关键词: 岩爆预测, 粗糙集理论, 遗传算法, 径向基函数神经网络, 引水隧洞

Abstract: The rockburst mechanism is very complicate, which is influenced by many factors. By the attribute reduction and the condition properties importance evaluating of rough set theory, it could determine the main factors of rockburst under specific geological conditions, and remove redundant data. Using genetic algorithm (GA) to optimize parameters of radial basis function (RBF) neural network, low-dimensional model inputs were transformed to high-dimensional space through hidden layer; the nonlinear relationship between impact factors and rockburst was fitted. Finally, a rockburst prediction model of GA-RBF neural network based on genetics of rough set theory was established; it had not yet seen the application to rockburst prediction. Based on several theoretical criteria according to the project, the established model was applied to rockburst prediction of practical engineering. It is approved that the evaluation results are consistent with the actual situation well; and it could play a very good guide for latter construction.

Key words: rockburst prediction, rough set theory, genetic algorithm, radial basis function neural network, diversion tunnel