A Study on the Spatial Sensitivity of Material Parameters for a Rockfill Dam Considering Multiple Sections and Material Zoning

Parameter sensitivity analysis is an important method for quantitatively measuring the contribution of uncertain input variables to structural response outputs, and is of great significance for identifying key parameters in geotechnical constitutive models and simplifying analytical models. To address the limitation that existing parameter sensitivity analyses for rockfill dams are mostly confined to a single section or a limited number of nodal displacements, making it difficult to capture the spatial distribution characteristics of sensitivity, this paper proposes a spatial sensitivity analysis method for rockfill dam material parameters that considers multiple sections and material zoning. The method constructs a surrogate model for structural analysis based on sPCE, employs OMP to reconstruct polynomial basis functions to improve the efficiency of high-dimensional analysis, and quantifies the global sensitivity of each parameter using Sobol indices. In the analysis, multiple transverse and longitudinal sections of the dam body are selected, and a systematic parameter sensitivity analysis is performed for all nodes. Simultaneously, the material zoning of the dam is further considered to reveal the spatial distribution patterns of material parameter sensitivity in each zone. Taking a large concrete face rockfill dam as a case study and the DuncanChang E-B model as the analytical object, the results demonstrate that the sPCE surrogate model maintains good simulation accuracy even with a relatively small sample size (RMSPE < 0.047, MAPE < 0.03, R > 0.88). The parameter sensitivity of displacements in different directions varies across sections, but overall, the parameters n, Kb, and φ0 exhibit high sensitivity. The high-sensitivity regions are mainly concentrated within the corresponding material zone and exhibit significant spatial distribution characteristics along both the transverse and longitudinal directions. The proposed method reveals the spatial patterns of parameter sensitivity, providing a new reference for material parameter inversion and structural reliability assessment of rockfill dams.