How to fix high equivolume and centroid skewness in pointwise?

Laurens_NLD · May 18, 2020, 11:13

Hi all,

I'm having trouble with my mesh of a birds wing in a windtunnel. The centroid and equivolume skewness are quite high (0.92 and 0.9999 something respectively). I'm having trouble with running the meshes in Fluent and according to the errors Fluent is giving, I strongly expect that this high skewness is the problem. The skewness of the surface mesh on the birds wing was quite high at first too, but I have managed to reduce it to around 0.80 for both the equiangle and equiarea skewness.

However, if I try to reduce the equivolume and centroid skewness (i've tried for example changing the grow rate, making the farfield finer, changing the skewness criteria when generating the 3D Trex), the skewness doesn't change or it even resulted in pointwise stopping after the boundary layer was created when I tried changing the skewness criteria.

equivolume_skewnessbiggerorequalthen_0.80_1.jpg
equivolume_skewnessbiggerorequalthen_0.80_2.jpg
Centroid_skewness_biggerorequalthen_0.80_1.jpg
Centroid_skewness_biggerorequalthen_0.80_2.jpg
Centroid_skewness_biggerorequalthen_0.80_3.jpg

In the pictures you can see where the elements are located with an equivolume or equiskewnes >= 0.80. It seems like it that they are all located at the transition region between boundary layer and farfield.

Some info that might be usefull:
-All the cells that have a equivolume >= 0.80 are tets.
-At the leading edge the cells with a high centroid skewness are at the transition from hexes to prisms.

I would appreciate any advice that you could give me regarding the decreasing of the skewness values.

Thanks in advance!

With kind regards,
Laurens

Bertdevries · May 20, 2020, 09:45

Honestly, it looks like you really do not need to fix your grid. In my experience, Ansys CFX is fine working with these skewness levels. If you really need to get rid of the high skewness, you might also consider grid refinement,but this will of course increase computational complexity. Hope this helps!

tcarrigan · May 20, 2020, 12:34

Hello Laurens,

There are quite a few skewness metrics and each is calculated differently as you've found. The metric that is triggering minimum orthogonal quality in Fluent is equivolume skewness and isn’t one we use for most CFD simulations as it is overly restrictive, is not tied to the numerics of the solver, and is not a proper measure of element quality since any stretching (even good, aligned, anisotropic elements) gets flagged as poor. Instead, we recommend looking at element geometry such as max included angle, cell non-orthongality, and centroid skewness. For this particular grid, your metrics look fine.

Regards,
Travis

Laurens_NLD · May 27, 2020, 04:41

Hi Travis,

Thanks for your reply! That explains a lot.

With kind regards,
Laurens

May 18, 2020, 11:13	How to fix high equivolume and centroid skewness in pointwise?	#1
Laurens_NLD New Member Laurens Schalk Join Date: May 2020 Posts: 3 Rep Power: 6	Hi all, I'm having trouble with my mesh of a birds wing in a windtunnel. The centroid and equivolume skewness are quite high (0.92 and 0.9999 something respectively). I'm having trouble with running the meshes in Fluent and according to the errors Fluent is giving, I strongly expect that this high skewness is the problem. The skewness of the surface mesh on the birds wing was quite high at first too, but I have managed to reduce it to around 0.80 for both the equiangle and equiarea skewness. However, if I try to reduce the equivolume and centroid skewness (i've tried for example changing the grow rate, making the farfield finer, changing the skewness criteria when generating the 3D Trex), the skewness doesn't change or it even resulted in pointwise stopping after the boundary layer was created when I tried changing the skewness criteria. equivolume_skewnessbiggerorequalthen_0.80_1.jpg equivolume_skewnessbiggerorequalthen_0.80_2.jpg Centroid_skewness_biggerorequalthen_0.80_1.jpg Centroid_skewness_biggerorequalthen_0.80_2.jpg Centroid_skewness_biggerorequalthen_0.80_3.jpg In the pictures you can see where the elements are located with an equivolume or equiskewnes >= 0.80. It seems like it that they are all located at the transition region between boundary layer and farfield. Some info that might be usefull: -All the cells that have a equivolume >= 0.80 are tets. -At the leading edge the cells with a high centroid skewness are at the transition from hexes to prisms. I would appreciate any advice that you could give me regarding the decreasing of the skewness values. Thanks in advance! With kind regards, Laurens

May 20, 2020, 12:34		#3
tcarrigan Senior Member Travis Carrigan Join Date: Jul 2010 Location: Arlington, TX Posts: 161 Rep Power: 16	Hello Laurens, There are quite a few skewness metrics and each is calculated differently as you've found. The metric that is triggering minimum orthogonal quality in Fluent is equivolume skewness and isn’t one we use for most CFD simulations as it is overly restrictive, is not tied to the numerics of the solver, and is not a proper measure of element quality since any stretching (even good, aligned, anisotropic elements) gets flagged as poor. Instead, we recommend looking at element geometry such as max included angle, cell non-orthongality, and centroid skewness. For this particular grid, your metrics look fine. Regards, Travis __________________ Travis Carrigan Manager, Business Development Pointwise, Inc.

May 20, 2020, 09:45		#2
Bertdevries New Member Join Date: May 2020 Posts: 1 Rep Power: 0	Honestly, it looks like you really do not need to fix your grid. In my experience, Ansys CFX is fine working with these skewness levels. If you really need to get rid of the high skewness, you might also consider grid refinement,but this will of course increase computational complexity. Hope this helps!

May 27, 2020, 04:41		#4
Laurens_NLD New Member Laurens Schalk Join Date: May 2020 Posts: 3 Rep Power: 6	Hi Travis, Thanks for your reply! That explains a lot. With kind regards, Laurens