CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   Langtry-Menter transition modell issue? (http://www.cfd-online.com/Forums/cfx/22484-langtry-menter-transition-modell-issue.html)

Chebeba April 22, 2006 02:34

Langtry-Menter transition modell issue?
 
Using CFX 10.0.1, SST turbulence model with Langtry/Menter correlation (Gamma Theta Model) at default settings on a NACA 63-012 profile at alfa = 4 deg, Re=6500000.

I get a very strange result where the transition onset is jumping around every second or third grid point. I have refined my mesh to be well inside every specification discussed in the Modelling Manual p103-111. Increasing the y axis mesh density does not make the problem go away, it only compresses it :(

The images below is from a mesh with:

y+ around 1-2

90 streamwise nodes

Normal expansion 1.2

30 layers in the boundary mesh

(for reference XFoil says transition is at x/C = 0.65 for this case)

Any ideas what causes this weirdness?

/C

http://img177.imageshack.us/img177/3...menter21hw.jpg

http://img83.imageshack.us/img83/885...menter16zm.jpg


kdrbrk June 8, 2010 11:18

as far as I know, NACA 63 series gives laminar flow to %30 of the chord (the 3 after the 6)
For %65 laminar flow, try NACA 66 or 67 series.

ghorrocks June 8, 2010 18:37

The behaviour described above is typical of turbulence transition models. Just like real turbulence transition points they move around and advance and retreat on the surface. In my experience you frequently have to run it 3D and transient even though the flow looks 2D and steady.

Will Anderson July 3, 2013 10:41

laminar flow using turbulent transition model?
 
Has anyone got the Transition-SST model to capture purely laminar flow, before the onset of transition? Particularly on perfectly round objects like cylinders or spheres?

I cite here a figure from Langtry's own thesis, in which the laminar region doesn't appear to have been accurately computed, encircled in blue:

https://lh3.googleusercontent.com/-v...ryCylinder.png

Langtry does write:

"It is possible that smaller time steps are required to properly
capture the drag coefficient.. However, what is more likely is that a hybrid RANS-LES approach such as Detached Eddy Simulation (DES) will be required to predict this case correctly."

Has anyone done that?

ghorrocks July 3, 2013 19:09

It has been a while since I did this sort of stuff but I do not recall a problem in the laminar region. I cannot cite and reference for this, just a vague recollection of ancient history.


All times are GMT -4. The time now is 18:10.