[ICEM] improving mesh quality

kannstmasehn · April 24, 2013, 04:28

hi all,

I have a question concerning my mesh for a cyclone separator. You can see the mesh quality in the added picture. The bad quality can be seen in the other picture. I am using Star CCM+ and the mesh has 4,000k cells. With k-epsilon it converges but not with k-omega sst. Does someone has some tipps for me? Do I need to improve the quality at that place? If yes, how can I do it? I did a refinement but it didn't really help. Should the mesh be over a quality of 0.7? Is there a difference between k-epsilon and k-omega concerning the mesh quality?

thank you very much

energy382 · April 26, 2013, 06:15

Quote:

Originally Posted by kannstmasehn

hi all,

I have a question concerning my mesh for a cyclone separator. You can see the mesh quality in the added picture. The bad quality can be seen in the other picture. I am using Star CCM+ and the mesh has 4,000k cells. With k-epsilon it converges but not with k-omega sst. Does someone has some tipps for me? Do I need to improve the quality at that place? If yes, how can I do it? I did a refinement but it didn't really help. Should the mesh be over a quality of 0.7? Is there a difference between k-epsilon and k-omega concerning the mesh quality?

thank you very much

Please make a angle and det 3x3 plot and use another color, as the mesh is imperceptible.
If avaiable, upload blocking file or provide some pics.

Quality criterion isn't that important. Angle and orthogonal quality are more important in my point of view. Make a mesh independence study (coarse, reference, fine mesh) and look, who the variable of interest changes.

What kind of simulation will you run....transient? And is there any heat exchange etc..?

RodriguezFatz · April 29, 2013, 02:31

1) I agree, but that quality looks horrible. This is a hexa mesh, right? You will definitely be able to improve that. Can you show us some pictures of the blocking at the relevant locations?
2) This looks like some cyclone. Don't worry if not every single turbulence model will converge. Some are just not made for these strong swirling motions. Altough both models are not made for that purpose, k-e converges anyway... that doesn't sound unusual.
3) Please tell us the settings of your solver (especially solution methods: discretizations)! If you don't already do that, give this a quick try: Set pressure to "body force weighted" and try again with k-w.

kannstmasehn · April 29, 2013, 03:23

I made some screenshots for you. I hope they can help.

Here is the blocking

A link to another blocking picture:
http://s14.directupload.net/images/130429/5cm6q9gv.jpg

Here the quality is shown:

Here are my settings for the k-omega.

And here are my settings for the solver.

Thank you very much!

RodriguezFatz · April 29, 2013, 03:50

Ok, maybe someone else can help with the solver settings in StarCCM - I just know Fluent.
Can you upload the ICEM files (geometry and blocking)? I would like to give it a try...

kannstmasehn · April 29, 2013, 04:12

i hope the .tin file is right for the geometry, is it? both files are in the .zip file

RodriguezFatz · April 29, 2013, 04:45

Alright! Still disappointing, but I was able to enhance the grid partially.
Look at the picture:
screen.jpg
1) Disassociate the edge and the vertrex from geometry.
2) Now associate the two faces to the wall and the edge (2b) to the outer radius+the inlet curve
3) You will be able to move the vertex to the right, along the linet curve.

The green region is much better now. The blue region is still bad. Unfortunately the same trick doesn't work here, because your outer cylinder goes further down.
stairmand_blk.zip
Maybe someone else has a good idea.

kannstmasehn · April 29, 2013, 05:08

ok thank you, I'll give that a try.

BrolY · April 29, 2013, 05:28

That's the big problem with such a geometry. You will always have bad elements quality. For this kind of geometry, the min angle is the most critical quality criterion. As you can see, your determinant is very good (>0.2 is often good for any solvers), but your min angle is awful (<9° is often very bad most of solvers).

Adding a 1/4 of Ogrid at the junction would probably help.
Search on this forum about cyclone, that's something we have already worked on

About the differences between k-eps and k-omega, the Y+ is not the same, and so is the length of your first layer !!!

April 24, 2013, 04:28	improving mesh quality	#1
kannstmasehn New Member Tobi Join Date: Mar 2013 Posts: 26 Rep Power: 13	hi all, I have a question concerning my mesh for a cyclone separator. You can see the mesh quality in the added picture. The bad quality can be seen in the other picture. I am using Star CCM+ and the mesh has 4,000k cells. With k-epsilon it converges but not with k-omega sst. Does someone has some tipps for me? Do I need to improve the quality at that place? If yes, how can I do it? I did a refinement but it didn't really help. Should the mesh be over a quality of 0.7? Is there a difference between k-epsilon and k-omega concerning the mesh quality? thank you very much

April 29, 2013, 02:31		#3
RodriguezFatz Senior Member Philipp Join Date: Jun 2011 Location: Germany Posts: 1,297 Rep Power: 26	1) I agree, but that quality looks horrible. This is a hexa mesh, right? You will definitely be able to improve that. Can you show us some pictures of the blocking at the relevant locations? 2) This looks like some cyclone. Don't worry if not every single turbulence model will converge. Some are just not made for these strong swirling motions. Altough both models are not made for that purpose, k-e converges anyway... that doesn't sound unusual. 3) Please tell us the settings of your solver (especially solution methods: discretizations)! If you don't already do that, give this a quick try: Set pressure to "body force weighted" and try again with k-w. __________________ The skeleton ran out of shampoo in the shower.

April 29, 2013, 03:50		#5
RodriguezFatz Senior Member Philipp Join Date: Jun 2011 Location: Germany Posts: 1,297 Rep Power: 26	Ok, maybe someone else can help with the solver settings in StarCCM - I just know Fluent. Can you upload the ICEM files (geometry and blocking)? I would like to give it a try... __________________ The skeleton ran out of shampoo in the shower.

April 29, 2013, 04:45		#7
RodriguezFatz Senior Member Philipp Join Date: Jun 2011 Location: Germany Posts: 1,297 Rep Power: 26	Alright! Still disappointing, but I was able to enhance the grid partially. Look at the picture: screen.jpg 1) Disassociate the edge and the vertrex from geometry. 2) Now associate the two faces to the wall and the edge (2b) to the outer radius+the inlet curve 3) You will be able to move the vertex to the right, along the linet curve. The green region is much better now. The blue region is still bad. Unfortunately the same trick doesn't work here, because your outer cylinder goes further down. stairmand_blk.zip Maybe someone else has a good idea. __________________ The skeleton ran out of shampoo in the shower.

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
3D Hybrid Mesh Errors	DarrenC	ANSYS Meshing & Geometry	11	August 5, 2013 06:42
early stall, poor convergence, and mesh quality	everest	CFX	2	May 12, 2010 16:27
Icemcfd 11: Loss of mesh from surface mesh option?	Joe	CFX	2	March 26, 2007 18:10
mesh quality: urgent	Madhukar Rapaka	FLUENT	2	June 8, 2006 09:43

April 29, 2013, 03:23		#4
kannstmasehn New Member Tobi Join Date: Mar 2013 Posts: 26 Rep Power: 13	I made some screenshots for you. I hope they can help. Here is the blocking A link to another blocking picture: http://s14.directupload.net/images/130429/5cm6q9gv.jpg Here the quality is shown: Here are my settings for the k-omega. And here are my settings for the solver. Thank you very much!

April 29, 2013, 05:08		#8
kannstmasehn New Member Tobi Join Date: Mar 2013 Posts: 26 Rep Power: 13	ok thank you, I'll give that a try.

April 29, 2013, 05:28		#9
BrolY Senior Member AB Join Date: Sep 2009 Location: France Posts: 323 Rep Power: 21	That's the big problem with such a geometry. You will always have bad elements quality. For this kind of geometry, the min angle is the most critical quality criterion. As you can see, your determinant is very good (>0.2 is often good for any solvers), but your min angle is awful (<9° is often very bad most of solvers). Adding a 1/4 of Ogrid at the junction would probably help. Search on this forum about cyclone, that's something we have already worked on About the differences between k-eps and k-omega, the Y+ is not the same, and so is the length of your first layer !!!