Abstract | There has been a lot of discussion on intensity measure optimality for conventional structures, touting the advantages of novel metrics of ground motion intensity to improve upon the efficiency and fidelity of seismic assessment. Yet, somehow this revolution of sorts has not transitioned to nuclear power plant assessments, which cling to the time-honored tradition of the peak ground acceleration (PGA). They do so for the simple reasons of stiffness and mass. That would be stiffness in the assets themselves, typically leading to periods of the order of 0.1 to 0.2sec for both the structures and their nested components, but also in the rigidity of regulations in an understandably ultra-cautious industry. Adding the mass (and cost) of engineering effort required to reassess already established fragilities for hundreds of standardized components, it is no wonder that there is too much inertia to allow moving away from PGA. Would it not be great if someone came along and offered a minimal-error approach for converting existing fragility curves from PGA to any intensity measure of choice? Interestingly, the response characteristics of nuclear power plants may actually favor an equivalent one/two-degree-of-freedom-model based procedure that allows disaggregating existing fragilities back into ground-motion-level constituents and reconstructing them anew with the desired intensity measure parameterization. There is little doubt that safeguarding the integrity of nuclear power plants would still require massive computations rather than rely on shortcuts, yet such an approach can give novel intensity measures a fighting chance to prove that they are worth the trouble for nuclear engineering.

Abstract | The safety of a nuclear power plant is influenced by both aleatory randomness and epistemic uncertainty, as well as the potential inter-component and intra-component correlations. Aleatory randomness arises from inherent variability in the data, while epistemic uncertainty stems from limitations, or incomplete knowledge in models or data. Component correlation refers to the extent to which the properties of various components within a Nuclear Power Plant (NPP) are interdependent and how they may co-vary within a single component (intra-component correlation) or among similar/identical ones (inter-component). Evaluating their effects to completion is a non-trivial operation that requires a full model of the power plant and its components, as well as the overall fault tree. When the goal is the evaluation of alternative approaches to safety assessment, one need not set the bar so high. In this, we offer a pared-down model, comprising simplified models of the reactor building and of one or more non-structural components, together with a simplified fault tree that leads to loss of core cooling capacity. As an example, three alternative cases of perfect, partial, and no correlation are employed to test common causes of failure. Uncertainty is propagated using a Monte Carlo simulation with either classic or progressive Latin hypercube sampling, using the simplified model as an efficient benchmark for NPP-compatible applications.

Abstract | Recent advances in Performance-Based Earthquake Engineering (PBEE) have followed their own revolutionary path over the past two decades. Focusing on the assessment of conventional buildings and infrastructure, several innovative proposals on improving fragility assessment have appeared over the years, focusing on record selection, improved intensity measures, as well as novel uncertainty propagation approaches. Whether these are also useful for nuclear powerplant assessment remains a question. Partially owing to the well-established practice and guidelines, introducing novelty in such procedures faces significant hurdles in terms of modeling, computational power, and complexity. In an effort to overcome said difficulties in the context of the METIS Euratom project, we propose a minimalistic computational testbed comprising simplified models of a building, one or more components, and a bare-bones fault tree
to tie them together and propagate uncertainties. The METIS simplified testbed is realized in open source, using Python and OpenSees to offer a fast assessment platform, whereby one can test any number of ideas in a setting that resembles a real case study. Among them are important questions, such as the influence of aftershocks, or clustered seismicity in general, new uncertainty propagation techniques, the use of hazard consistent record selection approaches in place of selection based on the uniform hazard spectrum, as well as the use of new intensity measures that can potentially reduce the aleatory uncertainty in the estimates of risk. Focusing on the latter, we present an example of application of the testbed and its results.