Uncertainty quantification

Share the news!

SESAME Newsletter #1 (March 2016) 

Access calendar of events

ALL ARTICLES

Access Newsletter here >


WP5UNCERTAINTY QUANTIFICATION

Throughout the modeling process of a facility, the model developer faces situations in which choices have to be made regarding the use of a specific model to account for a given physical phenomenon. A series of parameters may have to be determined for such models in order to best characterise the facility. Experimental values for these parameters are often only available in the case of separate effect test facilities. Otherwise, they have to be specified based on best engineering judgment. Furthermore, boundary and initial conditions are commonly subjected to measurement uncertainty. The effect of the so-called modeling uncertainty must be quantified, as it may significantly influence the accuracy of the simulations of the facility’s behaviour carried out with these models.

Uncertainty and sensitivity analysis offers a bundle of methods in order to quantify the variability of the simulation’s outcome resulting from the variability of the model’s inputs (uncertainty) and to identify the most important contributors to this variability (sensitivity). It can be applied whenever significant model modifications are made. Not only do these methods help the model developer check the consistency of his model, but they also help setting priorities for further model development and work on their assessment and improvement. The validation and assessment of the predictive capability of computer codes that make use of these models can be carried by comparing the codes’ results with available experimental data. If the experiments are well designed, the performance of specific physical models in the codes can be analysed, and the conclusions used to improve them, or to determine their contribution to the overall uncertainty in the codes´ simulation results.

In the context of the SESAME project, coupled system thermal hydraulics and CFD codes are to be qualified for application to LBE flows in pool type reactors. The data obtained from the transients performed at the TALL-3D facility will be used to validate the ATHLET-ANSYS CFX model of the facility.

Uncertainty and Sensitivity Analysis of the T06.01 transient (from natural to forced circulation with both heaters switched on at constant power) have been performed by running 153 coupled calculations. Twenty three model input parameters were included in the analysis. In the following figure, the pool outlet temperature during natural circulation is considered. The results of the uncertainty analysis consist in tolerance limits (the two vertical red lines) which delimit at least 95% of the output variability with a 95% confidence level. In short: the most likely results lie between the two red lines.

The experimental value can be compared with the simulation’s outcome of the ATHLET-ANSYS CFX model (the middle orange vertical line represents the experimental value whereas the other two lines account for the measurement uncertainty). From these results we can conclude that this temperature is correctly predicted by the coupled codes even though it tends to be over-estimated.

In the present case, the results from the sensitivity analysis also performed for this model advocate for a better calibration of the pressure losses at the pump, a parameter with a large influence, and for further investigations on the reasons why the temperature tends to be overestimated. Since it is not a result of the uncertainty in all of the 23 variables considered, it may be due to shortcomings in some physical or geometric/boundary conditions models used.