Abstract: Featured with inferior geological adaptability, TBM (Full Face Rock Tunnel Boring Machine) is prone to cause geological disasters such as jamming and collapses when encountered with unfavorable geological conditions or poor surrounding rock quality, hence affecting construction progress and threatening personnel safety. Through TBM data preprocessing, the original data is first divided into complete driving segments, and the torque penetration index (TPI) is calculated. The quality of surrounding rock is then predicted before boring by using time series method and neural network, and the quality of surrounding rock is judged in the rising segment of boring based on the Gini impurity of TPI. Results demonstrate that TPI well reflects the geological conditions of surrounding rock. TPI can be accurately predicted by using time series method and neural network. The quality of surrounding rock can be well judged by the Gini impurity of TPI.

Key words: TBM, TPI, neural network, surrounding rock quality, Gini impurity