Post Processing Tetrahedral mesh

AliMadayen · October 12, 2014, 19:42

Hi everyone

I'm studying a very complex geometry which I cannot generate structured mesh for it. Hence, I used ICEM to mesh it with commonly tetrahedral elements (more than 90%)
Checking the grid independency with number of the elements varying from 300K to 1000K, results are independent from the mesh.
But the problem is, the chart that I draw for checking independency is not smooth, I've uploaded an image of it.
Is this due to poor mesh quality, or it is an intrinsic characteristic of the unstructured meshes? I should mention that the solution is well converged (Residuals 1e-5 and imbalance 1e-3.
Is there a way to make these graphs smooth?
I have not worked with unstructured grids that much so sadly i'm not familiar with them

Best
Ali

ghorrocks · October 12, 2014, 20:05

Firstly - a mesh sensitivity sweep from 300k to 1M nodes is unlikely to be wide enough to capture anything useful. The parameter you should be sweeping is the element edge length. So a halving of the element edge length will result in approximately 2^3 = 8 times the number of nodes.

Your graph shows a parameter which should be smoothish but looks bumpy. You are correct in saying:
* Yes, this can be due to poor mesh quality.
* Yes, unstructured meshes frequently do this.

But mesh refinement should smooth it out more and reduce the wobblyness.

AliMadayen · October 13, 2014, 07:59

Thanks for your help.
I've checked the 2.2M grid but the problem still exists.
I've extracted geometry from CT angio imaging by Mimics software, and I doubt smoothness of the surface. I plotted velocity in cross section for various meshes and they were identical.
Is this parameters for the mesh acceptable?
I couldn't find any suitable range for edge-length ratio, but for face angle values more than 10 is said to be acceptable.

60<Face Angle<160
1<Edge Length Ratio<13
Connectivity number: 2-42
volume ratio: 1-125
(I guess high volume ratio is due to the coarse-mesh entrance volume that i've placed in the inlet for having fully developed flow at the main volume inlet)

I would appreciate if you could comment on these parameters.

ghorrocks · October 13, 2014, 08:02

Before considering any of the issues you discuss, have a look at the geometry you are modelling? How smooth is it? If it is rough - especially with roughness of the same length scale as your mesh - then no matter what you do you are going to get a lumpy results curve.

AliMadayen · October 13, 2014, 08:11

I could plot this. X coordinate on the surface polyline.
I guess for a grid element size of 2.1e-4 these oscillations are not suitable.
Is there any better way to determine the smoothness?

ghorrocks · October 13, 2014, 18:24

I have no idea what that graph is showing. You would have to show an image of your geometry for that to make sense.

Have a look at the wall boundary facets. Is the wall being resolved as a smooth curve, or does each facet have its own angle?

October 12, 2014, 20:05		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,870 Rep Power: 144	Firstly - a mesh sensitivity sweep from 300k to 1M nodes is unlikely to be wide enough to capture anything useful. The parameter you should be sweeping is the element edge length. So a halving of the element edge length will result in approximately 2^3 = 8 times the number of nodes. Your graph shows a parameter which should be smoothish but looks bumpy. You are correct in saying: * Yes, this can be due to poor mesh quality. * Yes, unstructured meshes frequently do this. But mesh refinement should smooth it out more and reduce the wobblyness. AliMadayen likes this.

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
engrid -> save as .stl with boundarie codes	Zymon	enGrid	31	August 29, 2011 14:40
Icemcfd 11: Loss of mesh from surface mesh option?	Joe	CFX	2	March 26, 2007 19:10
How to control Minximum mesh space?	hung	FLUENT	7	April 18, 2005 10:38

October 13, 2014, 07:59		#3
AliMadayen New Member Ali Madayen Join Date: Apr 2014 Posts: 22 Rep Power: 12	Thanks for your help. I've checked the 2.2M grid but the problem still exists. I've extracted geometry from CT angio imaging by Mimics software, and I doubt smoothness of the surface. I plotted velocity in cross section for various meshes and they were identical. Is this parameters for the mesh acceptable? I couldn't find any suitable range for edge-length ratio, but for face angle values more than 10 is said to be acceptable. 60<Face Angle<160 1<Edge Length Ratio<13 Connectivity number: 2-42 volume ratio: 1-125 (I guess high volume ratio is due to the coarse-mesh entrance volume that i've placed in the inlet for having fully developed flow at the main volume inlet) I would appreciate if you could comment on these parameters.

October 13, 2014, 08:02		#4
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,870 Rep Power: 144	Before considering any of the issues you discuss, have a look at the geometry you are modelling? How smooth is it? If it is rough - especially with roughness of the same length scale as your mesh - then no matter what you do you are going to get a lumpy results curve.

October 13, 2014, 18:24		#6
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,870 Rep Power: 144	I have no idea what that graph is showing. You would have to show an image of your geometry for that to make sense. Have a look at the wall boundary facets. Is the wall being resolved as a smooth curve, or does each facet have its own angle?