Personal tools
You are here: Home Forum Use Re: Local sizes fails near edge

Re: Local sizes fails near edge

Up to Use

Re: Local sizes fails near edge

Posted by Giovanni Medici at June 03. 2020

Dear all,

I'm having some issues with refinement (local sizes).

I'm meshing a 2D case (axisymmetric), use NETGEN-1D-2D, to mesh the face. It has three local sizes defined (three solids), with different values of refinement, moreover it has viscous layers on the body.

For some reasons, the refinement is doing fine inside the face; but when it reaches an edge, it sort of does not make it. I tried to disable the "optimize" option, but the problem persists.

It is partially mitigated if I set on that edge a 1D algorithm, but matching it with the refinement zones is hard.

I add here a picture of the resulting mesh.

 

Is it a known problem / expected behavior? What do you suggest to do to mitigate it?

 

Thanks

 

Tested on SALOME 9.3.0. Ubuntu, and similarly it happens under windows.

 

Attachments

Re: Local sizes fails near edge

Posted by Christophe Bourcier at June 04. 2020

Hi,

This happens when you use "Wire discretisation" in 1D or "Netgen 2D simple parameters" (which is a shortcut to Netgen 2D and wire discretisation).

Can you check that you use "Netgen 2D parameters" and that you define the edge refinament on the "local size" tab of Netgen 2D parameters?

Christophe

Re: Local sizes fails near edge

Posted by Giovanni Medici at June 08. 2020

Thanks Christophe for the answer. 

I confirm that I used Netgen 2D parameters, nevertheless in "local size" tab I added only solids (3 refinement solid that intersect the 2D face I'm meshing). 

I don't understand how to can I "define the edge refinament on the "local size" tab of Netgen 2D parameters", as the edge (symmetry axis) shall be refined only partially.

Shall I create another edge superimposed to the symmetry axis, whose length covers the refinement zone? 

Thanks

P.S. in order to try to mitigate the problem I also tried Netgen + Wire discretization, trying to match the right growth rate, but it is hardly achievable.

 

Re: Local sizes fails near edge

Posted by Saint Michael at June 08. 2020

Hi

It's hard to say what is wrong w/o seeing how algorithms are defined. Can you share your study? Or can you reproduce the issue on simple boxes?

St.Michael

Re: Local sizes fails near edge

Posted by Giovanni Medici at June 08. 2020

Hi Saint Michael, 

thanks, please find below a simple study where the problem close to the symmetry edge (x axis) is visible.

https://gofile.io/d/roSVt9

In the file I shared, I just ran the Compute mesh.

 

On my local file, after the mesh creation, I created a filter on the TRIAS, to laplace smooth them, things improve a bit, but still the mesh has some severely skewed TRIAS (not saved in the study I shared).

 

(SALOME 9.3.0, windows). 

P.S. By the way, is there any way to deal with sharp edges in the viscous layers? Maybe something similar to the option there is in 3D meshing (nose offset) could slightly improve the mesh quality around sharp edges.

 

Previously Saint Michael wrote:

Hi

It's hard to say what is wrong w/o seeing how algorithms are defined. Can you share your study? Or can you reproduce the issue on simple boxes?

St.Michael

 

Re: Local sizes fails near edge

Posted by Saint Michael at June 09. 2020

Hi

I checked the assigned hypotheses and everything seems OK. I don't know why it does not work ((. Apparently there are some bugs.

St.Michael

 

Re: Local sizes fails near edge

Posted by SMESH expert at June 09. 2020

Hello,

I investigated why refinement is not done on an edge. This is because of a NETGEN bug that it defines edge discretization by dividing an edge into 1000 samples where it checks element size. This causes problem if the element size at some sample is smaller than a 1/1000 of edge length.

It is possible to workaround this bug by dividing the "too long" edge into halves. I did this workaround and the refinement became OK, nonetheless a 2D mesh near the edge was invalid.  I got a valid 2D mesh only by applying 3D algorithm and 3D Viscous Layers.

Best regards

Edward

Re: Local sizes fails near edge

Posted by Giovanni Medici at June 09. 2020

Thanks St. Michael and Edward for your time and hints. 

I'll try to partition the long edges, didn't know about the 1/1000 limit. As far as the viscous layer behavior around sharp corners it is interesting the option of 3D meshing and then take only the symmetry plane face.

It would be nice to have an option to "share" the sharp corners (say over an user defined value of degrees e.g. 60ยบ), between an user defined number of nodes (3 on each edge, for example), the resulting viscous layers would be skewed to a certain degree, but the overall size growth would be a little less extreme, and aspect ratios could be kept under contorl (possibly).

 

Thanks to the community for the help!

Powered by Ploneboard
Document Actions