Learning Forum

[vedama]

Hello,

As a follow-up to my earlier question (https://community.ansys.com/?status%2F38737-38737-1658880053%2F), what do you mean by the mesh is revolved but is not 2.5D? It is still a one-cell thick mesh, which is what is specified in the documentation.

If I cannot revolve/rotate the mesh, and have to just extrude (translate), that doen't seem to work either. I tried to just extrude another 2D mesh to make it one cell thick, ie 2.5D, and I get the same error message:

"~~~~~~~~~~~~~~~~~~~~~~~~~ Adaption is not permitted ~~~~~~~~~~~~~~~~~~~~~~~~~~
Reason:
- Selected adaption mode: PUMA (2.5D)
is inappropriate for the current mesh
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Here is an isometric view of my extruded/translated one-cell thick 2.5D mesh:

[Rob]

In the other image it looks like the volume collapses to an axis. What did you extrude, and which Fluent solver are you using? Ie 2d, 3d etc. 

[vedama]

I made a 2D grid which I extruded (rotated for the first case, translated for the second) by one cell. So it's a 3D prism grid but just one cell thick.

I'm using Fluent's 3D solver in parallel

[Rob]

Anything set that conflicts with https://ansyshelp.ansys.com/account/Secured?returnurl=/Views/Secured/corp/v221/en/flu_th/flu_th_chp_adaption.html this lot? 

[vedama]

No, I believe my grid conform to all the limitations. I'd be happy to send it to you, if you wish or if that's possible.

[Rob]

We're not able to do that. By all means post the mesh on here and someone other than Ansys Staff may have a look. 

What's the cell quality like? Also, why do you want it to be one cell thick? Fluent allows 2d and 2d-axisymmetric models. 

[vedama]

 

I was using 2D but I ran into a couple of issues. There were these non-physical oscillations present in the solution that I couldn’t explain. I asked about them here (/forum/forums/topic/wild-oscillations-in-cp-and-cf/) but to no reasonable answer. At which point I was told to switch over to PUMA since it is more robust. I also keep getting a notification that hanging node adaption for 2D will be phased out and that I should switch to PUMA 2.5D.

As for mesh quality, it’s a poor mesh and I’ve done this intentionally, to evaluate the improvement in grid + solution after adaption. Does the mesh quality affect why PUMA 2.5D doesn’t work though? This is what I get when I evaluate mesh quality:

Minimum Orthogonal Quality = 5.24949e-01 cell 69 on zone 2 (ID: 37 on partition: 19) at location (-1.86587e-02 1.40334e-01 -1.22467e-03)
Maximum Aspect Ratio = 3.12470e+04 cell 38 on zone 2 (ID: 388 on partition: 5) at location ( 5.54834e-01 7.61255e-02 -6.64337e-04)

 

 

[Rob]

I wonder if the high aspect ratio is causing a problem. When the cell is split the aspect ratio will increase by about a factor of 4, and that's potentially going to mean the cell centre can't be accurately placed in the mesh.  

[vedama]

Why would the 2.5D algorithm be deemed incompatible if there is a high aspect ratio cell? It should still perform the adaption and then if I end up with a negative volume or a bad cell, so be it. But I'm not even able to activate the algorithm.

[Rob]

There may be a check, the PUMA tool is designed for poly cells so it'll be working out the centroid position: that calculation may be failing. 

[vedama]

So how do I fix it?

Image 1) I tried a much finer mesh and I get the same error.

Image 2) I converted the domain to polyhedra cells (against Fluent's recommendation) and I get the same error.

Image 3) I made it so there's only tetrahedrals (inviscid, no high aspect ratio boundary layer cells) and I get the same error.

Image 4) I converted tets-only grid to polyhedral cells and I get the same error.

[Rob]

What is the aspect ratio in each case?