Comparison of heat transfer models at the pebble, gas and reflector interface in the PBMR

Abstract

It is a great challenge in the design of the PBMR to accurately predict gas flow and heat transfer in the reactor. Understanding the heat transfer at the core-reflector interface in particular is a very important aspect as the reactivity of the control rods housed in the reflectors is highly temperature dependent. It is also very important because the core-reflector interface is on the critical path for heat removal during accident conditions. PBMR has developed an OECD/NEA coupled neutronic/thermal-hydraulic benchmark to aid in the understanding of the different modelling approaches currently employed at PBMR. A comparison of THERMIX-KONVEK and DIREKT results showed large temperature differences at the core-reflector interfaces. Further investigation showed that these differences are as a result of the numerical methods used i.e. Cell-Centred (CC) vs. Vertex-Centered (VC). The present study extended this comparison to Star-CD (CC) and Flownex (VC) which are also used to simulate the reactor at PBMR. An ID MATLAB program that mimics the CC and VC numerical methods was verified against Star-CD and Flownex. This program was then used to model an ID version of the OECD/NEA benchmark. Results were compared with DIREKT and THERMIX-KONVEK. Although the results compared well, there were significant errors at the core-reflector interfaces. The findings of this study were that different numerical methods will predict different temperatures, heat fluxes and (temperature-dependent) sink terms. It was also shown that in addition to the differences resulting from numerical methods, differences were seen between Star-CD and DIREKT and Flownex and THERMIX-KONVEK in the region of the core-reflector boundary. In general, for complicated simulations like that of the pebble bed, the numerical basis of software used to simulate the problem needs to be understood for the problem to be correctly modelled.