CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   ANSYS Meshing & Geometry (https://www.cfd-online.com/Forums/ansys-meshing/)
-   -   [ICEM] Dramatic decrease in Mesh Quality After creating Prism Layers? (https://www.cfd-online.com/Forums/ansys-meshing/222495-dramatic-decrease-mesh-quality-after-creating-prism-layers.html)

pamstad November 26, 2019 13:20

Dramatic decrease in Mesh Quality After creating Prism Layers?
 
Hello everybody,

I have used the following procedure to get an unstrucutred volume mesh:
  • First I created a Volume mesh with Octree Algorithm
  • In a second step I delete the volume mesh and used the surface Mesh from Octree as an Input for creating a Volume Mesh with Delaunay.
  • So far so good; I get a godd quality, the lowest element is around 0.47
  • The last step is including prism layers
  • There I give as an input the growth ratio, growth law (exponential), height and number of layers
after inserting the prism layers the quality drops drastically, and even after smoothing the lowest element does not get above 0.2.

The problem is that I dont really see what I could do differently?

Thanks a lot for your answer and have a nice day

siw November 28, 2019 04:56

This always happens. Your procedure is the best way of making hybrid meshes in ICEM. Smoothing prism layers is difficult. Have you read Simon's Tips & Tricks in https://www.cfd-online.com/Forums/an...ers-here.html? See Slide 62.

pamstad November 28, 2019 09:23

Thank you siw for your reply.

I have found the following post which helped me a lot and I was actually able to get a quite good quality (badest element 0.31), including the prism layers. So here is the post I am talking about:

"I promised you some instructions on improving prism layer generation. Prism generates prism layers by extruding the surface mesh into the volume. As it marches into the mesh, it uses the existing Octree mesh to find it's way. After the last prism layer is generated, the neighboring tet mesh is split, if necessary and the prisms and surrounding tet mesh are smoothed.

Most of the problems encountered with prism generation are due to this last step. If you have a refined surface mesh, but the volume mesh expands quickly away from the surface, the neighboring tets after the prism layer may be 4 to 16 times larger than the triangular face of the last prism. This makes it very difficult to split and smooth the mesh effectively. Furthermore, since the prisms are being extruded through progressively larger tets, the smoothness of the prism layers may suffer as well (although this has been improved in version 5.0).

With this in mind, I suggest using the 'width' parameter when defining the surface element sizes. The width parameter is an integer value that tells Tetra to refine the elements in the volume next to a surface to be the same size as the surface mesh, up to 'width' elements deep. In other words, if you specify width as 4, Tetra will make the first 4 elements away from the surface the same size as the surface element. To achieve a smooth transition from the last prism layer into the volume, you should specify about 1/2 as many width layers as you want prism layers. So if you want 20 prism layers, use a width of 10. You may refine this based on your own experience as well.

A good prism layer has two main qualities; it should have a constant first node height (at least along a given surface) and it should provide a low volume ratio between the last prism element and the first tet. Prism allows you to ensure both of these criteria. The "Initial Height" parameter specifies the distance to the first node. The "Max height over base" parameter specifies the maximum ratio of thickness of the extruded dimension (height) to the edge length of the triangle being extruded (base). Generally, if you assume the first tet is nearly isotropic, a base/height ratio of about .8 will give you a volume ratio of nearly one.

If you follow these instructions, you should be able to generate a pretty good prism mesh. The three parameters to remember are "width", "initial height" and "max base over height". Please feel free to forward this to your colleagues and provide feedback to me on this advice."


What I experienced:

- I got the best results as follows: When I created the prism layers I did not give any initial height as criterion. Initial height and Total height were set to zero.
- Then after the prism layers were created in went to the edit mesh tab --> move nodes --> Redistribute Prism Edge and then fixed the inital height to the desired one.

Maybe this helps someone.

Have a nice day!

siw November 28, 2019 10:08

A rule of thumb is for ICEM element quality > 0.3 is good. But that does not apply to prism element shapes. You should inspect the prism elements in your mesh.

shereez234 November 29, 2019 16:58

Quote:

Originally Posted by pamstad (Post 750983)
Thank you siw for your reply.

I have found the following post which helped me a lot and I was actually able to get a quite good quality (badest element 0.31), including the prism layers. So here is the post I am talking about:

"I promised you some instructions on improving prism layer generation. Prism generates prism layers by extruding the surface mesh into the volume. As it marches into the mesh, it uses the existing Octree mesh to find it's way. After the last prism layer is generated, the neighboring tet mesh is split, if necessary and the prisms and surrounding tet mesh are smoothed.

Most of the problems encountered with prism generation are due to this last step. If you have a refined surface mesh, but the volume mesh expands quickly away from the surface, the neighboring tets after the prism layer may be 4 to 16 times larger than the triangular face of the last prism. This makes it very difficult to split and smooth the mesh effectively. Furthermore, since the prisms are being extruded through progressively larger tets, the smoothness of the prism layers may suffer as well (although this has been improved in version 5.0).

With this in mind, I suggest using the 'width' parameter when defining the surface element sizes. The width parameter is an integer value that tells Tetra to refine the elements in the volume next to a surface to be the same size as the surface mesh, up to 'width' elements deep. In other words, if you specify width as 4, Tetra will make the first 4 elements away from the surface the same size as the surface element. To achieve a smooth transition from the last prism layer into the volume, you should specify about 1/2 as many width layers as you want prism layers. So if you want 20 prism layers, use a width of 10. You may refine this based on your own experience as well.

A good prism layer has two main qualities; it should have a constant first node height (at least along a given surface) and it should provide a low volume ratio between the last prism element and the first tet. Prism allows you to ensure both of these criteria. The "Initial Height" parameter specifies the distance to the first node. The "Max height over base" parameter specifies the maximum ratio of thickness of the extruded dimension (height) to the edge length of the triangle being extruded (base). Generally, if you assume the first tet is nearly isotropic, a base/height ratio of about .8 will give you a volume ratio of nearly one.

If you follow these instructions, you should be able to generate a pretty good prism mesh. The three parameters to remember are "width", "initial height" and "max base over height". Please feel free to forward this to your colleagues and provide feedback to me on this advice."


What I experienced:

- I got the best results as follows: When I created the prism layers I did not give any initial height as criterion. Initial height and Total height were set to zero.
- Then after the prism layers were created in went to the edit mesh tab --> move nodes --> Redistribute Prism Edge and then fixed the inital height to the desired one.

Maybe this helps someone.

Have a nice day!

You are right. It's a nice cheat. Create prism without any height. You could try splitting the prisms and equally and then re-distributing too.

One other thing you can do is increase the chord-wise and span-wise elements just on the surface of the body. But this is of course limited by your computing capability. ps; I prefer hexa meshing over any unstructured methods.

pamstad December 5, 2019 02:23

Thank you siw and sheerez234 for your answers.

@siw: You said: "You should inspect the prism elements in your mesh." What do you mean exactly with inspect? With the Quality tool? Or how do you do it? Thanks a lot for your answer

@sheerez234: You mentioned that you would increase chord and span-wise elements just on the surface of the body. So you mean that after having my volume mesh I should make the surface mesh finer? Is that even possible? Or dou you mean that I should make the surface mesh I've got from the Octree Volume meshing finer before I create again a Volume Mesh with Delaunay? Sorry, I dn't get what you mean...


Have a nice day!

Patrick


All times are GMT -4. The time now is 00:40.