ICEM smoothing function and distorted elements

Nick R · March 22, 2011, 19:26

Hi there,

I've meshed an airfoil using the C-Grid scheme. My problem is that when I initiate smoothing on hex elements I get distorted parts towards the trailing edge. My intention is to smooth the mesh for the transition from the outer block shown in the picture to the inner boundary layer block. Please see the pictures of the mesh before and after smoothing. Any comments are much appreciated.

Nick

alastormoody11 · March 24, 2011, 05:30

Hi,

the mesh smoothener is there for the unstructured hexa mesh which is generated in a way similar to an unstructured tet mesh.

What you have created using blocking is a structured hex mesh using blocking,etc for which the use of smoothener is not advisable.

The smoothener will go all out to improve the quality of the mesh and thereby destroying the structured nature of your mesh which would results in a loss of accuracy during the simulation since the loss of structured nature of your mesh will cost you much more in terms of accuracy than the minor gain you will gain by using a better quality mesh

Nick R · March 24, 2011, 06:23

Thanks for the response. I saw this done in a tutorial that's why I thought I'd try it. Would you mind commenting on quality inwhen it comes to creating a finer mesh to yield lower y+ plus values? The problem I have is that once I decrease the first node distance from the surface of the foil my mesh determinant goes from 0.7 to elements in the 0.1 region. Can this be corrected? Or is it not a big deal...I thought the smoother would fix this issue but it's obviously not working. Thanks again.

alastormoody11 · March 24, 2011, 06:29

Hi,

How does your quality reduce so drastically?

You might want to try using a mesh law half cosinus to capture the boundary layer.

If you want to reduce the size of the forst layer of cells you can set a large number of nodes on the curve representing the aerofoil.

Nick R · March 24, 2011, 06:37

The quality of most elements remain the same but about 70 cells fall between 0.1 and 0.7. I don't know if this should be a concern. I'll give it a shot with the cosine law.
How does increasing the number of nodes on the foil help with y+ though? Doesn't the first element height determine the y+ value?

alastormoody11 · March 24, 2011, 06:47

Ya,

Sorry about that, you are correct it wont make a difference.

If you are using blocking to create a structured mesh, you ideally should not elements below quality 0.5, just check which are your problem elements, also check association.

BrolY · March 24, 2011, 06:53

You can have elements with a quality around O.3 with blocking (especially around o-grid ..), it would be ok. It really depends on your geometry, and then on your blocking.

You should post pictures of your blocking where your quality elements is dropping. It might helps !

PSYMN · March 24, 2011, 21:14

There are two smoothers... The old unstructured smoother that has been in ICEM CFD for a long long time is not ideal for structured grids because it optimizes individual elements... But there is a newer orthogonal smoother that he may have seen in a video or two, this elliptical smoother is getting better and better each release...

My guess is that the problem is due to associations... The smoothers (like all mesh movement tools) rely on association to know what features need to be maintained... If you don't associate the trailing edge (blocking) with the trailing curve (geometry) of the wing, then the nodes along that edge are not associated with the curve and the smoother may try to move them along the surfaces causing all sorts of problems. Similarly for vertex to point associations...

Nick R · March 24, 2011, 23:05

@alastormoody11 : That's cool. Thanks for suggestions.

@Simon: That was it, association was the culprit! Although in the pre-mesh that wouldn't cause any misalignment, after smoothing it'd show up! That was excellent. Many thanks Simon.

Nick R · March 27, 2011, 04:28

@BroLY

Please see the elements shown in the picture that drop in quality once I reduce the first layer height. Any suggestions to improve the mesh quality would be great!

PSYMN · March 27, 2011, 20:48

Zoom in a lot closer to see why... There is probably some projection problem due to geometry issues. These can usually be fixed by one of a few settings under settings => meshing => Hexa/Mixed that will either relax or improve the projection.

Post a zoomed in picture of the poor mesh near a curve and I can tell you which...

Nick R · March 27, 2011, 22:36

Thanks for your attention. I am posting three screenshots. I can see how the mesh doesn't quite follow the curves at those spots..don't know how to fix it..tried moving the vertices around and increasing the nodes but didn't work..your advice would be invaluable

March 27, 2011, 20:48		#11
PSYMN Senior Member Simon Pereira Join Date: Mar 2009 Location: Ann Arbor, MI Posts: 2,663 Blog Entries: 1 Rep Power: 47	Zoom in a lot closer to see why... There is probably some projection problem due to geometry issues. These can usually be fixed by one of a few settings under settings => meshing => Hexa/Mixed that will either relax or improve the projection. Post a zoomed in picture of the poor mesh near a curve and I can tell you which... __________________ ----------------------------------------- Please help guide development at ANSYS by filling in these surveys Public ANSYS ICEM CFD Users Survey This second one is more general (Gambit, TGrid and ANSYS Meshing users welcome)... CFD Online Users Survey

March 24, 2011, 05:30		#2
alastormoody11 Senior Member Join Date: Jun 2009 Posts: 100 Rep Power: 17	Hi, the mesh smoothener is there for the unstructured hexa mesh which is generated in a way similar to an unstructured tet mesh. What you have created using blocking is a structured hex mesh using blocking,etc for which the use of smoothener is not advisable. The smoothener will go all out to improve the quality of the mesh and thereby destroying the structured nature of your mesh which would results in a loss of accuracy during the simulation since the loss of structured nature of your mesh will cost you much more in terms of accuracy than the minor gain you will gain by using a better quality mesh

March 24, 2011, 06:23		#3
Nick R Senior Member Nick Join Date: Nov 2010 Posts: 126 Rep Power: 16	Thanks for the response. I saw this done in a tutorial that's why I thought I'd try it. Would you mind commenting on quality inwhen it comes to creating a finer mesh to yield lower y+ plus values? The problem I have is that once I decrease the first node distance from the surface of the foil my mesh determinant goes from 0.7 to elements in the 0.1 region. Can this be corrected? Or is it not a big deal...I thought the smoother would fix this issue but it's obviously not working. Thanks again.

March 24, 2011, 06:29		#4
alastormoody11 Senior Member Join Date: Jun 2009 Posts: 100 Rep Power: 17	Hi, How does your quality reduce so drastically? You might want to try using a mesh law half cosinus to capture the boundary layer. If you want to reduce the size of the forst layer of cells you can set a large number of nodes on the curve representing the aerofoil.

March 24, 2011, 06:37		#5
Nick R Senior Member Nick Join Date: Nov 2010 Posts: 126 Rep Power: 16	The quality of most elements remain the same but about 70 cells fall between 0.1 and 0.7. I don't know if this should be a concern. I'll give it a shot with the cosine law. How does increasing the number of nodes on the foil help with y+ though? Doesn't the first element height determine the y+ value?

March 24, 2011, 06:47		#6
alastormoody11 Senior Member Join Date: Jun 2009 Posts: 100 Rep Power: 17	Ya, Sorry about that, you are correct it wont make a difference. If you are using blocking to create a structured mesh, you ideally should not elements below quality 0.5, just check which are your problem elements, also check association.

March 24, 2011, 06:53		#7
BrolY Senior Member AB Join Date: Sep 2009 Location: France Posts: 323 Rep Power: 22	You can have elements with a quality around O.3 with blocking (especially around o-grid ..), it would be ok. It really depends on your geometry, and then on your blocking. You should post pictures of your blocking where your quality elements is dropping. It might helps !

March 24, 2011, 21:14		#8
PSYMN Senior Member Simon Pereira Join Date: Mar 2009 Location: Ann Arbor, MI Posts: 2,663 Blog Entries: 1 Rep Power: 47	There are two smoothers... The old unstructured smoother that has been in ICEM CFD for a long long time is not ideal for structured grids because it optimizes individual elements... But there is a newer orthogonal smoother that he may have seen in a video or two, this elliptical smoother is getting better and better each release... My guess is that the problem is due to associations... The smoothers (like all mesh movement tools) rely on association to know what features need to be maintained... If you don't associate the trailing edge (blocking) with the trailing curve (geometry) of the wing, then the nodes along that edge are not associated with the curve and the smoother may try to move them along the surfaces causing all sorts of problems. Similarly for vertex to point associations...

March 24, 2011, 23:05		#9
Nick R Senior Member Nick Join Date: Nov 2010 Posts: 126 Rep Power: 16	@alastormoody11 : That's cool. Thanks for suggestions. @Simon: That was it, association was the culprit! Although in the pre-mesh that wouldn't cause any misalignment, after smoothing it'd show up! That was excellent. Many thanks Simon.