A multi-index evaluation method for rockburst proneness of deep underground rock openings with attribute recognition model and its application

Abstract

As the tunnel excavation and resource exploitation go deeper into the Earth, increasing number of rockburst occur all over the world. However, the understanding of rockburst prediction and rockburst proneness evaluation is still at its infancy. In this paper, to accurately and efficiently predict rockburst proneness, a multi-index evaluation method for rockburst proneness of deep underground rock openings is established, verified and applied. The multi-index evaluation method is put forward based on attribute recognition model and combined weighting method. Eight discriminate indexes are taken into account, i.e., in-situ stress, rock brittleness, elastic deformation energy, maximum storage elastic strain energy, rock integrity, groundwater condition, section design size and site construction status. Through analyzing a variety of practical engineering cases, the weight values of each index are evaluated by combined method of analytic hierarchy process and random forest algorithm. It is found that the groundwater condition, in-situ stress, rock integrity and elastic deformation energy are the most important evaluation indexes. The attribute recognition model of rockburst proneness is established with weight evaluation and the attribute measure function of each index, where the attribute measure function is developed by analyzing the rockburst proneness of each index and setting up the standards for risk classes. The accuracy and applicability of the proposed multi-index evaluation method are verified through analyzing rockburst cases of twenty engineering projects. It is then applied to the rockburst hazard evaluation of the Caoguoshan Tunnel. Results showed that the predictions for rockburst proneness are consistent with the actual rockburst occurrence. The findings in this paper could facilitate evaluating of rockburst proneness for deep underground rock opening and be of great significance for rockburst risk prediction.