TG mesh density

Red Ember · February 10, 2016, 02:23

I generated mesh for several compressor stages using turbogrid. Looking at borders between stages (e.g. - first stage and second stage) it's obvious that mesh density differs. Are there any suggestions for volume/length ratios for cells that belong to different domains? I use GGI interface option in CFX-PRE, but I think that's not enough...

ghorrocks · February 10, 2016, 06:48

I would go back to turbogrid and try to make the mesh size more even. That does not look good to me.

To prove whether it is a problem or not - do a simulation on this mesh and then significantly change the mesh and repeat it. If the results are significantly different then you know the mesh change you did is significant - which probably means the mesh quality at the interfaces is not adequate.

turbo · February 10, 2016, 10:14

Increase grid numbers in both inlet and outlet domains in TG.

You will see the function allows only a uniform spacing there, which is so frustrating. I had argued with Ansys support on this issue, and they said they would submit a request to improve it. I doubt they will do. That's what CFX is now.

Red Ember · February 11, 2016, 02:13

ghorrocks,
yes, I tried to make elements at borders more equal, although it caused some increase of nodes number. I'll try to compare results.

turbo,
I used "increase edge refinement" function for local mesh density smoothing. Number of elements at inlet/outlet is linked to number at blade surface, so it's not easy to create mesh without local high concentration of elements. Moreover, I see no way to open all domains using TG, I compare grids using postprocesor. Rather annoying too...

Red Ember · February 12, 2016, 01:16

My refinements yielded no significant differences.
Too bad, I thought mesh was a reason why calculated compressor efficiency value is very low - about 60%...

-Maxim- · February 12, 2016, 09:52

in order to get a 'nicer' mesh around the blades, you could try another topology in TG. Activate advanced/beta features to get more options.
I usually work with edge refinements as well, so that sounds good to me.

Maybe post in the meshing subforum in case you would like more improvements there.
Otherwise you might share and discuss the rest of your project so that we could try to help to find out why you don't get your expected compressor efficiency.

edit: just saw your other post about the compressor efficiency

turbo · February 12, 2016, 11:08

Quote:

Originally Posted by Red Ember

My refinements yielded no significant differences.
Too bad, I thought mesh was a reason why calculated compressor efficiency value is very low - about 60%...

Because your blade design was wrong, not the mesh.

Red Ember · February 12, 2016, 14:29

-Maxim-,
actually I don't blame mesh now, mesh was fine within every domain. Then I made cell sizes near domain interfaces more equal, so that's enoufg for now. I just wonder whether exist any ratio criteria for elements that belong to different domains but have common border or not.

turbo,
I made several assumptions that increase numerical errors.
1. I have no geometry for the last, 7th stage, so I simulate only 6 stages (plus inlet guide vanes). I get outlet pressure via interpolation (not proper accurate, but better than nothing). I use outlet flow direction normal to outlet surface for now. I'll estimate approximate angle value a bit later.
2. I used merge operation to get one flowpath sketch for 6 stages at once. That distorted hub surface:
http://www.cfd-online.com/Forums/ans...istortion.html
I think I should chop this sketch in 13 parts to improve hub line, although I suspect somewhat problems with domain interfaces geometry can pop out.
3. Shroud tips and hub tips - not all of them can be simulated via turbogrid tip options (or may be I just am not aware of it). Tip span and absolute values are not constant! Tip and hub/shroud curves are not equidistant in my case. So I had to accept normal distance tip option.
4. I faced problems with curves around fillets. No exportpoints were created. I'll try to create more curves around fillets.
5. Inlet guide vanes and first stage guide vanes are variable. So there are dashed hub and shroud tips that I can't simulate using TG.

But blades and vanes design is good enough due to static pressure contours.

turbo · February 15, 2016, 22:56

I assume your task is to simulate an existing 7-stage axial compressor in CFX by importing blade solids from CAD. What you did is not a recommended procedure. You need to build a BladeGen model of each blade row from xyz surface coordinates from CAD. You have to visit BG 15 times for your case. Then TG > CFX.

Red Ember · February 17, 2016, 14:42

turbo, thanks for advice!
Yep, existing compressor. But why it is not recommended? That's definitely new for me, can you give me reason why? Or share a link, please! I'd like to read smth about it.
So! If I get you right, I should use BG instead of Design Modeler and Blade Modeler preference?

turbo · February 17, 2016, 17:14

That is why You made such many assumptions in the geometry including flowpath curves. If you split each row to model in BG, you can minimize any mismatch with the real blades.

Red Ember · February 17, 2016, 23:44

turbo, is TG able to reproduce fillets and non-trivial curvature at end clearance if I use BladeGen before TurboGrid? And how I should get coords for BladeGen - import 3D models or get coords in CAD soft at first and then use it in BladeGen?

turbo · February 18, 2016, 14:45

If you want blade fillets and/or custom tip clearance, BG > DM > TG > CFX

Red Ember · February 18, 2016, 23:45

And what about coords?

turbo · February 20, 2016, 10:04

Get all xyz surface coordinates at spanwise sections > BGD model > Fillets in DM > Export to TG > CFX

February 12, 2016, 09:52		#6
-Maxim- Senior Member Maxim Join Date: Aug 2015 Location: Germany Posts: 413 Rep Power: 13	in order to get a 'nicer' mesh around the blades, you could try another topology in TG. Activate advanced/beta features to get more options. I usually work with edge refinements as well, so that sounds good to me. Maybe post in the meshing subforum in case you would like more improvements there. Otherwise you might share and discuss the rest of your project so that we could try to help to find out why you don't get your expected compressor efficiency. edit: just saw your other post about the compressor efficiency Last edited by -Maxim-; February 12, 2016 at 09:53. Reason: saw other post

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
sliding mesh problem in CFX	Saima	CFX	46	September 11, 2021 08:38
[ICEM] Using mesh Density	Daniel_Khazaei	ANSYS Meshing & Geometry	0	May 2, 2014 19:00
Mesh motion with Translation & Rotation	Doginal	CFX	2	January 12, 2014 07:21
[snappyHexMesh] Layers:problem with curvature	giulio.topazio	OpenFOAM Meshing & Mesh Conversion	10	August 22, 2012 10:03
[snappyHexMesh] snappyHexMesh won't work - zeros everywhere!	sc298	OpenFOAM Meshing & Mesh Conversion	2	March 27, 2011 22:11

February 10, 2016, 06:48		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,871 Rep Power: 144	I would go back to turbogrid and try to make the mesh size more even. That does not look good to me. To prove whether it is a problem or not - do a simulation on this mesh and then significantly change the mesh and repeat it. If the results are significantly different then you know the mesh change you did is significant - which probably means the mesh quality at the interfaces is not adequate.

February 10, 2016, 10:14		#3
turbo Senior Member Join Date: Jun 2009 Posts: 174 Rep Power: 17	Increase grid numbers in both inlet and outlet domains in TG. You will see the function allows only a uniform spacing there, which is so frustrating. I had argued with Ansys support on this issue, and they said they would submit a request to improve it. I doubt they will do. That's what CFX is now.

February 11, 2016, 02:13		#4
Red Ember Member Alex Join Date: Feb 2016 Posts: 81 Rep Power: 10	ghorrocks, yes, I tried to make elements at borders more equal, although it caused some increase of nodes number. I'll try to compare results. turbo, I used "increase edge refinement" function for local mesh density smoothing. Number of elements at inlet/outlet is linked to number at blade surface, so it's not easy to create mesh without local high concentration of elements. Moreover, I see no way to open all domains using TG, I compare grids using postprocesor. Rather annoying too...

February 12, 2016, 01:16		#5
Red Ember Member Alex Join Date: Feb 2016 Posts: 81 Rep Power: 10	My refinements yielded no significant differences. Too bad, I thought mesh was a reason why calculated compressor efficiency value is very low - about 60%...

February 12, 2016, 14:29		#8
Red Ember Member Alex Join Date: Feb 2016 Posts: 81 Rep Power: 10	-Maxim-, actually I don't blame mesh now, mesh was fine within every domain. Then I made cell sizes near domain interfaces more equal, so that's enoufg for now. I just wonder whether exist any ratio criteria for elements that belong to different domains but have common border or not. turbo, I made several assumptions that increase numerical errors. 1. I have no geometry for the last, 7th stage, so I simulate only 6 stages (plus inlet guide vanes). I get outlet pressure via interpolation (not proper accurate, but better than nothing). I use outlet flow direction normal to outlet surface for now. I'll estimate approximate angle value a bit later. 2. I used merge operation to get one flowpath sketch for 6 stages at once. That distorted hub surface: http://www.cfd-online.com/Forums/ans...istortion.html I think I should chop this sketch in 13 parts to improve hub line, although I suspect somewhat problems with domain interfaces geometry can pop out. 3. Shroud tips and hub tips - not all of them can be simulated via turbogrid tip options (or may be I just am not aware of it). Tip span and absolute values are not constant! Tip and hub/shroud curves are not equidistant in my case. So I had to accept normal distance tip option. 4. I faced problems with curves around fillets. No exportpoints were created. I'll try to create more curves around fillets. 5. Inlet guide vanes and first stage guide vanes are variable. So there are dashed hub and shroud tips that I can't simulate using TG. But blades and vanes design is good enough due to static pressure contours.

February 15, 2016, 22:56		#9
turbo Senior Member Join Date: Jun 2009 Posts: 174 Rep Power: 17	I assume your task is to simulate an existing 7-stage axial compressor in CFX by importing blade solids from CAD. What you did is not a recommended procedure. You need to build a BladeGen model of each blade row from xyz surface coordinates from CAD. You have to visit BG 15 times for your case. Then TG > CFX.

February 17, 2016, 14:42		#10
Red Ember Member Alex Join Date: Feb 2016 Posts: 81 Rep Power: 10	turbo, thanks for advice! Yep, existing compressor. But why it is not recommended? That's definitely new for me, can you give me reason why? Or share a link, please! I'd like to read smth about it. So! If I get you right, I should use BG instead of Design Modeler and Blade Modeler preference?

February 17, 2016, 17:14		#11
turbo Senior Member Join Date: Jun 2009 Posts: 174 Rep Power: 17	That is why You made such many assumptions in the geometry including flowpath curves. If you split each row to model in BG, you can minimize any mismatch with the real blades.

February 17, 2016, 23:44		#12
Red Ember Member Alex Join Date: Feb 2016 Posts: 81 Rep Power: 10	turbo, is TG able to reproduce fillets and non-trivial curvature at end clearance if I use BladeGen before TurboGrid? And how I should get coords for BladeGen - import 3D models or get coords in CAD soft at first and then use it in BladeGen?

February 18, 2016, 14:45		#13
turbo Senior Member Join Date: Jun 2009 Posts: 174 Rep Power: 17	If you want blade fillets and/or custom tip clearance, BG > DM > TG > CFX

February 18, 2016, 23:45		#14
Red Ember Member Alex Join Date: Feb 2016 Posts: 81 Rep Power: 10	And what about coords?

February 20, 2016, 10:04		#15
turbo Senior Member Join Date: Jun 2009 Posts: 174 Rep Power: 17	Get all xyz surface coordinates at spanwise sections > BGD model > Fillets in DM > Export to TG > CFX