IRTG 2657 Research Research Projects
Model reduction for the construction of fragility curves in earthquake engineering

Model reduction for the construction of fragility curves in earthquake engineering

Team:  Alexandre Daby-Seesaram, David Néron, Amélie Fau, Michael Beer
Year:  2020

Fragility curves play a key role in the design of structures in seismic zones, the prediction of potential losses on existing structures or the rapid assessment of their condition after an earthquake (in order to decide on appropriate response measures) are at the heart of the nuclear industry's concerns. The construction of relevant numerical twins, fed by data and allowing to test in real time different plant configurations, is therefore a major issue. From a structural model, considered with its possible misunderstandings, the numerical obtaining of a brittleness curve supposes the realisation of a very large number of nonlinear simulations in order to calculate the response to all the seismic signals defining the hazard. Each of these simulations requires several weeks on a high-performance simulator, which is currently a real bottleneck. The approach proposed herein is based on model reduction techniques, which offer a huge potential to develop innovative tools for computing. The basic ingredient is the LATIN-PGD method, a strategy for solving nonlinear parametric problems based on model reduction by separation of variables. This technique allows considerable gains in terms of computational cost and storage and has been proven in many problems (including the simulation of fatigue at a very large number of cycles in structures with damage, which is typically the case in earthquake engineering). In this work, the strategy will be extended to build fragility curves. For that purpose, some new approaches will be introduced to take into account non-conventional parameters, such as the map of initial damage in the structure.


Doctoral Researcher: Alexandre Daby-Seesaram

Scientific Advisors: David Néron, Amélie Fau, Michael Beer, Dominik Schillinger