Choosing a turbulence model is not a preference but an engineering decision, and the reasoning should be written down [1] [2].
The decision path
Does the flow separate? If separation is expected — diffuser, volute tongue, blade wake, sudden expansion — k-ω SST is noticeably better. Under an adverse pressure gradient k-ε predicts separation late, or misses it.
Is the flow largely wall-parallel and attached? Long pipes, ducts, the attached region in external flow — k-ε is sufficient and converges more cheaply.
Rotating machinery? k-ω SST is the default; boundary layer behaviour on the blade surface drives the result directly.
Does transition matter? At low Reynolds numbers, where the laminar-turbulent transition region sets the drag, a transition model such as γ-Reθ is needed.
Its relationship with y+
The model choice dictates the mesh requirement:
- Using wall functions, stay in the range y+ ≈ 30–300
- Resolving to the wall (where SST earns its keep) requires y+ < 1
The right model with the wrong y+ is not the right model. Choosing SST and running at y+ = 50 throws away exactly the near-wall advantage you selected it for.
What belongs in the report
The chosen model, the reason for choosing it, the resulting y+ distribution (mean and maximum), and how wall treatment was handled. Three sentences will do — but those three sentences must be there.
References
- Menter, F. R. (1994). Two-equation eddy-viscosity turbulence models for engineering applications. AIAA Journal, 32(8), 1598–1605. DOI: 10.2514/3.12149
- Launder, B. E. & Spalding, D. B. (1974). The numerical computation of turbulent flows. Computer Methods in Applied Mechanics and Engineering, 3(2), 269–289. DOI: 10.1016/0045-7825(74)90029-2
- Langtry, R. B. & Menter, F. R. (2009). Correlation-based transition modeling for unstructured parallelized computational fluid dynamics codes. AIAA Journal, 47(12), 2894–2906. DOI: 10.2514/1.42362