Construction Safety Analysis And Surrounding Rock Deformation Prediction Of High-speed Railway Tunnel

High-speed railway tunnel is gradually developing to the direction of long distance,large diameter,large buried depth,so the construction safety requirements are also improved.Rock mass is characterized by inhomogeneity,discreteness and high nonlinearity,which leads to complex deformation and failure characteristics of rock mass.The monitoring data of tunnel deformation reflects its current state and future trend,which can provide reliable information for developing the mechanics model of surrounding rock.It is also an important basis for tunnel deformation prediction and construction safety evaluation.In this paper,based on the monitoring data during the construction of Yangshan high-speed railway tunnel,the safety analysis of tunnel construction and prediction of tunnel deformation are carried out.The main contents of this thesis mainly include:(1)Parameter back analysis of surrounding rock of high-speed railway tunnel.The numerical simulation of the shallow buried section is carried out by using FALC3 D.Based on support vector regression algorithm optimized by genetic algorithm,a back analysis model of the surrounding rock mechanical parameters is established.The mechanical parameters of the surrounding rock then are calculated.The results show that the established back analysis model has high accuracy and can provide reliable mechanical parameters of surrounding rock for subsequent construction safety analysis.(2)Safety analysis of tunnel construction based on FLAC3 D.The numerical simulations of different tunnel construction methods and different construction step lengths are established by using FLAC3 D.The distribution characteristics of deformation,stress and plastic zone of surrounding rock are analyzed,which reveals the deformation mechanism and stress redistribution rule of surrounding rock under different excavation methods.Through comparative analysis,the excavation method of shallow buried section of Yangshan tunnel is optimized.(3)Data reconstruction of tunnel deformation based on multi-task learning.The monitoring data of tunnel deformation from adjacent monitoring sections is selected,and the least squares support vector regression algorithm based on multi-task learning is developed to reconstruct the missing deformation data.On this basis,the data reconstruction performance of three similar single-task algorithms is compared for comparison purposes,and the feasibility of the proposed multi-task reconstruction model is verified.The complete data is provided for the subsequent deformation prediction and analysis of surrounding rock.(4)Prediction of surrounding rock deformation based on combination model.According to the measured data of convergence deformation,the empirical prediction model based on Bayesian parameter optimization is adopted to obtain preliminary prediction results.On this basis,relevance vector machine algorithm is used to modify prediction results of the empirical model.Then a combined prediction model of convergence deformation is developed.Finally,the prediction accuracy of the four models is compared,and the validity and reliability of the proposed prediction model are verified.