The paper introduced Centrifugal Simulative test of loess high embarkmen and analysis method of loess slope stabiciation and Sum up studied achievement of Plane Centrifuge modal test and three -dimensional Centrifuge model test,and put forward K-G model to react loess basic character and Some Suggests to improve losess high embankment design for the Sumnaty of high embankment thece dimensional nonlinear finite dlement analysis.
Therefore, this paper, based on model test, puts forward the effective measures to reduce and control the welding stress and concaving deformation from various aspects of welding specification, welding process, manufacture of compensating stress ring, slope type of jpint, welding direction and welding order.
On this basis,a linear programming method is used to estimate the model's parameter,and the optimal background value parameter λ and the optimal boundary improvement condition ε are calculated by the genetic algorithm so as to ensure the minimum estimation error under corresponding model test criterion.
Research by this paper indicates that both the background value parameter λ and the boundary condition of GM(1,1) model have an effect on model prediction accuracy, then presents a method of achieving an optimal background value parameterand λ and an optimal boundary improvement condition ε based on genetic algorithm, in order to minimize the average relative error criterion or to minimize the maximum relative error criterion, respectively. This method also ensures the smallest prediction error under corresponding model test criterion.