Uncertainty analysis of PKL SBLOCA G7.1 test simulation using TRACE with Wilks and GAM surrogate methods

•Uncertainty analysis using order statistics and metamodelling.•Plackett-Burman design of experiments to select significant variables.•Uncertainty analysis of an experiment conducted at PKL facility. The Nuclear Energy Agency auspices simulation of experiments in different facilities under several p...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Nuclear engineering and design 2017-08, Vol.319, p.61-72
Hauptverfasser: Sanchez-Saez, F., Carlos, S., Villanueva, J.F., Sanchez, A.I., Martorell, S.
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:•Uncertainty analysis using order statistics and metamodelling.•Plackett-Burman design of experiments to select significant variables.•Uncertainty analysis of an experiment conducted at PKL facility. The Nuclear Energy Agency auspices simulation of experiments in different facilities under several programs. One on them consisted of performing a counterpart test between ROSA/LSTF and PKL facilities, with the main objective of determining the effectiveness of late accident management actions in a small break loss of coolant accident. The results obtained by TRACE code for PKL experiment SBLOCA G7.1 (a scaled model of Konvoi reactor) were in good agreement with the experiments. However, in the simulation process, uncertainty was not accounted. Uncertainty analysis, following the principles of Best Estimate Plus Uncertainty (BEPU) approach, must be performed to measure the effect of uncertainties on the evolution of safety variables of interest, such as the maximum of the Peak Cladding Temperature (PCTmax) in the experiment. In this paper we present a comparison between two uncertainty analysis techniques. The first technique is based on order statistics that makes use of Wilks’ formula. The second technique is based on a Generalized Additive Model (GAM) that substitutes the thermal-hydraulic code, without and with consideration of errors in adjusting the GAM model. The comparison of the uncertainty analysis results makes use of several performance metrics such as coverage, Coefficient of Variation and conservativeness. Based on the results of these metrics it can be concluded that the GAMPE (GAM Plus Error) provides the best performance, in particular, when using small sample size, i.e. n=59, 93. For larger sample sizes, i.e. n=124, 153, GAMPE and Wilks’ results presents similar performance.
ISSN:0029-5493
1872-759X
DOI:10.1016/j.nucengdes.2017.04.037