Research On TBM Surrounding Rock Classification And Tunneling Performance Prediction Based On Artificial Intelligence

The tunneling efficiency of Tunnel Boring Machine(TBM)is determined by the selection of tunneling parameters.The core problem is the interaction between TBM and surrounding rock.TBM construction data is the most realistic reflection of the machine-rock interaction in the current construction geological environment,and the traditional statistical analysis method of construction data is not enough to analyze the big data generated during TBM construction.Based on the data of TBM construction projects such as a water supply project in Xinjiang and Zhuxi Reservoir,this thesis systematically analyzes the TBM-rock mapping relationship,summarizes the distribution law of tunneling parameter values,and establishes the identification method of surrounding rock drivability based on FPI and TPI.On this basis,the prediction model of tunneling parameters is established based on LSTM and CNN algorithms,which provides a theoretical basis for the safe,efficient and intelligent tunneling of TBM.The main work and achievements are as follows:(1)Based on a water diversion project in Xinjiang,Zhuxi Reservoir and other projects,the number,type and characteristics of data are collected and determined,and the data requirements and functional requirements of the database are analyzed.Aiming at the problems of multi-source heterogeneity of TBM construction and inconvenience of cross-table retrieval,the corresponding database is designed and the data is imported into the database to solve the integrated storage of structured and semi-structured construction data.On this basis,based on Python,Java and SQL programming languages,the front-end and back-end separated TBM construction database operating system is developed,which improves the utilization rate of data and provides a certain data basis for machine-rock correlation analysis and tunneling performance prediction.(2)Based on the TBM database operating system,the correlation between tunneling parameters under different surrounding rock conditions is studied.The TPI calculation method was established,and the FPI-UCS and TPI-UCS models were fitted.The model can effectively eliminate the influence of cutterhead diameter.The variation rules of FPI and TPI under different surrounding rock types are summarized,and the identification method of surrounding rock drivability based on FPI and TPI is constructed.The data verification of different sections shows that this method can realize the real-time identification of surrounding rock grade in the process of TBM tunneling in the relatively uniform geological section.(3)Using the memory mechanism of LSTM and the attention mechanism of CNN,the prediction model of TBM tunneling parameters under different surrounding rock types is established.MSE,RMSE and R~2 are used as evaluation indexes to evaluate the model as a whole.Subsequently,the absolute error and relative error were used as evaluation indexes to discuss the prediction effects of the early,middle and late stages of excavation.The prediction results show that the LSTM model has better prediction effect on stable data,and the CNN model has better prediction effect when the data fluctuates.The prediction accuracy in the middle of excavation is higher than that in the middle and later stages of excavation.