how many cells with RANS

Tobi · September 14, 2012, 04:55

Hello all,

I have a new

question. I had a look at a few presentations on the 7th OFW and other stuff. I realized that the meshes they use with RANS are not fine at all. Alberto Cuoci (for an example) uses very small cell meshes with RANS modelling of his catalystic and flamelet solver. In work I see always big and fine meshes used with RANS and so I am a bit confused what is the best way.

I think there is no exact answer but most of the meshes that I see I think you can calculate all fluctuations couse of the fine mesh. In a case a student meshed a geometry of about (highxlenthxwidth) 20cm x 50 cm x 20 cm with round about 4.5 million cells.

Is there any litrature or something like that where I can find some informations. Couse I think that its not worth to build a mesh with 4.5 million cells if you get about the same resolution with 800.000 - 1.000.000 cells.

For any suggestions I ll be very thankful

Tobi

blacksquirrel · September 28, 2012, 09:08

The correct way is to estimate you discretization error with a mesh refinement study.
(e.g.: http://www.grc.nasa.gov/WWW/wind/val...spatconv.html; Roache grid refinement studies)

in short:
You do the same simulation on different meshes (coarse and fine) and compare the solution. Your grid is fine enough, when the difference to the result of a simultaion of a finer grid is less than your margin of error.

colinB · September 28, 2012, 09:40

Dear Tobi,

the number of cells in your mesh depend highly on what you are actually
calculating.
Unfortunately I cannot comment on the case from Alberto you mentioned
since I don't do such type of calculations, but for ship hull evaluation
I use the following parameters:

- check the resolution of the boundary layer (y+value e.g. with
checkMesh) this value should be below 300

- resolution of the free surface: over the expected wave hight there
should be a layer of at least 12 cells

- in the area where you are expecting the waves the cells cross section
should be less than a square meter

- finally use a smooth transition from a dense grid to a coarse grid.

in the other areas you might use a coarser grid ( for me sometimes cells
with a volume of several cubic meters
(domain size 500 x 200 x 55 [m x m x m])) and the total number of cells
is sometimes up to 6M cells when using a very dense mesh, but usually
much less (approx. a third of it).

So this is how I end up with my number of cells. But I guess for other
problems the required resolution might be different.

I hope I could contribute

regards Colin

PS: Grüße ins Ländle meine alte Heimat!

September 14, 2012, 04:55	how many cells with RANS	#1
Tobi Super Moderator Tobias Holzmann Join Date: Oct 2010 Location: Tussenhausen Posts: 2,708 Blog Entries: 6 Rep Power: 51	Hello all, I have a new question. I had a look at a few presentations on the 7th OFW and other stuff. I realized that the meshes they use with RANS are not fine at all. Alberto Cuoci (for an example) uses very small cell meshes with RANS modelling of his catalystic and flamelet solver. In work I see always big and fine meshes used with RANS and so I am a bit confused what is the best way. I think there is no exact answer but most of the meshes that I see I think you can calculate all fluctuations couse of the fine mesh. In a case a student meshed a geometry of about (highxlenthxwidth) 20cm x 50 cm x 20 cm with round about 4.5 million cells. Is there any litrature or something like that where I can find some informations. Couse I think that its not worth to build a mesh with 4.5 million cells if you get about the same resolution with 800.000 - 1.000.000 cells. For any suggestions I ll be very thankful Tobi

September 28, 2012, 09:08		#2
blacksquirrel Member Join Date: Jun 2011 Posts: 53 Rep Power: 14	The correct way is to estimate you discretization error with a mesh refinement study. (e.g.: http://www.grc.nasa.gov/WWW/wind/val...spatconv.html; Roache grid refinement studies) in short: You do the same simulation on different meshes (coarse and fine) and compare the solution. Your grid is fine enough, when the difference to the result of a simultaion of a finer grid is less than your margin of error. plucas likes this.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
[snappyHexMesh] No layers in a small gap	bobburnquist	OpenFOAM Meshing & Mesh Conversion	6	August 26, 2015 09:38
[Netgen] Import netgen mesh to OpenFOAM	hsieh	OpenFOAM Meshing & Mesh Conversion	32	September 13, 2011 05:50
[snappyHexMesh] snappyHexMesh won't work - zeros everywhere!	sc298	OpenFOAM Meshing & Mesh Conversion	2	March 27, 2011 21:11
[snappyHexMesh] snappyHexMesh aborting	Tobi	OpenFOAM Meshing & Mesh Conversion	0	November 10, 2010 03:23
physical boundary error!!	kris	Siemens	2	August 3, 2005 00:32

September 28, 2012, 09:40		#3
colinB Senior Member Join Date: Aug 2010 Location: Groningen, The Netherlands Posts: 216 Rep Power: 18	Dear Tobi, the number of cells in your mesh depend highly on what you are actually calculating. Unfortunately I cannot comment on the case from Alberto you mentioned since I don't do such type of calculations, but for ship hull evaluation I use the following parameters: - check the resolution of the boundary layer (y+value e.g. with checkMesh) this value should be below 300 - resolution of the free surface: over the expected wave hight there should be a layer of at least 12 cells - in the area where you are expecting the waves the cells cross section should be less than a square meter - finally use a smooth transition from a dense grid to a coarse grid. in the other areas you might use a coarser grid ( for me sometimes cells with a volume of several cubic meters (domain size 500 x 200 x 55 [m x m x m])) and the total number of cells is sometimes up to 6M cells when using a very dense mesh, but usually much less (approx. a third of it). So this is how I end up with my number of cells. But I guess for other problems the required resolution might be different. I hope I could contribute regards Colin PS: Grüße ins Ländle meine alte Heimat!