Back Analysis Of Rockfill Material Parameters Of Na Ba Reservoir Concrete Face Rockfill Dam

Concrete face rockfill dam(CFRD)is one of the most common and widely used dam types in hydraulic structures.CFRD has many advantages,such as making full use of local materials,being able to be constructed in stages,being convenient for mechanized construction and being less affected by local climatic conditions,etc.it has gradually become a dam type with strong competitiveness and development prospect.With the increasing of the number of face rockfill dams and the extending of the operation time of the face rockfill dams that has long-established,there is an urgent need for a method to analyze and predict the stress and deformation characteristics and trends of the face rockfill dams during the construction and operation of the face rockfill dams,and to guide the construction of the dams or evaluate the operation status of the dams.However,the material composition of CFRD is varied and the construction conditions are complicated.At present,it is difficult to accurately simulate the stress and deformation characteristics of rockfill by the commonly used constitutive model of rockfill,and it is impossible to measure the real material parameters of point rockfill through laboratory tests.Therefore,it is of great significance to use the monitoring data of face rockfill dam to back analysis the actual material parameters of rockfill,so as to correct the model parameters to obtain the accurate status of face rockfill dam.In this paper,taking the face rockfill dam of Na Ba reservoir as an example,based on the monitoring data of the dam,the parameters of Duncan-Chang EB model during the impoundment period of the dam are inversed by using the direct back analysis method of displacement combining BP neural network and genetic algorithm,so as to analyze the stress and strain of the face rockfill dam.Since Duncan-Chang EB model can only reflect the instantaneous deformation of rockfill,the creep component of rockfill is deducted by introducing panel rockfill dam deformation statistical model.Then the BP neural network is trained by using the parameters sample and the settlement calculated by finite element method,which can accurately reflect the mapping relationship between the parameters and the settlement value of rockfill dam.Then,the fitness value corresponding to each parameter sample is calculated according to the fitness function,and the genetic algorithm is used to optimize and output the optimal parameters.Finally,the output parameters are used to analyze the stress and strain of CFRD.Based on the parameters obtained from the inversion to calculate the settlement of the individual monitoring-station of comparative analysis with the actual monitoring data,found that based on the inversion parameters calculated settlement monitoring points are in good agreement with the actual monitoring of settlement,it shows that the parameters obtained from the inversion analysis are accurate.