Abstract: The on-line real-time identification and early warning of TBM surrounding rock excavatability grade is significant for safe, high-efficient and intelligent TBM tunneling. An open-type TBM with a diameter of 7. 0 m is applied to EH tunnel in Xinjiang, and the practical boring data and geological data are analyzed. The surrounding rock excavatability is classified according to the characteristic parameter indicators reflecting the TBM tunneling performance and construction risk. Further, after analyzing the excavation parameters of better discrimination quality under different surrounding rocks with principal component analysis method, two principal component indicators for characterizing the excavatability grade of surrounding rock are obtained; and based on which, the BP neural network is constructed to identify the surrounding rock excavatability grade. Meanwhile, in order to increase the response speed of the model, a MATLAB program is designed to obtain real-time identification method of surrounding rock excavatability grade with better practicability.

Key words: TBM, rock excavatability grade, principal component analysis, BP neural network, real-time identification model