Research On The Reliability Of Tunnel Structure Based On FEM & ANN

With the large scale development of highway construction in China, the design standard and tunnel construction in Southwest China especially in Chongqing is becoming higher. Owing to the complexity and uncertainty of tunnel, the current design method and evaluation of tunnel construction could not meet the need. It's necessary to evaluate and predict the reliability of tunnel support structure with a feasible method.Referring to research on the stability and reliability of tunnel and underground space structure(50334060), a key project of National Nature Science Foundation of China, proceeding from the uncertainty and fuzziness of tunnel and underground engineering, and also based on the comprehensive evaluation of varieties of structures reliability calculation methods and their limitations in the applying reliability of tunnel support structure evaluation and the probability distribution features of rock mass mechanical parameter, this paper used Shilong tunnel between Shuijiang and Jieshi in Yu-Xiang Highway as engineering background and simulated the surrounding rock and supports mechanical features during construction. Applying Monte-Carlo finite element method based on neural network, this paper analyzed the reliability of initial rock mass and support structure.The main conclusions of the paper are showed as following:(1)The main methods of reliability index calculation and the merits were discussed systematically. Shortcomings and limitations of these methods in tunnel structure reliability were analyzed. The paper expounded the principle of neural networks and its application in rock engineering synthetically. The BP neural networks program written by ourselves was used in tunnel structure reliability analysis. Neural networks, Monte-Carlo method and finite element method were integrated commendably.(2) Combined with the mechanical parameters and probability distribution features gained from Shilong Tunnel exploration survey and test., this paper selected the following random variables E 1,μ,γ, C 1,φ1 and E 2 as input. An obliquitous experiment was carried out. The giant general finite program-ANSYS was used to simulate and analyze the mechanical behavior of tunnel shallow buried section to gain the mechanical and displacement features of rock mass and initial support structure.(3) This paper applied finite element program results as training sample for BP neural networks, net training practice and building correct net link. Using Monte-Carlo...