Research on TBM Tunneling Parameters Prediction and Stratum Recognition Based on Ascending Section Tunneling Data

Tunnel boring machine (TBM) has been widely used in deep buried and long tunnels in recent years because of its practicality in complex geological conditions. TBM will generate rich operational data during tunneling in different geological conditions. The aim of this study is to achieve intelligent prediction of tunnelling parameters and real-time sensing of rock information using ascending section data. Using a water diversion tunnel in Xinjiang as the research background, firstly, the typical TBM boring cycle data under each surrounding rock level were statistically analyzed. The time and rate of the rising phase to smooth running phase of the TBM under different surrounding rock grades were found to vary greatly. Therefore, in this study, Bi-LSTM real-time prediction model of boring parameters was established using the TBM rising section boring data, in which the R2 values of thrust and penetration prediction were 0.961 and 0.956. Then, based on the real-time prediction results of the Bi-LSTM model, a particle swarm optimization SVM surrounding rock grading model was established to predict the surrounding rock grade from the rising section data with a prediction accuracy of 93.2%. This study realizes the rapid identification of the surrounding rock grade, which provides guidance for the intelligent TBM tunneling construction.