Abstract: For the safe disposal of high-level radioactive waste, China plans to establish an underground laboratory at buried depth of about 500 m in the granite rocks to carry out preliminary study on the disposal. However, as a common dynamic failure in deep rock engineering, rockburst always cause serious consequences. In the aim of guiding the selection of the underground laboratory site and the safe design and construction of the project, rockburst risks of shaft and tunnel excavation at different sites were predicted and analyzed based on support vector machine optimized by particle swarm optimization (PSO-SVM). One hundred groups of measured rockburst data as well as the geo-stress values and the mechanical parameters of rock mass of three candidate sites (Jiujing,Jijicao,and Xinchang) in Beishan pre-selected area were also taken as basis. Evaluation parameters including maximum tangential stress σ_θ of surrounding rock, uniaxial compressive strength σ_c,uniaxial tensile strengh σ_t,stress coefficient T_s, and brittleness coefficient B were chosen. Results show that PSO-SVM algorithm is feasible for rockburst prediction. The rockburst risk of engineering excavation in the depth of 300-600 m at Xinchang is the lowest. Therefore, selecting Xinchang as the construction site of underground laboratory for the disposal of high-level radioactive waste is the most secure.

Key words: disposal of high-level radioactive waste, PSO-SVM, rockburst prediction, Beishan pre-selected area, underground laboratory