Term-by-term tests of stress-transport turbulence models in a three-dimensional boundary layer

Current Reynolds-averaged turbulence models for use in three-dimensional flows have mostly been derived by straightforward generalization of two-dimensional models; the results are rather unsatisfactory. Reynolds-stress transport methods perform better than methods based on an (isotropic) eddy viscosity. One approach to the problem is a term-by-term check of Reynolds-stress transport models against experiments or simulation data in representative three-dimensional flows. In this paper a comparison of four models for the turbulent-transport ‘‘diffusion’’ terms, and four models for the pressure–strain ‘‘redistribution’’ terms, with measurements (including all six Reynolds stresses and all ten triple products) in the thin boundary layer on the flat floor of a duct with a 30° bend in the horizontal plane is presented. The conclusion is that the pressure–strain models perform remarkably well, and that although the triple-product models often perform badly, they represent rather small terms and therefore do not cause large errors in Reynolds-stress predictions. The implication is that inadequate modeling of the dissipation must be the main cause of errors in predictions of three-dimensional boundary layers.