NEA Benchmark Exercise: Computational Fluid Dynamic Prediction and Uncertainty Quantification of a GEMIX Mixing Layer Test

NEA/CSNI/R(2017)19
Adobe Acrobat PDF Document - on 8/5/22 at 5:31 PM
- Adobe Acrobat PDF Document on 1/15/20 at 5:31 PM
The main objectives of the current activity are the promotion, test and evaluation of various methodologies for uncertainty quantification in computational fluid dynamics (UQCFD) for nuclear reactor safety (NRS) applications. The assessment of the UQ methodologies was based on a blind test case from an experiment with a density gradient, which is of practical interest to NRS. The product is a synthesis report presenting a detailed assessment of numerical predictions – including the uncertainty bands – with measured data. This report was presented at the CFD4NRS-6 Workshop held in Boston in 2016.
 
The numerical benchmark exercise is restricted to single-phase flow, with turbulent mixing in the presence of density gradients, which is a typical situation encountered in many reactor issues, where computational fluid dynamics (CFD) is currently used. The exercise was based on the GEMIX experiment, which was carried out at the Paul Scherrer Institute (PSI) in Switzerland. Participants submitted a calculation for the blind test case, where they presented their predictions for mean velocity, turbulence kinetic energy and concentration profiles. All the results included uncertainty bands. Since some methodologies for UQCFD use data from a validation step (for the definition of the model, its calibration and/or extrapolation of errors), three open tests cases were provided to the participants. It is noteworthy to mention that for the blind test, the density ratio between the two mixing streams was 1%, which is much lower than the values encountered, for example, in pressured thermal shocks.