Post-BEMUSE Reflood Model input uncertainty methods (PREMIUM) Benchmark Phase II: identification of influential Parameters

The objective of the Post-BEMUSE Reflood Model Input Uncertainty Methods (PREMIUM) benchmark is to progress on the issue of the quantification of the uncertainty of the physical models in system thermal-hydraulic codes by considering a concrete case: the physical models involved in the prediction of core reflooding. The PREMIUM benchmark consists of five phases: - Phase I: mainly, definition of the different uncertainty methods; - Phase II: determination of the physical models influential in reflooding using the test 216 of the FEBA programme; - Phase III: quantification of the uncertainties of the parameters associated with the physical models identified as influential within Phase II, using FEBA/SEFLEX experimental results; - Phase IV: validation of the found uncertainties within Phase III by propagating them in the 2-D PERICLES reflood experiment; this phase will be performed blindly except for the coordinators; - Phase V: synthesis report. This report presents the results of Phase II. Phase II is dedicated to the identification of the uncertain code parameters associated with physical models used in the simulation of reflooding conditions. This identification is made on the basis of the Test 216 of the FEBA/SEFLEX programme according to the following steps: - identification of influential phenomena; - identification of the associated physical models and parameters, depending on the used code; - quantification of the variation range of identified input parameters through a series of sensitivity calculations. A procedure for the identification of potentially influential code input parameters has been set up in the Specifications of Phase II of PREMIUM benchmark. A set of quantitative criteria has been as well proposed for the identification of influential IP and their respective variation range. Thirteen participating organisations, using 8 different codes (7 system thermal-hydraulic codes and 1 sub-channel module of a system thermal-hydraulic code) submitted Phase II results. The base case calculations show spread in predicted cladding temperatures and quench front propagation that has been characterized. All the participants, except one, predict a too fast quench front progression. Besides, the cladding temperature time trends obtained by almost all the participants show oscillatory behaviour which may have numeric origins. Adopted criteria for identification of influential input parameters differ between the participants: some organisations used the set of criteria proposed in Specifications “as is”, some modified the quantitative thresholds proposed in Specifications, and others used their own methodologies. This fact was a partial reason for the different ranges of input parameter variation identified by participants, in addition to differences of the physical models adopted by the different codes. Therefore, such different variation ranges of IP and, correspondingly, such different variation ranges of cladding temperature and time of rewet, make rather difficult the task of meaningful and easy-comprehendible comparison of Phase II results. Out of a total of 72 input parameters, initially considered by all participants, only 6 were identified as influential by more than 4 participants that are: - bundle power; - wall heat transfer coefficient; - interphase friction coefficient; - interphase heat transfer coefficient; - heat transfer (enhancement) at the quench front; - droplet diameter. It should be noted that actual parameters considered in parameter “Heat transfer (enhancement) at the quench front” are code-specific and may have different influence on calculation results. Several participants discarded some identified influential parameters (e.g., droplet diameter) due to existing relation between this kind of parameters so-called “Input Coefficient Parameters” and more global parameters (e.g. interfacial friction coefficient and interphase heat transfer coefficient which use the droplet diameter) so-called “Input Global Parameters”. Some participants also discarded identified influential so-called “Input Basic Coefficients” (e.g. bundle power) since their uncertainty has not to be determined in the Phase III but will be provided by the coordinator from experimental data. The behaviour of the variation of the responses at the extremes of IP range of variation greatly depends on the type of input parameter and on the code used. Mainly, the following two different behaviours can be characterized: - For some parameters, like power, wall heat transfer and interphase heat transfer coefficients, a qualitative (but not quantitative) agreement between different codes is observed. - For other parameters, like interphase friction coefficient and droplet diameter, a contrary behaviour (i.e. in correspondence of one of the extreme of the IP range, the direction of change of the responses is different) between different codes and even between different selected models within the same code can be observed. This suggests that the effect of such parameters on the cladding temperatures is quite complex, probably because it involves a lot of physical models (e.g., via interphase friction and interphase heat transfer coefficients for the droplet diameter). It shall be noted that the analysis of differences between the reflood models of different codes is out of scope of the PREMIUM benchmark. Nevertheless, it is recommended to take into account the physical models/ input parameters found as influential by the other participants in order to select the influential input parameters for which uncertainties are to be quantified within the Phase III of PREMIUM. In particular, input parameters identified as influential by other participants using the same code should be considered.