Solutions to the Reynolds-averaged Navier Stokes equations using four basic classes of turbulence models are compared with experimental data for flow across a row of confined cylinders. The configuration is intended to mimic the lower-plenum model for a generic Next Generation Nuclear Plant. Turbulence models considered are the standard and realizable k - &egr; models, the standard and SST k - ω models, the ν² - f model, and a differential Reynolds-stress transport model. Comparisons between the computational results and experimental data are made for the streamwise pressure gradient, recirculation lengths behind cylinders, separation points from cylinder surfaces, velocity profiles, boundary layer profiles, and turbulence profiles. Calculations between 2D steady, 2D unsteady, and 3D unsteady formulations of the governing equations are compared.

No single turbulence model was sufficiently accurate in all areas considered. The ν² - f model provided the most accurate results for high Reynolds numbers, but misrepresented the transition point between laminar and turbulent flow. The k - ω class of models provided the most accurate results for low Reynolds numbers.

Results between the 2D steady, 2D unsteady, and 3D unsteady formulations were consistent in general. Considerations of computational requirements and accuracy of results showed that the 2D steady formulation was the most efficient for this simulation. The 3D unsteady formulation proved to be the least efficient and least accurate of the formulations considered.