Based On Rbfnn Coal And Gas Outburst Prediction Research

Coal and gas outburst is a complex dynamic phenomenon in the underground coal mine. Coal and gas outburst disaster not only caused heavy casualties, but also damaged the mine very severely, is recognized as a serious threat to mine's safety and production of a natural disaster. Currently, coal and gas outburst is still the world's major coal-producing countries solved the problems, and the realization of coal and gas outburst forecast quickly and accurately is an urgent requirement for safety in coal mines. Now there are many coal and gas outburst prediction methods at home and abroad, but mostly used by a single factor. However, the occurrence of coal and gas outburst mechanism is extremely complex, and has many affecting factors which are mostly in a complex nonlinear state. And so far the study has not yet been reached consensus. In recent years, many researchers try to use artificial neural network model to predict coal and gas outburst. Due to the defects and deficiencies in its structure, the predicted effects and results are not very satisfactory.The current prediction models mostly built by BPNN have some disadvantages. In view of the problems existing in the prediction models of coal and gas outburst based on BPNN, and in order to get a fast convergence and more accurate prediction before the outburst accidents, the RBFNN model is built to predict it. The kernel k-means clustering algorithm, which is universal for the samples, is used to determine the central value of the basis function. Its width and its weight are optimized and adjusted by the gradient descent adaptive algorithm and the recursive least square algorithm respectively. And then, the hybrid algorithm and the model are verified with the measured data of coal and gas outburst in China. The simulation results show that the method in the paper has better forecasting accuracy and superior convergence rate than the BPNN and the improved RBFNN based on the classical k-means clustering algorithm, and indicate the practicability and the efficiency of the new model. Then, the each affecting factor's correlation is calculated by the grey relational analysis model in this paper, and use the correlation as the old predicting model's input weights. Finally, the new simulation results show that the input-weighted model has higher prediction accuracy, better classification ability and more representation, and provides a good theoretical support for the prediction study on the coal and gas outburst.