Choose a turbulence model
Pick from the sixteen models under Models, and know what the choice commits you to.
Runs on the free tier.
Select Models. Model under Turbulence lists, in this order:
Laminar, k-epsilon, realizable k-epsilon, RNG k-epsilon, k-omega,
k-omega SST, k-omega SST LM, GEKO, EBRSM, Spalart-Allmaras,
LRR (RSM), SSG (RSM), LES WALE, LES Smagorinsky, SA-DDES,
SST-IDDES.
Each one shows a hint when selected. Wall treatment sits directly below and
must agree with the mesh you built — it selects a coherent set of conditions
across nut, k, omega and alphat.
Practical selection
Section titled “Practical selection”- External aerodynamics, adverse pressure gradients, separation —
k-omega SST. This is the application’s default and the right first choice for most external flow. - Attached external aerodynamics, and you want it cheap and robust —
Spalart-Allmaras. One equation. - Transition matters —
k-omega SST LM. It needs a wall-resolved mesh and a believable inlet turbulence specification; without both it is worse than the model it is refining. - Low Reynolds number, genuinely laminar —
Laminar. Not a fallback for a case that will not converge. - You need the unsteady structures, not a mean —
LES WALEorLES Smagorinsky. Both require a Transient run and a fine mesh in the region of interest, so budget for it before choosing. - Detached or hybrid —
SA-DDES,SST-IDDES. Same transient requirement. - A tunable model spanning several behaviours —
GEKO. ESI v2606 only.
Anything not in that list is available and unopinionated: choose it because you have data that says so, not because it appeared next in the menu.
What changing it costs
Section titled “What changing it costs”Changing Model — or Analysis type — re-derives the field set. Conditions on fields that no longer exist are discarded, so revisit Boundary conditions afterwards. The Derived block on the same pane shows the resulting solver and field list.
Check it worked
Section titled “Check it worked”- Derived lists the fields you expect, and the solver has not changed under you.
- The Problems tab is clear — an incompatible wall treatment or a model that needs a transient run is reported there, before the run.
- Run the same case with a second model in a second simulation and use Comparison. If the answer moves, the result is model-dependent, and that is a fact about your problem you have to report rather than a fault to fix.
What the models are actually choosing between: Concepts.