| With the durative, steady, fast development of economy, our country expands the investment to the basic establishment, so more and more tunnels have been built. As we all know, tunnels are built in the complicated ground mass. Ground mass behaves great randomness, fuzziness, uncertainty, non-integrality of information, so it is intractable to ascertain parameter of rock and soil exactly. On the bases of massive research, back analysis is a good method to get parameter of rock. During the construction, the design can be timely amended according to mechanical parameters from back analysis. No knowing nonlinear -relationship between deformation and mechanical parameters, we can solve question of back analysis. But because of high nonlinear function, learning ability, associative memory, and strong capacity of error tolerance of artificial neural network, it is able to construct the mapping between mechanical parameters and displacements from practical samples.This paper's main contents and research results are:1. This paper takes the eastern passageway in Xiamen for example, introduces the engineering and hydrologic geological situation, and expounds the relationship of monitoring measurement and displacement back-analysis and measurement's effects in the tunnel construction. In this paper, some measuring projects such as ground settlement, stress monitoring, horizontal displacement and vault crown settlement are introduced, and uses professional statistical analytical software called SPSS to make a recursive analysis, thus finds the distributed regulation of the measured data of the rock and confirms the final deformation of surrounding rock based on the practical geological conditions. The value of vault crown settlement is 84.2996mm, and the value of horizontal displacement is 81.5969mm.2. The dissertation selects three-dimensional fast lagrangian analysis of continua which is a geotechnical software as platform to simulate the surrounding rock's deformation. By analyzing the cross-section ZK6+857's cloud picture of maximum principal stress, the minimum principal stress, vertical displacement and horizontal displacement, the simulation results equate with the fact essentially. Basis on the tunnel engineering's gelolgical conditions, we can determine the value of mechanical parameters and the range. There are elastic modulus E/GPa (0.1,0.105,0.11,0.115,0.12,0.125,0.13) , poisson ratioμ(0.42,0.43,0.44,0.45,0.46,0.47,0.48,0.49) , cohesion C/KPa (20,25,30,35,40,45,50,55), internal friction angleφ/°(45,47.5,50,52.5,55,57.5,60,62.5) .The training sample of BP neural network is constructed using the orthogonal experimental design to guarantee the equilibrium dispersibility and tidiness comparability of sample. Inputing the parameter combination, and outputting the displacements by simulating, the nonlinear mathematical model of parameters and displacements is constructed by trainding and learning technology in BP neural network. And then the distributed rule of deformation value is substituted into the nonlinear model constructed and is used to inverse the mechanical parameters of rock. There are elastic modulus 0.1121GPa, poisson ratio 0.4717, cohesion 25.61KPa, internal friction angle 56.64°separately.3. The paper simulates the intersion result again, and the displacement values of tunnel surrounding rock simulated are the value of vault crown settlement 86.77mm and the value of horizontal displacement 74.68mm. Compared the displacement amount of surrounding rock simulated with the measured displacement value, the result shows that the simulation value equates with the measured value and it attains to precision requirement of BP neural network. The error of vault crown settlement is 2.93% and horizontal displacement is -8.48%. And the paper selects the cross-section ZK6+872 and ZKK6+885 to verify the result of back analysis. The error of vault crown settlement in the cross-section ZK6+872 is 4.01% and horizontal displacement is 1.21%. And the error of vault crown settlement in the cross-section ZK6+885 is -5.46% and horizontal displacement is -5.80%. The comparison results show that the parameters by back analysis are accurate. Therefore, we consider that the surrounding rock's mechanical parameters inversed accord with the practical engineering geological condition and can be applied to practical engineering project and assist the construction decision of tunnel engineering.Trougth research, artificial neural network can solve the system's modeling and controlling problems which is a complicated system with uncertainties,serious non-linearity, time varying posteriority. The neural network has special non-linearity, non-convexity, non-locality, adaptability and hurge ability of calculation and information processing. It can learn the change of environment adaptively. The neural network can be applied to back-analysis research of tunnel rock's deformation monitor with its typical black-box mode. With every kind traditional monitoring analytical method applied maturely, the neural network's effects obtain the practical test. Therefore, it is easy for technical staffs to master and use back analysis, because they needn't learn profound elastic-plastic mechanics and finite elements theory. This theory of back analysis provides a new means and a good method for further research. |
PDF Full Text Request