Back Analysis Of Maopingxi Asphaltic Concrete Core Rockfill Embankment Dam

High asphaltic concrete core dam is one of the new developing directions in dam construction of the world. It is important to study the parameters back analysis. Combined with the behavior in 1st constructing period of Maopingxi dam, nonlinear FEM calculation and material parameter back analysis are applied to the Maopingxi asphaltic concrete core dam. The main contents are as follows:1) 3σ eliminate error criterion method and step wise regression analysis method is used to analyze the errors of observation date. In the stepwise regression analysis, aging, loading and temperature which bulkily affects the dam's vertical deformation are selected as the regression factors, then regression equation and curve of each point are worked out.2) Duncan E-u model is applied to the material of the dam. The feasibility of using Duncan model is discussed. The Duncan E~ u model which considers the creep character of asphaltic concrete is adopted. Mid-point incremental method and initial stress method are used to reflect the nonlinear of the material's stress and strain. At the same time, the process of load during construction is simulated. The incremental method is used to analyze the observation data. The calculated values and the observed values are approximate. It reveals that the incremental method can reflect the dam's behavior law.3) The six parameters of the asphaltic concrete core dam is inversed by least square , Bayes random and BP neural network method. The sensitivity analysis of the parameters is studied.4) The precision of the three inversion method are compared. In the Bayes back analysis, The parameters' statistical character is discussed. The conclusion is that the back analysis results have the character of unbiasedness, efficiency and consistency. Bayes random method which considers the prior errors is better in the back analysis of embankment dam's nonlinear materials.5) The nonlinear FEM analysis is carried on the dam. In this process, Bayes inversion values are selected. The FEM results are consistent with the observation data.