Multi-scale uncertainty and sensitivity analysis of the TALL-3D experiment

• The ATHLET-CFX model of the TALL-3D facility behaves in a monotonic way regarding the propagation of the modeling uncertainty.
• The biggest variations are observed in the temperature behavior.
• A screening analysis identifies the most influential parameters.

Over the last decades, the increase of the computational power has made feasible the computer modeling of complex thermal-hydraulic phenomena. These complex models use physical models to account for specific thermal-hydraulic phenomena. Each physical model requires a set of model input data. For several reasons (e.g. measurement uncertainties for stationary and time-dependent values, cost of the measurement campaign), the input data for the physical models cannot always be determined with precision. This lack of accuracy can significantly impair the model results. The analysis of the influence of these input uncertainties is therefore a key step to understand the model behavior and possibly improve its accuracy.The TALL-3D facility, built by KTH in the scope of the THINS project, aims at investigating challenging phenomena in a facility filled with lead–bismuth eutectic (LBE) containing a pool. The experimental data will be used for the validation of the models developed by the project partners. Based on the coupling between ANSYS CFX (CFD) and ATHLET (system code) implemented by the GRS, TUM performed an uncertainty and sensitivity analysis on the model of the TALL-3D facility. This analysis investigates the uncertainty in the output which is due to the uncertainty on the input (uncertainty analysis) and assesses the influence of the uncertain parameters (sensitivity analysis).