I would appreciate some input on the problems I have been experiencing
with the SST model. I have generally been using the KE model but am
trying to improve on the heat transfer predictions by using a K-omega
model near the wall (i.e. SST). Unfortunately I have a separation point
for which the SST model cannot calculate a converged solution:
Model (see fig3 of attachment):
- Rotor-Stator gas turbine cooling chamber (1/60th segment modelled)
- Axisymmetric slot inlet in the stator at low radius
- Discrete hole outlet in rotor at high radius (represents entrance to
blade cooling passage)
- All walls adiabatic except rotor
Mesh:
- Tetrahedral mesh with inflated boundary regions (created in CAD2MESH)
- Y+ is generally 20ish to make use of wall functions
- The refined mesh in the area of interest has Y+ < 1
KE:
- The K-Epsilon model has given good results for fluid dynamics but
over-estimates heat transfer from the rotor (as expected)
Problem:
- Below the inlet are two secondary recirculating flows (fig 1 & 2 of
attachment). Between these regions, at the point highlighted the total
velocity of the flow relative to the surface is zero, i.ethe is flow
separating from the surface and the shear stress goes to zero. The SST
model will not converge(Momentum RMS = 5x10^4), and the maximum
residuals for all properties occur at this point.
Attempted:
- Using the KE results as an initial value file
- Refining the grid in the area
- "tef numerics option" = 1
Do you have any suggestions how I can get results using this model?
|
The problem you experience could very well be related to the fact that the SST model is a much more accurate model than the KE model. I assume that the purpose of changing to the SST model is that you would like a more accurate prediction of the boundary layer. CFX has another model which blends the KE and the KO model. It is the BSL model. You can not access this model from the GUI, but you can just replace SST with BSL in the CCL under turbulence model. For this problem, the BSL model works very well, and provides improved heat transfer predictions for the rotating disc. |
||
|