Transient LOFA computations for a VHTR using one-twelfth core flow models

Abstract A prismatic gas-cooled very high temperature reactor (VHTR) is being developed under the next generation nuclear program. One of the concerns for the reactor design is the effects of a loss of flow accident (LOFA) where the coolant circulators are lost for some reason, causing a loss of forced coolant flow through the core. In the previous studies, the natural circulation in the whole reactor vessel (RV) was obtained by segmentation strategies if the computational fluid dynamic (CFD) analysis with a sufficiently refined mesh was conducted, due to the limits of computer capability. The computational domains in the previous studies were segmented sections which were small flow region models, such as 1/12 sectors, or a combination of a few number of the 1/12 sector (ranging from 2 to 15) using geometric symmetry, for a full dome region. The present paper investigates the flow and heat transfer for a much larger flow region model, a 1/12 core model, using high performance computing. The computation meshes for 1/12 sector and 1/12 reactor core are of 7.8 M and ∼531 M, respectively. Over 85,000 and 35,000 iterations for steady and transient (100 s) calculations are required to achieve convergence, respectively. ∼0.1 min CPU time was required using 192 computer cores for the 1/12 sector model and ∼1.3 min CPU time using 768 cores in parallel for the 1/12 core model, for every iteration, using ALPS, Advanced Large-scale Parallel Superclusters. For the LOFA transient condition, this study employs both laminar flow and different turbulence models to characterize the phenomenon of natural convection. The result of Realizable k–ɛ turbulence model (RKE) model is the most conservative one in the heated blocks (channels) regions, but the result of laminar flow is the most conservative one in the plenum regions. The comparisons of the results from large and small models suggest that large flow model is needed to obtain accurate prediction.