| Summary: | In view of problems of randomness and subjectivity in determining weight of existing rockburst prediction methods，a discrete Hopfield neural network (DHNN) model for prediction of classification of rockburst intensity was proposed。The model selects stress coefficient, rockbrittleness coefficient and elastic energy index as evaluation index, divides rockburst grade into 4 stages, such as strong rockburst, medium rockburst, weak rockburst and no rockburst, then encodes them. The model need't normalize sample data with simpler encoding ,lesser iterations of network and better associative memory ability, only be converted to '1' and '-1' of the two value model, therefore, the classification prediction of rockburst intensity is more scientific and reasonable. The model can provide a new way for classification prediction of rockburst intensity in deep underground engineering. The prediction results of typical rockburst engineering examples prove the correctness of the model.
|
|---|