Gerontati A., Vamvatsikos D. (2023). The effect of intensity measure selection and epistemic uncertainties on the estimated seismic performance for non-structural components of nuclear powerplants. Proceedings of the SECED 2023 Conference, Cambridge, UK.
Abstract | The seismic performance of a structure/component is influenced by aleatory randomness and epistemic uncertainty, but also by the intensity measure (IM) selected for the assessment. Aleatory randomness results from natural ground motion record variability, while epistemic uncertainty corresponds to modelling assumptions, parameter variability, omissions or simplifications. IM selection, though, depends on the analyst and the data available. Potential
candidate IMs are the peak ground acceleration (a nuclear industry standard), spectral acceleration at a fundamental period of the structure (the relative newcomer), and average spectral acceleration in the range of short periods (the novel option). Their performance in quantifying uncertainty for short-period nuclear powerplants is not given, nor is it necessarily obvious given the sizeable uncertainties involved. To provide a basis for discussion, a singledegree-of-freedom non-structural component in an AP1000 reactor building is used as casestudy. Three alternative uncertainty propagation approaches are employed: (a) Monte Carlo simulation with classic Latin hypercube sampling, (b) Monte Carlo simulation with progressive Latin hypercube sampling and (c) a first-order second-moment method, representing different compromises between speed and accuracy. The resulting fragility curves of the non-structural component are compared in terms of efficiency for assessing its performance, offering evidence in support of optimizing IM selection.
[paper]