[ICEM] courant number <-> mesh quality

kpax · November 20, 2012, 05:20

hey folks,

i am currently encountering the problem that even for very small time-steps (0.0001), the courant number of my problem explodes at some point of the simulation (in icoFoam). Since i have to simulate ~50s in total, I really dont wanna go to even smaller time steps.
I think the problem is mesh quality, because the simulation stayed stable a bit longer when I used more non-orthogonal corrector steps. plus i could see in the results that the velocity exploded in regions of low mesh quality...

so my question is this: what kind of mesh errors have the biggest influence on the courant number, according to your experience?
so far, i have mainly focused on improving equiangle skewness, but i'm not sure if that is really the best choice.

btw: both "smooth elements globally" and "smooth hexahedral mesh - orthogonal" did not work for me at all (made it all worse) - is there someone else who experienced that?

best,
kpax

BrolY · November 20, 2012, 05:43

First of all, you have to know that if you increase the determinant too much, it could decrease the angle, and vice versa .. But, if you increase the aspect ratio, it could increase the determinant. Most of the criteria are linked.

From my experience, angle is worth for hexa. A very bad angle could crash your simulation. Could you post some pictures of your problem. For example, a picture of the region where your simulation crashes with the picture of your mesh at the same location.

Also, it depends on the conditions of your simulation : very high velocity in area where your mesh is not refinned is not good at all !

kpax · November 20, 2012, 09:03

ok, attached you find screenshots of the areas of high velocity (at the outlet), the mesh at the outlet, and areas of low skew values (0.4-0.5, i thought that would not be too bad?)..
obviously, the high u values appear at the corners of the o-grid. i tried moving the vertices around, but it didn't improve results. weird, isnt it?

Quote:

Originally Posted by BrolY

Also, it depends on the conditions of your simulation : very high velocity in area where your mesh is not refinned is not good at all !

i thought a fine mesh and high velocity would be the worst that can happen in terms of the courant number? because c = u*dt/dx then becomes large...

BrolY · November 20, 2012, 10:46

About the high velocity and fine mesh, the problem doesn't come from the Courant, but comes from the solver. You need more elements to capture more information.

About the velocity profile at the outlet, it looks weird.
What solver do you use ? Are you sure everything is correctly configured with your solver ? Because, from the picture of your mesh, it looks like there is no trouble with the mesh !
Unless other users see something I do not, maybe you should ask your question in the solver forum.

Far · November 21, 2012, 02:07

did you try the elliptical smoother? It comes under blocking and use the option orthogonality.

kpax · November 21, 2012, 10:02

Quote:

Originally Posted by BrolY

About the velocity profile at the outlet, it looks weird.
What solver do you use ? Are you sure everything is correctly configured with your solver ? Because, from the picture of your mesh, it looks like there is no trouble with the mesh !
Unless other users see something I do not, maybe you should ask your question in the solver forum.

broly, you were absolutely right. some of my BC were not adjusted correctly... once if fixed that the velocity profile looked alright.
thx a lot for pointing me in the right direction!!

Quote:

Originally Posted by Far

did you try the elliptical smoother? It comes under blocking and use the option orthogonality.

yep, made things worse. but maybe that is because my mesh is in fact not too bad to begin with? or i might not have chosen the best configuration, there are a lot of options...

kpax · November 26, 2012, 09:22

hey everybody,

turns out i still have some problems with the courant number. after around 50 time steps (each with dt = 0.0001), the simulation crashes due to very high U and p values. i attached a picture of the area where p values get very high - obviously not a reasonable result.
strangely though, this area does not seem to have worse-than-average mesh elements, as you can see in the other picture, which shows elements with determinant between 0.45 and 0.65 (all other elements have higher determinant).

in some publications, i have seen reasonable results for my kind of simulation with roughly the same number of mesh elements, but dt = 0.01. i wonder what's the difference?

sharonyue · December 20, 2012, 12:06

Hi kpax,
I am using OpenFOAM too, and I am facing some similary problem with you.
http://www.cfd-online.com/Forums/ope...behaviour.html
in unstructured mesh the quality is not good. in structured mesh the quality is above 0.8. I think thats the mesh quality which incur the result explode on twophaseeulerfoam. If you handle your problem, I can get it a hint.~

November 20, 2012, 05:20	courant number <-> mesh quality	#1
kpax Member Join Date: Jul 2012 Posts: 31 Rep Power: 14	hey folks, i am currently encountering the problem that even for very small time-steps (0.0001), the courant number of my problem explodes at some point of the simulation (in icoFoam). Since i have to simulate ~50s in total, I really dont wanna go to even smaller time steps. I think the problem is mesh quality, because the simulation stayed stable a bit longer when I used more non-orthogonal corrector steps. plus i could see in the results that the velocity exploded in regions of low mesh quality... so my question is this: what kind of mesh errors have the biggest influence on the courant number, according to your experience? so far, i have mainly focused on improving equiangle skewness, but i'm not sure if that is really the best choice. btw: both "smooth elements globally" and "smooth hexahedral mesh - orthogonal" did not work for me at all (made it all worse) - is there someone else who experienced that? best, kpax

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
[snappyHexMesh] Add Mesh Layers doesnt work on the whole surface	Kryo	OpenFOAM Meshing & Mesh Conversion	13	February 17, 2022 08:34
Moving mesh	Niklas Wikstrom (Wikstrom)	OpenFOAM Running, Solving & CFD	122	June 15, 2014 07:20
[snappyHexMesh] How to Do External Mesh for Airfoil sHM	msuaeronautics	OpenFOAM Meshing & Mesh Conversion	1	September 23, 2012 05:00
[snappyHexMesh] snappyHexMesh won't work - zeros everywhere!	sc298	OpenFOAM Meshing & Mesh Conversion	2	March 27, 2011 22:11
[blockMesh] BlockMeshmergePatchPairs	hjasak	OpenFOAM Meshing & Mesh Conversion	11	August 15, 2008 08:36

November 20, 2012, 05:43		#2
BrolY Senior Member AB Join Date: Sep 2009 Location: France Posts: 323 Rep Power: 22	First of all, you have to know that if you increase the determinant too much, it could decrease the angle, and vice versa .. But, if you increase the aspect ratio, it could increase the determinant. Most of the criteria are linked. From my experience, angle is worth for hexa. A very bad angle could crash your simulation. Could you post some pictures of your problem. For example, a picture of the region where your simulation crashes with the picture of your mesh at the same location. Also, it depends on the conditions of your simulation : very high velocity in area where your mesh is not refinned is not good at all !

November 20, 2012, 10:46		#4
BrolY Senior Member AB Join Date: Sep 2009 Location: France Posts: 323 Rep Power: 22	About the high velocity and fine mesh, the problem doesn't come from the Courant, but comes from the solver. You need more elements to capture more information. About the velocity profile at the outlet, it looks weird. What solver do you use ? Are you sure everything is correctly configured with your solver ? Because, from the picture of your mesh, it looks like there is no trouble with the mesh ! Unless other users see something I do not, maybe you should ask your question in the solver forum.

November 21, 2012, 02:07		#5
Far Senior Member Sijal Join Date: Mar 2009 Location: Islamabad Posts: 4,558 Blog Entries: 6 Rep Power: 54	did you try the elliptical smoother? It comes under blocking and use the option orthogonality.

December 20, 2012, 12:06		#8
sharonyue Senior Member Dongyue Li Join Date: Jun 2012 Location: Beijing, China Posts: 849 Rep Power: 18	Hi kpax, I am using OpenFOAM too, and I am facing some similary problem with you. http://www.cfd-online.com/Forums/ope...behaviour.html in unstructured mesh the quality is not good. in structured mesh the quality is above 0.8. I think thats the mesh quality which incur the result explode on twophaseeulerfoam. If you handle your problem, I can get it a hint.~