mine area changed smoothly and the time separate of observation is much mean, all the prediction methods can attain relatively high prediction precision. The results of adaptive collocation, AR model and BP methods are better than those of others because of their ability to estimate trend term and random items, and prediction precision of the adaptive Collocation model is the best in overall.