Efficient Metamodeling-Based Sensitivity Analysis of Uncertain Material Parameters of ACCRDs Under Random Endurance Time Excitations

Fragility curves with variable damage probability are significant in performance-based earthquake engineering since they depict the seismic performance of dams over numerous damage states. To fully comprehend the structural performance under intense earthquake excitation, it is essential to account for uncertainties in the structural model. Considering that such uncertainties are challenging to estimate due to computational complexity, global sensitivity analysis (GSA) with Sobol indices is an essential method in seismic performance assessment for the purposes of factor prioritization and mechanism reduction. While GSA is a reliable technique, it entails significant processing demands and is time intensive, particularly when nonlinear finite element analysis needs to be conducted for each sample. This paper introduces an efficient framework that integrates traditional fragility analysis methods to address both aleatoric and epistemic concerns. The approach integrates the generalized F-discrepancy (GF) sampling method, a sparse grid-based polynomial chaos expansion metamodel, and the endurance time history method to effectively account for material uncertainties with a limited number of time history studies. The findings demonstrate the substantial computational advantages of the proposed framework over the MC-based method without sacrificing accuracy. Moreover, the critical state parameters and stress path ratio of the multiple yield surface model can significantly affect the seismic resistance of a dam under nonlinear conditions.