Coolability of heat-releasing debris bed. Part 1: Sensitivity analysis and model calibration

Coolability of heat-releasing debris bed is an important issue in the severe accident analysis and management. Traditionally, theoretical studies of top or bottom-fed debris bed coolability have been focused on obtaining a “best estimate” value for the Dryout Heat Flux (DHF) as a function of debris bed parameters (mean particle diameter and porosity). However, an important question for safety analysis is the quantification of uncertainties inherent in the problem. In this paper, a one-dimensional coolability problem is considered, with the aim of analyzing the influence of aleatory uncertainties in input physical parameters and modeling (epistemic) uncertainties on the prediction of DHF. Global sensitivity analysis is applied to rank the aleatory and epistemic parameters according to their effects on DHF and average pressure drop. The most influential model parameters are then calibrated to achieve the best fit to experimental data available. On the one hand, we demonstrate that model calibration is instrumental in achieving considerable improvement of quantitative agreement between the experimental and simulation data. On the other hand, experience of model calibration also suggested that (i) optimization of model parameters with respect to available experimental data on DHF is an ill-posed problem, and (ii) model calibration with respect to one-dimensional pressure drop experiments does not automatically improve the prediction of DHF and in some cases can even worsen it. Based on these insights, one can speculate that further analytical and experimental efforts are necessary to establish a better consistency between model form and experimental data on pressure drop and DHF.

► Sensitivity analysis is carried out for the model and physical input parameters.
► Interphase drag has minor effect on the Dryout Heat Flux (DHF) in 1D configuration.
► Model calibration on pressure drop experiments fails to improve prediction of DHF.
► Calibrated classical model provides the best agreement with DHF data from 1D tests.
► Further validation of drag models requires data from 2D and 3D experiments on DHF.