LAUSR

Abstract The dynamic parameters of rockfill are of great significance for the seismic safety analysis of high earth-rockfill dams. However, due to the size and time effects, the dynamic parameters obtained from the laboratory may not accurately reflect the actual properties of the dam. Therefore, it is necessary to analyse the dynamic response and deformation of the dam through inversion. In this paper, based on the combination of a viscoelastic artificial boundary, a radial basis function neural network (RBFNN) and a multi-population genetic algorithm (MPGA), an inversion method of the dynamic parameters of a rockfill dam considering the interaction effect is proposed. Additionally, to ensure the rationality of the ground motion input mechanism of a three-dimensional (3D) canyon, a transfer function method is adopted. Using the inversion method, the rockfill material dynamic parameters of the Nuozhadu earth core rockfill dam are identified. The acceleration values calculated by the identified parameters are consistent with the prototype monitoring data of the dam under small or weak earthquake conditions, indicating that the inversion method proposed in this paper is reasonable and accurate. The method is especially suitable for earth-rockfill dams with many weak earthquake records and few strong seismic records.