关键词: 隧道工程, 岩爆, 影响因素, 判断准则, MATLAB-BP神经网络, 改进预测模型

Abstract: In view of the worldwide problem of rock burst prediction in the field of underground engineering, the relevant research results of many scholars on the influencing factors, prediction methods and criteria of rock burst were summarized and expounded firstly. Furthermore, 9 main predictive correlation indexes were introduced, including maximum principal stress σmax, tangential stress of rock mass σθ, uniaxial tensile strength σt, uniaxial compressive strength σc, point load strength of rock examples Is, σθ/σc，σc/σt, buried depth of tunnel h and impact tendency index Wet. Under the condition of collecting typical underground rock burst projects at home and abroad, based on in-situ stress monitoring data and guided by non-linear scientific theory, an improved BP neural network prediction model considering the above 9 key prediction indexes was established. The proposed model was validated by an engineering example. The research results show that: the improved algorithm model realizes the organic combination of the non-linear theory and the network analysis method, avoids the shortcomings of the ordinary algorithm, such as slow convergence speed of the network, easy to fall into the local minimum point. The proposed prediction model has obvious advantages over the complex structure. The prediction results of the improved algorithm model are in good agreement with the actual situation, which can provide guidance and reference for the rock burst prediction of tunnel and underground engineering.

Key words: tunnel engineering, rock burst, influencing factor, judgment criteria, MATLAB-BP neural network, improved prediction model