How to make best use of numerical simulation, experience feedback and expert judgement in seismic fragility analysis for nuclear installations

Abstract

In nuclear engineering practice, the so-called safety factor (or separation of variables) approach is generally used to develop fragility curves due to its systematic applicability and the possibility to deal with a large number of SSCs (Structure, Systems and Components). With increasing computational capabilities, it becomes now feasible and more and more common to develop numerical models representing complex and possibly nonlinear behavior for components at stake. This talk addresses different aspects related to the numerical evaluation of fragility curves, including the choice of intensity measures, uncertainty propagation, reliability of numerical models, possible surrogates and the introduction of knowledge through expert judgement and in-situ experience data. Different sources of information such as expert judgement, numerical simulation, qualification tests and experience feedback can be combined in a Bayesian framework to develop best-informed fragility curves. Here, we present an approach that allows for the consideration of generic fragility parameters and simulation to develop priors and update fragility curves using experience feedback considering both epistemic and aleatory uncertainty. In particular, we use a database that contains failure data collected in industrial plants that have experienced an earthquake. We discuss opportunities and difficulties of this approach, related to the lack of specific data for nuclear equipment despite growing experience feedback and awareness. The PGA is generally used as intensity measure when developing fragility and hazard curves. Eventually, we consider an approach to deal with vector hazard and vector fragility curves and discuss possible benefit for seismic risk assessment of nuclear plants. Indeed, while PGA proves to be a very good damage or failure indicator for a large number of SSCs, there are a few with low frequency behavior that could be better characterized by introducing a second intensity measure such as low frequency spectral acceleration.

Biography

Dr Irmela Zentner is a civil engineer specialized in structural dynamics and earthquake engineering with a focus on probabilistic assessments. She is currently research engineer at EDF R&D Lab Paris-Saclay where she acts as an expert for probabilistic seismic risk assessment. She holds a PhD carried out at the French Aerospace Lab in the area of aeroelasticity and stochastic dynamics, after working in consulting and receiving master degrees at Ecole Centrale Paris and RWTH Aachen. She is member of IMSIA laboratory and SEISM institute Paris-Saclay. She is currently coordinator of the H2020 funded project EURATOM METIS and has been involved in several national and international research projects such as the recent SIGMA-2 and ANR EXAMIN projects.