Abstract: High core-wall rockfill dam is featured with many partitions of different materials with varied instantaneous and rheological parameters. The inversion of instantaneous and rheological parameters simultaneously is accompanied with ponderous calculation. To improve the accuracy of material parameters in calculation model and deformation prediction, a decoupling inversion of instantaneous and rheological parameters was conducted in consideration of the decoupling relationship between instantaneous and rheological parameters. With Pubugou high core-wall rockfill dam as an example, the sensitivity of parameters to the deformation of rockfill dam was examined,and highly sensitive parameters were selected for inversion. Furthermore, the decoupling inversion of instantaneous and rheological parameters was conducted by adopting modified genetic algorithm and radial basis function neural network (RBF) according to the monitoring data of deformation during dam construction, first impoundment, and second impoundment. The inversion result proved that the calculated settlements agreed well in general with the measured data in terms of values and change rules.

Key words: core-wall rockfill dam, instantaneous and rheological deformation, parametric inversion, decoupling, RBF, genetic algorithm, Pubugou