Displacement Back-analysis Of Soil Parameters For Shui Bu Ya CFRD

For high concrete-faced rock-fill dam (CFRD), the exact mechanical parameters of the embankment materials should be given before computing the deformation and stabilization of the dam. Because the material properties are so complicated, including the effect of time and environment and the limitation of experiment technology, it is difficult for us to make them certain. The back-analysis method provides an effective way to solve this problem. This thesis does some research in the back-analysis to confirm the mechanical parameters of the dam materials, according to the deformation observation data of Shui Bu Ya rock-fill dam.After reading a variety of relative documents about back-analysis and comparing different methods, the thesis adopt the direct back-analysis method which combines the BP Neural Network and Genetic Algorithm to calculate the Tsinghua non-linear K-G model parameters of the rock-fill of Shui Bu Ya dam. BP Neural Networks, which establish the mapping between mechanical parameters and displacement of the materials instead of Finite Element Method, can improve the computation efficiency greatly in back-analysis; Genetic Algorithm can find out the optimization overcoming the local minimum; Tsinghua non-linear K-G model does an excellent performance in reflecting the constitutive relation of rock-fill materials.Based on Matlab, the BP Neural Network is established. The method of comprehensive test is adopted to obtain the training samples. The training parameters and the corresponding displacement constitute the training samples. After that, The BP Neural Network is trained to establish the mapping between the mechanical parameters and the displacement. Then, to prove the correction of the training result, some test samples are made to test the Network. When the BP Neural Network is completed, it is used to replace the Finite Element Method during displacement back-analysis. Finally, The Genetic Algorithm, associated by BP Neural Network, searches the optimum mechanical parameters of the rock-fill.The actual displacement observation data between December, 2005 and July, 2006 are collected and analyzed to make sure the credibility of data. And then, the Finite Element mesh is established to simulate the practical filling procedure and the materials distribution.Based on all above, the back-analysis method which combines the BP Neural Network and Genetic Algorithm is used to gain the mechanical parameters of the rock-fill materials used in Shui Bu Ya dam. Using the gained parameters, the dam's displacement is computed and the result is acceptable. The back-analysis parameters are also used to predict the displacement of the dam when it is completed.