Computational Fluid Dynamics & Validation

CFD is not a method where a finer mesh eventually yields the right answer. What decides the result is the model standing in for what cannot be resolved directly, turbulence above all. So before a computed number can support a design decision, it has to come with a statement of how well it holds and under which conditions.

The longest-running subject is the wire-wrapped fuel bundle of a sodium-cooled fast reactor, where a wire spiralled around each pin stirs the coolant. We began by establishing what that wire actually does (Nuclear Engineering and Design, 2017), then solved the three-dimensional flow in a JAEA 127-pin bundle and, the same year, carried the RANS methodology up to the full-scale 217-pin bundle of KAERI's PGSFR. There followed where the vortices form (2019), how flow divides among subchannels (2020), what happens when one channel is blocked (2021), whether the cladding holds during design basis events (2024), and how sodium carries heat through a 37-pin bundle in the PLANDTL facility (2023).

The turbulence model is chosen to fit the question. RANS suffices when only the mean field matters, but not when the result is decided by structures that fluctuate in time. The lower plenum of the High Temperature Test Facility was solved with LES (2023); cross flow in a helical coil steam generator was solved with LES and then separated into structures by proper orthogonal decomposition (NTHAS-13, 2024); and the cross flow and vortices induced by a spacer grid were compared side by side with URANS and DES (2025). The solver, too, is chosen per problem: Ansys, OpenFOAM and STAR-CCM+ are all in use.

The last step is to put the result against measurement. Velocity fields of a 19-pin wire-wrapped bundle, measured under index matching, were compared directly with RANS (Annals of Nuclear Energy, 2020), and three-dimensional fields measured by magnetic resonance velocimetry served as validation data for helical cruciform fuel (NURETH-21, 2025). Analyses we publish go through a grid convergence index (GCI) check and uncertainty quantification as part of the procedure. That accumulated procedure was written up as a CFD methodology for next-generation reactor cores (Korean Society for Energy, 2025).