| Hidden fault identification is one of the main hotspots of current research.This article focuses on the study of TSP advanced geological forecasting to detect hidden faults and mechanical research,using theoretical analysis,data query,and computer simulation methods to summarize the TSP advanced geological forecasting to detect hidden faults.In the process,the water permeability of hidden faults under different mechanical properties is analyzed.Using the Keras deep learning framework for P-wave images of TSP concealed faults on the Lanxin Line,Rongwu Expressway,Xinghe Line and other lines,an automatic detection model of TSP concealed fault P-wave images was realized through training.The faults in section 900～800 of Guanjiao Tunnel 4# Inclined Shaft are simulated.The main work and results of this paper are as follows:(1)According to the stress of the concealed faults,the concealed faults are divided into tensile or tensile-torsional water-conducting faults,tensile or tensile-torsional water-proof faults,compressive and compressive-torsional water-conducting faults,compressive and compressive For torsional water-resistant faults,torsional water-conducting faults,and torsional water-resistant faults,the reliability of this classification is proved based on engineering practice.(2)Build a TSP hidden fault P-wave image library,which is the P-wave image of each tunnel measured by TSP in the past ten years.There are 3 880 pictures in the gallery,including 2 463 pictures of hidden faults and 1,417 pictures of no abnormal disasters,which have certain value for the follow-up study of TSP hidden faults.(3)Constructing a convolutional neural network.In this paper,a total of three models are carried out to identify the P-wave images of the TSP hidden fault.The three models are:(1)simple convolutional neural network,(2)migration learning based on VGG16,(3)fine-tuning it through the migration learning of VGG16,and adopting data enhancement technology.The test accuracy,loss rate and time of the three models are different,and the three models are used to recognize the P-wave image of the TSP of the Guanjiao tunnel.The results show that the recognition rate of the migration learning of fine-tuning the VGG16 can be better recognized The existence of hidden faults.(4)Relying on the concealed faults in section 900～800 of Guanjiao Tunnel 4#Inclined Shaft,three models of intersection angles between the fault and the tunnel of60°,45°,and 30° were established.By comparing with the results of TSP advanced geological prediction.It is found that the model 60° and 45° are similar to the forecast results,inferring that the intersection angle between the fault and the tunnel is between 60° and 45°.The following results are obtained through numerical simulation research:(1)For the initial stress and initial displacement,the intersection angle between the fault and the tunnel is the maximum of 30°.(2)After the tunnel excavation,the influence range of the stress and displacement of the vault and invert arch is the largest when the intersection angle between the fault and the tunnel is 60°,the smallest is 30°,and the position of the fault arch toe has the largest influence at45°.(3)In the excavation process,for stress,the maximum stress of the dome is at the angle of intersection 45°,the minimum is 60°,the maximum stress of the invert is 60°,and the minimum is 30°;for displacement,the maximum stress of the dome and invert When the displacement occurs at an angle of 30°,it is the smallest at 60°;for the convergence of the side wall,the fault has the greatest influence on the convergence value and range of the side wall at 30°,and the smallest at 60°. |
PDF Full Text Request