Inverse Analysis Of Tunnel Surrounding Rock Based On DE-BP Neural Network And Research Of Support Parameters

There are lots of tunnel used in many fields such as transportation,municipal,irrigation,mine,military and so on.The construction of the tunnel was accompanied by excavation of soil,because the soil is sensitive to the external environment,once the land subsidence and the rock deformation are very large,it will cause the ground settlement,cave collapse and the loss of human life and property.Generally speaking,the construction process and supporting parameter were determined according to the material parameters of the surrounding rock,but as is known to all,rock is a type of nonlinear,heterogeneity and discontinuous material,how to obtain the material parameters in a simple and efficient way becomes a difficult problem.With Dalian Metro Line 5 tunnel as the research object,this research proposed the DE-BP algorithm based on the theory of the differential evolution and BP neural network.The rock parameters were back analyzed according to the displacement monitoring data and then we used the parametersto study the influence of different excavation methods and supporting parameters of the tunnel.The achievement of this thesis are as follows:(1)Collecting and reorganizing the engineering geology,hydrological conditions,construction technology and construction monitoring,using the hyperbola function in regression analysis to fit the curve of the rock deformation changing with time,and predicting the final deformation of the surrounding rock.(2)Using the finite element calculation software Abaqus to establish the model of tunnel,analyzing the sensitivity of mechanical parameters involved in the Mohr-Coulomb constitutive model.Finally,selecting the elasticity modulus,cohesive force and internal friction angle as the back analysis parameters of this study.(3)Established BP and DE-BP neural network model respectively.Using orthogonal experiment and uniform division to establish learning samples and training network.Two models’ predictive ability was evaluated by using posterior-variance-test.It is found that,the error of DE-BP algorithm is smaller than BP algorithm,which shows a good predictive ability.(4)The mechanical parameters of surrounding rock are applied to establish the dynamic excavation tunnel model.The CRD construction process,anchor length and anchor space of the tunnel were studied from the perspective of rock deformation,anchor axial force and supporting stress.The safety of each working condition was investigated to provide guidance and suggestions for tunnel design and construction.