About mesh-independent solutions

86lolo · June 27, 2014, 18:13

Hello folks
Firstly, I apologize if this should go somewhere else, but I couldn't find a specific forum to post this.

I am about to conduct a CFD study on the Strouhal-Reynolds dependence for arrays of tubes.
I want to be thorough with this study, and I will perform a mesh-independence study of it. This means:
1) I will solve the problem with an initial mesh and I will obtain a solution
2) I will solve the same problem again with a finer mesh and compare
3) If the solutions were significantly different, I will proceed with an even finer mesh
4) Keep doing so until the solution does not change anymore (within some tolerance)

Now that I am going to start I've got some doubts about how much I should refine the mesh in each step to obtain meaningful changes.
For instance, let's define N as a parameter which may be, for instance, the number of faces in a given contour of the domain. Thus, the number of cells of the mesh will grow (in 2D) proportionally to N^2.
My question is: How much should I refine the mesh for each following step? Should I take 2*N (thus the number of cells will be multiplied by 4), 1.5*N,...? Or is it too much?
I'd like to make moderate steps in order not to waste too much computing resources (it is transient problem, the computing time grows waaay faster than N^2), but if the mesh is not much finer than the previous one, then the solution I will obtain will not differ from the previous one either... and then it might look like the problem got grid-independent, but actually it didn't. Therefore, smaller refinement ratios would require tougher tolerances.

So my question is... what tolerances/refinement ratios would you guys reccomend? Is there any relationship between refinement ratio and tolerances? What is your experience about this?

Thanks for answers!

tas38 · June 29, 2014, 04:58

The Journal of Fluids Engineering Editorial policy gives some practical guidance on this issue. It is recommended that when comparing the results of two or more grids, the grid refinement ratio should be 1.3 or greater where the grid refinement ratio is defined as the coarse grid spatial size divided by the fine grid spatial size.

86lolo · June 29, 2014, 05:08

Thank you very much Tas!
Very much appreciate this.

June 27, 2014, 18:13	About mesh-independent solutions	#1
86lolo New Member Manuel Join Date: Jan 2014 Posts: 18 Rep Power: 12	Hello folks Firstly, I apologize if this should go somewhere else, but I couldn't find a specific forum to post this. I am about to conduct a CFD study on the Strouhal-Reynolds dependence for arrays of tubes. I want to be thorough with this study, and I will perform a mesh-independence study of it. This means: 1) I will solve the problem with an initial mesh and I will obtain a solution 2) I will solve the same problem again with a finer mesh and compare 3) If the solutions were significantly different, I will proceed with an even finer mesh 4) Keep doing so until the solution does not change anymore (within some tolerance) Now that I am going to start I've got some doubts about how much I should refine the mesh in each step to obtain meaningful changes. For instance, let's define N as a parameter which may be, for instance, the number of faces in a given contour of the domain. Thus, the number of cells of the mesh will grow (in 2D) proportionally to N^2. My question is: How much should I refine the mesh for each following step? Should I take 2N (thus the number of cells will be multiplied by 4), 1.5N,...? Or is it too much? I'd like to make moderate steps in order not to waste too much computing resources (it is transient problem, the computing time grows waaay faster than N^2), but if the mesh is not much finer than the previous one, then the solution I will obtain will not differ from the previous one either... and then it might look like the problem got grid-independent, but actually it didn't. Therefore, smaller refinement ratios would require tougher tolerances. So my question is... what tolerances/refinement ratios would you guys reccomend? Is there any relationship between refinement ratio and tolerances? What is your experience about this? Thanks for answers!

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
[ICEM] surface mesh merging problem	everest	ANSYS Meshing & Geometry	44	April 14, 2016 06:41
[ICEM] Hexa mesh, curve mesh setup, bunching law	Anorky	ANSYS Meshing & Geometry	4	November 12, 2014 00:27
[snappyHexMesh] snappyHexMesh won't work - zeros everywhere!	sc298	OpenFOAM Meshing & Mesh Conversion	2	March 27, 2011 21:11
Mesh motion independent mesh regions	philippose	OpenFOAM Running, Solving & CFD	12	August 5, 2008 16:16
fluent add additional zones for the mesh file	SSL	FLUENT	2	January 26, 2008 11:55

June 29, 2014, 04:58		#2
tas38 Senior Member Troy Snyder Join Date: Jul 2009 Location: Akron, OH Posts: 219 Rep Power: 18	The Journal of Fluids Engineering Editorial policy gives some practical guidance on this issue. It is recommended that when comparing the results of two or more grids, the grid refinement ratio should be 1.3 or greater where the grid refinement ratio is defined as the coarse grid spatial size divided by the fine grid spatial size.

June 29, 2014, 05:08		#3
86lolo New Member Manuel Join Date: Jan 2014 Posts: 18 Rep Power: 12	Thank you very much Tas! Very much appreciate this.