Classification Of Surrounding Rock And Intelligent Prediction Analysis Of Advance Rate Based On Real-time Monitoring Data Of Shield

In subway construction,tunnel boring machines(TBM/shield machines)are widely used due to their advantages such as high construction efficiency,safety and environmental protection during construction,safe and reliable construction quality,less interference with ground buildings,and no impact from the weather.However,shield machines are expensive and sensitive to geological conditions,Effective prediction of various tunneling parameters(advance rate,cutter head torque,and cutter head thrust)of subway shield machines under specific geological conditions is particularly important for the selection of construction methods,scheduling,and project budget.Previous studies have shown that using deep learning algorithms can achieve relatively ideal prediction results for shield tunneling parameters.Based on this,this article relies on the second phase project of Chengdu Metro Line 19,combines the measured data of shield tunneling and depth learning algorithms,studies the efficiency of shield tunneling,and conducts research on the prediction of shield tunneling speed and classification prediction of surrounding rock in shield tunneling based on depth learning.Through preprocessing and correlation analysis of shield tunneling data,a prediction model for tunneling speed is established.Based on geological data,the classification of surrounding rock in shield construction tunnels is studied,and a classification model of surrounding rock in shield construction tunnels is finally constructed.The classification of surrounding rock in tunnels is predicted using the tunneling parameters recorded by the shield.The main research results are as follows:(1)According to the periodic changes of various parameters in the circulation section,such as cutter head thrust F,advance rate V,cutter head rotation speed N,and cutter head torque T,the circulation section is divided into four stages:starting stage,transition stage,stable stage,and shutdown stage.Based on the measured data after the division of the shield circulation section,the cutter head rotation speed,cutter head torque,penetration,tunneling speed The variation rule of five excavation parameters such as cutter head thrust along the pile number in moderately weathered mudstone and sandstone provides a reference for evaluating construction efficiency and surrounding rock classification.(2)Surrounding rock classification based on HC method,the XGBoost model is used to classify the surrounding rock types of the selected sample tunnel sections,and the excavation parameters that affect the classification of shield tunnel surrounding rock are analyzed.A shield tunnel surrounding rock classification model is established,and the model is preliminarily tested.It provides important support for improving the performance prediction,risk prevention and control level of shield,and realizing shield informatization construction.(3)Based on the improved LSTM,GRU,and RNN depth network models of ResNet and Attention,a new prediction framework is proposed,which is used to predict the advanced rate of Chengdu subway tunnels.The prediction effects of RNN,LSTM,and GRU depth models,as well as the optimization model improved based on ResNet and the optimization model improved based on Attention,are compared and analyzed.(4)The established shield advanced rate prediction model was applied to the mudstone and sandstone formation section and compared with the field data collected by Chengdu Metro shield machine,the r-square of the model is above 0.90,the root-mean-square error is below 3.6,and the mean absolute error is below 1.3.The model has achieved good accuracy and stability,which can provide reference for the setting value of tunneling velocity.