Abstract: As rockfill dams advance in height from the 200-meter to 300-meter class, a coordinated control of dam deformation becomes particularly important, and their deformation behaviors are commonly assessed using parametric inversion-based finite element stress-deformation analysis. To investigate the applicability of different parametric inversion strategies to high-core-wall rockfill dams, this paper presents a comparative study of the two strategies－parametric joint inversion and decoupled inversion. We use the neural network to construct an agent model for the analysis by FEM, and adopt a population-intelligent iterative optimization algorithm to calculate the inversion of static, rheological, and humidification model parameters. Compared with the parameter joint inversion that features with less manual intervention and high efficiency, the decoupled inversion reflects more reasonably the influence of the filling and impounding process on rockfill dam deformation through a staged and zoned inverting procedure. It produces better model parameters and dam body deformation calculations in good agreement with measurements, thus achieving better effects in revealing deformation evolution and its spatial distribution trends in a dam body during filling, impounding and long-term operation.

Key words: high core wall rockfill dam, monitoring datum, parameter inversion, intelligent optimization algorithm, decoupled inversion