The OECD benchmark for uncertainty analysis in modelling for design, operation and safety analysis of LWRs (UAM) is currently at its second phase. This phase entails the determination of modeling uncertainties in separate neutronics, fuel and thermal-hydraulics calculations and serves as a starting point for Phase III, where uncertainties and sensitivities for coupled multi-physics simulations will be analyzed. To appropriately account for the relevant sources of uncertainties in coupled simulations, it is necessary to quantify the major sources of uncertainties in each of the different fields and on that basis to propagate these in the subsequent integral coupled simulations.
At PSI, the participation to all the components of Phase II was considered as a highly valuable exercise. In particular, the Phase II-3 that includes the thermal-hydraulics component of the benchmark was taken as an opportunity to initiate the development of a methodology for uncertainty quantification (UQ) and sensitivity analyses (SA) related to LWR sub-channel analysis. The latest version of the thermal-hydraulic (TH) sub-channel code COBRA-TF has been selected for this work. COBRA-TF is a two-fluid, three-field (continuous vapor, continuous liquid and entrained liquid droplets) TH sub-channel analysis code capable of simulating two-phase phenomena in light water reactor (LWR) cores.
Thus, in this study, a set of COBRA-TF models simulating the well-known OECD/NRC BFBT experiments is developed. To generate input samples and compute sensitivity related quantities for selected COBRA-TF outputs (void fractions), the freely available software SIMLAB has been used. Starting from a set of prior uncertain parameters suggested by the UAM benchmark organizers, a sensitivity analysis is conducted with focus on the radial void distributions predicted by the code. The uncertain parameters include the boundary conditions of the tests (pressure, inlet coolant temperature and mass flow, heating power) and selected parameters of the closure models of the code (mixing coefficient, interfacial drag and heat transfer terms).
Only a set of steady-state void distribution tests of BFBT is considered (assembly types 1 and 4) in order to reduce the number of phenomena involved and thus increase the quality of the SA analysis results. In particular, the Morris screening algorithm, designed for the elimination of the less influential parameters, and global sensitivity analysis methods (to rank the most influential parameters), available in SIMLAB, are evaluated and compared with results from a similar sensitivity study using the SUSA methodology.