Contribution Of Internal And External Factors And Disaster Prediction Methods For Rockburst At Great Depth

Rockburst is a difficult and hot issue in deep underground engineering in the world.In this paper,rockburst is the research object,and the existing methods and models can hardly meet the prediction requirements of current deep engineering safety excavation because of the unknown contribution mechanism of induced rockburst parameters and the difficulty of rockburst hazard prediction.In this paper,the research on the contribution of internal and external factors and disaster prediction methods for rockburst at great depth is carried out by means of literature research,theoretical analysis,and mathematical modeling,and the following results are obtained:(1)The risk of rockburst was simulated for the cavern form,engineering scale,stress level,rock mass quality,and explosion source characteristics parameters,and a corresponding prediction model was established.The influence of internal and external parameters on rockburst was analyzed by numerical simulation,and the association rules between internal and external parameters and rockburst were analyzed by data mining techniques.(2)The rockburst criterion index Rbi was proposed,and the integrity index,maximum tangential stress,uniaxial compressive strength,uniaxial tensile strength,and elastic energy index were selected as control factors.Finally,the engineering verification of 8 rockburst cases was carried out,and the highest accuracy of 87.5% was obtained.(3)An ensemble model for long-term rockburst hazard prediction based on the improved WOA algorithm was proposed,and the highest accuracy of 82.93% was achieved in the test.In the parametric contribution interpretation,the maximum tangential stress is the most influential feature in terms of integrated 4 intensity levels;for intense rockburst,the integrity index is the most influential feature.(4)An XGBoost short-term rockburst prediction hybrid model based on the BSA algorithm was proposed,and the model achieved the highest accuracy of 88.46% in the test.According to the parametric contribution interpretation,PPV is the largest contributing parameter for classes R3,R4,and R5,but the contribution of PPV is slightly inferior to GSSC for class R2.(5)To overcome the limitations of single model prediction,an ensemble framework for short-term rockburst was proposed based on RF,LGBM,XGBoost,SVM,and LR algorithms.The accuracy of the ensemble framework in the test was improved by about 3%,to 85.71%.The cumulative apparent volume is obtained as an important condition according to the parametric contribution interpretation.(6)To overcome the black-box property,a gene programming-based short-term rockburst hazard prediction formula was proposed to provide a more practical and convenient output.Intense and slight rockbursts were predicted with 100% accuracy in the overall performance test.100%accuracy was obtained in 7 real-case validations.The research in this paper can promote the improvement of rockburst criterias and prediction methods and provide theoretical guidance for the prevention and control of rockburst disaster at great depth.