An Application Research On Intelligent Back Analysis Of Tunnel Surrounding Rock Displacement In Xue Feng Mountion Highway Tunnel

With the durative, steady, fast development of economy, our country expand theinvestment to the basic establishment, so more and more tunnels have been built. Aswe all know, tunnels are built in the complicated ground mass. Ground mass behavesgreat randomness, fuzziness, uncertainty, non-integrality of information, so it isintractable to ascertain parameter of rock and soil exactly. On the bases of massiveresearch, back analysis is a good method to get parameter of rock. A intelligent BackAnalysis for tunnel surrounding rock displacement is proposed by combining the BPneural network and genetic algorithm, which is based on ANN (artificial neuralnetwork) and GA (Genetic Algorithm) toolbox of MATLAB. Then, it is applied toXue Feng Mountion highway tunnel in order to evaluate the practicality andfeasibility of the Displacement Intelligent Back Analysis.The displacement back analysis method is based on BP artificial neural networkwhich has powerful nonlinear functions mapping abilities, auto learning andgeneralization ability, the BP neural network is used to describe the relationshipbetween the parameters inversion and the surrounding rock displacements produceddue to excavation. It is a optimization back analysis method. The Parameter backanalysis problem is transformed into a function for the best answer problem. Thecalculation progress is similar to the forward analysis. Firstly giving the primitiveparameter to the ANN, we can get the calculated value. Comparing the calculatedvalue with the measured value, we modify the primitive parameter until the calculatedvalue approaching to the measured value, then we believe the modified parameter asthe true value.The dissertation make experiments with Orthogonal design to gain trainingsamples, make experiments with Uniform design to gain test samples. With parameterof every experiment, the simulation of excavation and support on tunnel was carriedout by Fast Lagrangian Analysis for Continuum to get the surrounding rock displacement. The parameters are used as the input, and the surrounding rockdisplacements are used as the output, so that we can gain training samples and testsamples of BP neural network.Based on traditional BP arithmetic, an enhanced BP arithmetic is obtained bycombining the additive momentum method with the self-adjusting learning speedmethod. Thus, the training speed can be fast. At the same time, the stability of thenetwork can be ensured. By introducing the parameter of adjusting learning speed, thetransfer process becomes more sensitive and the convergent speed becomes faster,thus increasing the calculating precision of the training function.In the course of Back analysis, Genetic Algorithm is used to search the optimumneural network structure, and the optimum neural network structure is trained by theenhanced BP arithmetic. At last, Back test is used to test the result. The result provesthat forecast capability of the optimum neural network is good enough Based on whathave done, the genetic algorithm associated with optimum neural network is adoptedagain to search the optimal parameters inversion in their global ranges.Displacement Intelligent Back Analysis is applied to the Xue Feng Mountionhighway Tunnel engineering, On account of the monitoring data of surrounding rockdisplacements, the paper interchanges the elastic modulus of surrounding rock and thecoefficient of initial stress of rock masses. The results are satisfactory and show thatthe intelligent network method for tunnel surrounding rock displacement backanalysis is accuracy, credible and operational in certain degree.