Problem with tetrahedral grid

[Mubashir Ali]

Hi all,
 
I am simulating a bluff body using tetrahedral mesh. No. of cells is around 1.84 million. I am facing two problems which are :
 
1)  i am getting strange values on the outer surface of the the computational domain (the blue triangular values in the picture)
 
2)  Getting very high value of drag co-efficient around 2.7  but i am expecting values around 1.4
 
My settings: 
K- omega SST model
Steady state
incompressible
velocity at inlet = 36 m/sec
gradient calculation by =  imrgra = 2
Solver :  upwind for velocity components
                Centered for K nd omega
 
Please see the attach pic of the outer surface of the mesh after reaching steady state                  

[Alexandre Douce]

Hi,
For a tetrahedral mesh, you should use SOLU for velocity, and keep Upwind for k and Omega.
Don't forget to choose the adapted gradient : Least squares with extended neighborhood.
If the monitoring points present oscillation with a relaxation parameters to 0.9 try 0.7.
 

[Mubashir Ali]

hi Mr Dounce,
 
I tried the settings u recommended but unfortunately nothing  changed .... still have the same problem while see velocit magnitude with strange value on the outer surface of the computational domain ....  may i send u the mesh file ... i m using version 1.3.3  .does tht hv some bug or is it i am doing something wrong ... ur help would be appriciated .. merci :D

[Yvan Fournier]

Hello,
Note that the value you visualize is a cell-centered value. With a tetrahedral mesh, some cell centers are closer to the boundary than others, so depending on the thickness of the boundary layer, what you see may be quite normal.
You may "smoothe" the visualization using ParaView's "cell data to point data" filter or EnSight's "ElementtoNode" calculator function.
If the result is still irregular, you may have an issue with the calculation. Otherwise, it is mainly a visualization issue.
In any case, having one or two prismatic layers (extruded triangles for example) at the boundary whould help both visualization and calculation quality, so if your friction coefficient is very far from what you would expect, improving the mesh (or at least doing some refinement for mesh sensitivity analysis) would help.
Best regards,
  Yvan

[Mubashir Ali]

Hi yvan,
 
Thnx for your reply.  your suggestion really worked for visualization worked.... i changed cell data to point data nd i see no problems. But for the friction co-efficient i still need to find out... i will update you about it...
btw what should be yplus values for k omega sst model .. does yplus value have effect on my results...
 
thnx in advance
Mubashir

[César Vecchio]

In theory SST is "independant" from Y+, it behaves like K-e when Y+>30 and simulates laminar sublayer with Y+<1. I read somewhere that SST shows flow detaching earlier than K-e, so that could explain something. I think, however, that a finer (specially around corners and vertices, eddies go mad in those areas :P) hexahedral mesh will yield better approximations like Yvan said before. Your geometry seems ideal like to apply a fully hexahedral mesh.