Efficient Seismic fragility curve estimation by Active Learning on Support Vector Machines

Rémi Sainct (1) , Cyril Feau (1) , Jean-Marc Martinez (2) , Josselin Garnier (1)

1 CEA-DES (ex-DEN) - CEA-Direction des Energies (ex-Direction de l'Energie Nucléaire)
2 LMA - Laboratoire de Mécanique et d'Acoustique [Marseille]

Fragility curves which express the failure probability of a structure, or critical components,as function of a loading intensity measure are nowadays widely used (i) inSeismic Probabilistic Risk Assessment studies, (ii) to evaluate impact of constructiondetails on the structural performance of installations under seismic excitationsor under other loading sources such as wind. To avoid the use of parametric modelssuch as lognormal model to estimate fragility curves from a reduced number ofnumerical calculations, a methodology based on Support Vector Machines coupledwith an active learning algorithm is proposed in this paper. In practice, input excitationis reduced to some relevant parameters and, given these parameters, SVMs areused for a binary classification of the structural responses relative to a limit thresholdof exceedance. Since the output is not only binary, this is a score, a probabilisticinterpretation of the output is exploited to estimate very efficiently fragility curvesas score functions or as functions of classical seismic intensity measures.

Keywords: Support Vector Machines; Fragility curve; Active Learning; seismic intensity measure indicator

Domains: Nuclear Experiment [nucl-ex]; Nuclear Theory [nucl-th]

Rémi Sainct, Cyril Feau, Jean-Marc Martinez, Josselin Garnier. Efficient Seismic fragility curve estimation by Active Learning on Support Vector Machines. 2018. ⟨cea-02339421⟩