Tools for global and target sensitivity analyses in the context of high-dimensional thermal-hydraulic numerical experiments - Archive ouverte HAL Accéder directement au contenu
Pré-Publication, Document De Travail Année : 2020

Tools for global and target sensitivity analyses in the context of high-dimensional thermal-hydraulic numerical experiments

Résumé

In nuclear engineering, the quantification of the input uncertainties and their propagation through high-fidelity computer models in order to estimate relevant statistical quantities of interest constitute the core phases of the so-called "best-estimate plus uncertainty" (BEPU) methodology. However, several methodological and numerical issues can make the BEPU methodology difficult to execute in practice, and in particular, the large number (e.g., up to one hundred in our case) of uncertain input variables. Recently, an advanced BEPU methodology, called ICSCREAM, has been proposed for the identification of penalizing configurations in high-dimensional framework. The present work aims to detail the first step of the ISCREAM method which consists in identifying and ranking the primary influential inputs. For this purpose, global and target sensitivity analyses based on dependence measures are performed. Their principles and estimation techniques are explained, as well as their pragmatic implementation. Their use is then illustrated on an industrial use case simulating an intermediate-break loss-of-coolant accident via the CATHARE2 simulator.
Fichier principal
Vignette du fichier
preprint_GSA_TSA_HSIC_bepu2020.pdf (214.46 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-02877385 , version 1 (22-06-2020)

Identifiants

  • HAL Id : hal-02877385 , version 1

Citer

Vincent Chabridon, A. Marrel, B. Iooss. Tools for global and target sensitivity analyses in the context of high-dimensional thermal-hydraulic numerical experiments. 2020. ⟨hal-02877385⟩
160 Consultations
134 Téléchargements

Partager

Gmail Facebook X LinkedIn More