An Efficient Computational Seismic Testbed of a Nuclear Power Plant Diesel Generator Building

The state-of-the-art in seismic hazard and risk assessment has evolved to encompass, among others, ground motion selection methods, clustered seismicity, and new fragility assessment approaches. In contrast, nuclear power plant safety assessment follows a more conservative path, adhering to established methods with limited scope for experimentation. Bridging the divide to test and potentially introduce novel concepts requires compatibility with current practice, while avoiding unnecessary modeling complexity. To address this, a low-order computational testbed is proposed within the EURATOM METIS project, to create a fast, open-source assessment platform. It focuses on the diesel generator building of the Zaporizhzhia VVER-1000 nuclear power plant and incorporates simplified models of the building, one non-structural component, and a bare-bones fault tree to propagate uncertainties efficiently for a scenario of loss of core cooling (LoCC). The impact of soil–structure interaction (SSI) is modeled using either equivalent-linear lumped-parameter assemblies with frequency-dependent properties or more advanced frequency- and intensity-dependent configurations. This streamlined setup enables researchers to explore advanced assessment methodologies and provides a practical benchmark for nuclear facilities. As part of the testbed evaluation, the effectiveness of two different ground motion intensity measures is assessed in conjunction with hazard-consistent versus randomly selected earthquake records for the site of interest. In all cases, incorporating SSI and hazard-consistent ground motions offers higher fidelity and more consistent risk estimates with lower conservatism. Coupling these with a more efficient intensity measure, such as average spectral acceleration, offers further improvements, to some extent ameliorating any lack of hazard consistency in record selection.