This study investigates the maximum dynamic load a geostructure, such as an embankment dam, can withstand under conditions of epistemic uncertainty. The development of “dynamic capacity functions” for such infrastructures has become technically and commercially viable due to advancements in numerical modeling techniques, enhanced hardware capabilities, and successful prior implementations. However, uncertainties in loading conditions and modeling assumptions are often neglected or addressed using empirical models, limiting their reliability. In this paper, a framework based on Intensifying Artificial Acceleration (IAA) is proposed to quantify the uncertainty in response quantities and estimate the failure capacity of a representative geostructure. The IAA methodology expedites the uncertainty quantification process by significantly reducing the computational demand associated with nonlinear transient simulations, offering a practical approach for engineering practitioners. This framework further generates failure fragility curves influenced exclusively by epistemic uncertainties, effectively decoupling them from aleatory uncertainties arising from ground motion record-to-record variability. Such a distinction facilitates the integration of the proposed framework with other studies that primarily address aleatory uncertainty such as performance-based earthquake engineering. Additionally, a series of sensitivity analyses are performed to evaluate the influence of material property variability and water level fluctuations on the response quantities. • A comprehensive framework is developed for quantifying material uncertainty in embankment dams. • Ground-motion randomness is explicitly separated from material-property uncertainty. • A novel intensifying artificial acceleration (IAA) procedure is introduced for failure analysis. • Sensitivity analysis is performed to identify the random variables that most strongly drive failure. • An innovative collapse fragility curve is constructed that reflects only epistemic uncertainty.
Hariri-Ardebili et al. (Fri,) studied this question.