CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   The problem with GAMBIT - using two mesh sizes. (http://www.cfd-online.com/Forums/fluent/37011-problem-gambit-using-two-mesh-sizes.html)

Owen June 21, 2005 05:55

The problem with GAMBIT - using two mesh sizes.
 
Hi CFDers,

My question concerns meshing a geometry with two different sized meshes.

My 3D problem is a U shaped pipe (0.5m diameter) with several small injection nozzles (10mm diameter). In GAMBIT I have tried to save on mesh cell count by creating a fine mesh inside the small nozzles and retaining a coarse mesh through the main pipe. Separating the regions using a connected split operation. The small mesh size is 10mm while the large mesh is 100mm, total cell count 90k.

In Fluent the problem will converge poorly for a basic flow problem (vel-Inlet, pressure-Oulet, single phase). I would like to improve convergence of the problem to <10-2 residuals. Any ideas?

Best regards,

Owen

Mehdi Ghoddosy June 21, 2005 09:49

Re: The problem with GAMBIT - using two mesh sizes
 
Dear Owen:

I suggest to Use non-conformal meshing. first create the U shape pipe and nozzels separately in mesh files. then merge them using tmerge utility. then assign interface 1 to the nozzel interfaces in interface creation in FLUENT.this would improve your convergence.

Also i suggest to use two step for fining your mesh. The ratio of ten seems to be large. Use 3-5 as maximum.

Regards

Mehdi Ghoddosy

Owen June 22, 2005 05:16

Re: The problem with GAMBIT - using two mesh sizes
 
Thank you for your help Mehdi. Until you suggested it I had never heard of the T-merge function.

Is there any advantage to creating separate mesh files and merging them in Fluent (with T-merge) over simply creating the two volumes in GAMBIT and meshing them with different size meshes? I ask this because I find GAMBIT easier to use since it offers more flexibility to manipulate the mesh interface and check skewness.

I have taken your advice to increase the mesh size in two steps - this appeared to reduce the number of highly skewed elements at the interface.


All times are GMT -4. The time now is 15:28.