Uncertainty analysis of CFD benchmark case using optimal statistical estimator

Abstract

The increase of computational power along with continuous development of local mechanistic models opens the space for detailed Computational Fluid Dynamics (CFD) codes in nuclear reactor applications. Unlike the mature uncertainty evaluation of system codes, the uncertainty evaluation for CFD code calculations is still under development.The main purpose of this study was the uncertainty quantification of the NEPTUNE_CFD calculation of the Generic Mixing eXperiment (GEMIX) test. The GEMIX experimental facility considers a square channel where two streams of isokinetic water, with different temperatures and/or densities, mix in a turbulent regime. For uncertainty analysis Monte Carlo analysis using surrogate model as replacement of code calculations has been used. In this work for surrogate model the proven optimal statistical estimator (OSE) method has been used for the response surface generation of the NEPTUNE_CFD predictions.In the uncertainty analysis two uncertain input parameters were considered: inlet velocity profile and inlet turbulence intensity. The calculational matrix consisting of 30 NEPTUNE_CFD calculations fully covers the space with calculated points. This gave the basis to study response surfaces constructed from less than 30 points.The results of the response surface generation together with the uncertainty quantification are presented. The influence of the number of NEPTUNE_CFD calculated points on the response surface generation and the uncertainty is shown. The obtained uncertainty results suggest that OSE is very powerful for uncertainty evaluation of CFD calculations (especially when few input parameters are varied) due to its ability to model highly nonlinear functions.