Back Analysis On Material Parameters Of Rockfill Dam Using Adaptive Differential Evolution Algorithm And Its Application

The consequence of the failure of high rockfill dam is very serious,so it is very important to evaluate the structural behaviors of the dam(i.e.stress and deformation).The finite element analysis of stress and strain is a significant way to evaluate dam safety.In order to obtain the objective and reliable results of the stress and strain analysis of rockfill dam,reliable constitutive model parameters for calculating the behaviors of the dam are necessary.However,the actual dam material parameters differ from the design parameters obtained through the indoor triaxial test,as a result of the differences of dam materials and the uncertainty of the actual construction.So it is necessary to obtain more reliable parameters by back analysis of the actual deformation of rockfill dam in order to evaluate dam safety more accurately.Intelligent algorithms have been widely applied to the back analysis of dam mechanical parameters,such as genetic algorithm(GA)and particle swarm optimization(PSO).However,most of the algorithms have certain disadvantages,such as the sensibility of control parameters of the algorithm,low searching efficiency and premature convergence.Aiming at these problems,the self-adaptation differential evolution algorithm for back analysis of rockfill dam mechanical parameters was proposed.The main contents and achievements of the thesis are as follows:(1)The method of establishing three-dimensional finite element model(FEM),considering the actual construction loading process,were presented.Then,the orthogonal experiment table of inversion parameters was designed and inversion samples were obtained by FEM.Finally,a neural network model,representing for the relationship between mechanical parameters and settlements,was established to replace the time-consuming and complicated finite element analysis.(2)The adaptive differential evolution algorithm(ADE)for the back-analysis of mechanical rockfill dam parameters of was designed,and the parameters inversion,combining ADE and BP neural network model,was presented with MATLAB programming.Also,the differences between the proposed algorithm and other intelligent algorithms were compared and analyzed.The algorithm could adaptively adjust the control parameters of each individual in each generation,so as to reduce the error caused by the constant control parameters during the whole optimization process.Meanwhile,it could also ensure the diversity of searching range and avoid local convergence.(3)Taking the main dam section of concrete face rockfill dam of a pumped storage power station for example,an application study was carried out.Using the proposed methods,the mechanical dam parameters were obtained by back analysis.It was found that the inversion parameters could reflect the characteristics of the dam materials more objectively,for that the settlement calculated by the inversed mechanical parameters is closer to the measured value than that calculated by the design mechanical parameters.Meanwhile,it was also found that ADE algorithm had more advantages than the other two algorithms in the convergence and searching effectiveness after comparing ADE,PSO and GA.So the availability and reliability of the proposed method was verified.