Displacement Back-analysis Of Embankment Dam Based On Neural Network And Evolutionary Algorithm

The safety evaluation of high embankment dam is becoming more and more important today. Using the in-situ observation data to back-analyze the actual running state of dam is one of the main methods of safety evaluation. So it is of great significance to study the theories and methods of back-analysis.In this thesis, the back-analysis of soil parameters is investigated and a practical method of displacement back-analysis is proposed using the neural network and evolutionary algorithm. The neural network is used to substitute the time-consuming finite element analysis and the evolutionary algorithm that is proved efficient and convergent in global range is employed to optimize the neural network and back-analyze the soil parameters.The displacement back-analysis system DBA_EANN is developed in Visual C++. The proposed method is applied to the displacement back-analysis of the Maopingxi embankment dam in Three Gorge Project and the effect of generation number and sample size on the simulation ability of neural network is investigated.The properties of the gravel-mixed core clay of the Nuozhadu rockfill dam are studied using the conventional triaxial tests and the back-analysis of the field loading test. A method to describe the over-consolidated condition resulted from the in-situ roller compaction test is proposed by setting the initial value of loading function in the finite element analysis. The above work provides a basis for determining the parameters of the soil in the feasibility study of the Nuozhadu project.The displacement back-analysis of the Gongboxia faced rockfill dam under construction is carried out using the in-situ observation data. The deformation properties of its four main rockfill materials are discussed by comparing the result of the back-analysis and laboratory tests. The above analyses have verified the proposed displacement back-analysis method based on the neural network and evolutionary algorithm to be effective, reasonable and practical.