too dense mesh deteriorates the simulation

kkpal · June 29, 2015, 20:38

hi,
I'm using pimpleFoam to simulate the flow around a circular cylinder. I found that many times that if a dense mesh is used for a low Reynolds number simulation, say Re=100, the simulation is not physically correct(in terms of the drag or lift force coefficients). However, if I maintain the same mesh and raise the reynolds number to like 1000, the simulation seems to go well.

does someone else also have the same experience with me?

Blanco · August 5, 2015, 03:44

Hi kkpal, this is very different to the normal results improvement due to mesh elemnt size decrease and it is not aligned to numerical theory...have you checked everything in your model? are you using wall resolved boundary layers in particular?

fabian_roesler · August 5, 2015, 07:57

Right, when using lowRe Models without wall functions, the refinement of the wall depends on the Reynolds number. You have to achieve a y+=1 for the first layer. When changing the Reynolds number from 100 to 1000, the first layer thickness to achieve y+=1 varies with round about factor 5.

Cheers

Fabian

kkpal · August 5, 2015, 12:19

Thank you, Andrea and Fabian,
I am not using any wallfunctions. Just pure DNS for low Re flows.
I think I've found the root to the problem.
It turned out a larger value of nCorrectors(in the fvSolution under PIMPLE directory) needs to be used.
While in the case of coarse mesh and small Re(100), a value of nCorrectors=1 suffices, when the mesh is refined, a larger value, say nCorrectors=3 needs to be used.
Another detail: I'm using fixed courant about 2 to dynamic adjust the timestep. So I think my previous problem is not a result of the timestep.
To be honest I do not know the reason behind this

. I would appreciate it if someone could give me some hints.

Blanco · August 30, 2015, 06:51

It is very out of theory that a fine mesh gives worst result than a coarse mesh...it sounds to me like what you get with the coarse mesh is a lucky result, while the fine grid tells that something is not resolved properly... I don't have experience with DNS, so I cannot help further in detail. But, are you comparing cfd results with experiment at same Reynolds number? are you comparing lift and drag coefficients only or do you have any other data available?

squick · August 31, 2015, 16:38

Which kind of mesh do you use, structured, unstructured ... ?

See the post Weird results on Cavity and unstructured meshes.

June 29, 2015, 20:38	too dense mesh deteriorates the simulation	#1
kkpal Senior Member Join Date: Jan 2013 Posts: 134 Rep Power: 13	hi, I'm using pimpleFoam to simulate the flow around a circular cylinder. I found that many times that if a dense mesh is used for a low Reynolds number simulation, say Re=100, the simulation is not physically correct(in terms of the drag or lift force coefficients). However, if I maintain the same mesh and raise the reynolds number to like 1000, the simulation seems to go well. does someone else also have the same experience with me?

August 5, 2015, 07:57	y+=1	#3
fabian_roesler Senior Member Fabian Roesler Join Date: Mar 2009 Location: Germany Posts: 213 Rep Power: 18	Right, when using lowRe Models without wall functions, the refinement of the wall depends on the Reynolds number. You have to achieve a y+=1 for the first layer. When changing the Reynolds number from 100 to 1000, the first layer thickness to achieve y+=1 varies with round about factor 5. Cheers Fabian

August 5, 2015, 12:19		#4
kkpal Senior Member Join Date: Jan 2013 Posts: 134 Rep Power: 13	Thank you, Andrea and Fabian, I am not using any wallfunctions. Just pure DNS for low Re flows. I think I've found the root to the problem. It turned out a larger value of nCorrectors(in the fvSolution under PIMPLE directory) needs to be used. While in the case of coarse mesh and small Re(100), a value of nCorrectors=1 suffices, when the mesh is refined, a larger value, say nCorrectors=3 needs to be used. Another detail: I'm using fixed courant about 2 to dynamic adjust the timestep. So I think my previous problem is not a result of the timestep. To be honest I do not know the reason behind this. I would appreciate it if someone could give me some hints.

August 30, 2015, 06:51		#5
Blanco Senior Member Blanco Join Date: Mar 2009 Location: Torino, Italy Posts: 193 Rep Power: 17	It is very out of theory that a fine mesh gives worst result than a coarse mesh...it sounds to me like what you get with the coarse mesh is a lucky result, while the fine grid tells that something is not resolved properly... I don't have experience with DNS, so I cannot help further in detail. But, are you comparing cfd results with experiment at same Reynolds number? are you comparing lift and drag coefficients only or do you have any other data available? Last edited by Blanco; August 30, 2015 at 08:14.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
airfoil mesh genaration and simulation with a trip on surface issue	rajibroy	Mesh Generation & Pre-Processing	1	December 3, 2014 02:04
[ANSYS Meshing] 3d wind turbine mesh for multiphase simulation	mingersai	ANSYS Meshing & Geometry	0	January 17, 2012 18:20
fluent add additional zones for the mesh file	SSL	FLUENT	2	January 26, 2008 11:55
Icemcfd 11: Loss of mesh from surface mesh option?	Joe	CFX	2	March 26, 2007 18:10
How to control Minximum mesh space?	hung	FLUENT	7	April 18, 2005 09:38

August 5, 2015, 03:44		#2
Blanco Senior Member Blanco Join Date: Mar 2009 Location: Torino, Italy Posts: 193 Rep Power: 17	Hi kkpal, this is very different to the normal results improvement due to mesh elemnt size decrease and it is not aligned to numerical theory...have you checked everything in your model? are you using wall resolved boundary layers in particular?

August 31, 2015, 16:38		#6
squick New Member Gerard Join Date: Dec 2014 Posts: 7 Rep Power: 11	Which kind of mesh do you use, structured, unstructured ... ? See the post Weird results on Cavity and unstructured meshes.