| Rockburst is a geological disaster with strong destructive power.The frequency and intensity of subterranean space development and construction are rising day by day,limiting the deeper growth of underground construction in the long term.In light of the current state of this project,this dissertation investigated rockburst influencing factors and forecasting methods from the perspective of rockburst advance forecasting,and developed a reasonable and effective rockburst forecasting model that could be used as a guide for rockburst risk initiatives prevention and management.The main research work was as follows:(1)The rockburst prediction benchmark index dependability was proved.The significant key elements of rockburst and their internal relationship with rockburst intensity were examined from three perspectives: lithology,wall rock stress,and excavation circumstances,with the degree of correlation between each index and rockburst intensity determined using the Spearman analysis method;(2)Plate crack buckling rockburst vertical stress situation was determined.The cusp catastrophe model of plate crack buckling rock explosion was created using the potential energy balance equation of the plate beam,and the vertical in-situ stress condition was produced.The analytical results were shown to be in excellent accord with the real scenario when combined with engineering examples;(3)Two models for predicting rock bursts were developed.Because the multidimensional redundancy of rockburst prediction indexes and data made model development difficult,this dissertation optimized them from two perspectives.The catastrophe progression method created the hierarchical structure of the rockburst prediction index,the catastrophe model processed the rockburst data,and the catastrophe progression method-BP neural network rockburst prediction model was created by combining the catastrophe progression method and the BP neural network.The characteristic information of rockburst data was extracted by designing and building the kernel principal component analysis approach with the capacity of dimension reduction of nonlinear absorption info,and the KPCA-BP neural network rockburst forecast was formed by blending with BP neural network;(4)The prediction model’s accuracy was checked.To compare and measure the forecast effectiveness of the model developed in this dissertation,various methodologies were employed.The findings revealed that the two complete models for rockburst prediction developed in this dissertation had great accuracy,both over 90%,and both frameworks were better and more efficient.The two prediction models developed in this dissertation were applied to the rockburst prediction of the Duoxiongla Tunnel Project in Tibet under various working conditions,and the prediction results were generally compatible with the real setting,indicating that the two models developed in this dissertation had practical engineering value. |
