Learning Forum

[bottini.henny]

Hello everybody,

 

I am doing FEA of a wind turbine rotor hub with an academic licence of Workbench (windows-10 workstation: 32 cores, Xeon 2.4 GHz, 512 RAM GB, Nvidia RTX 5000).

I am new to Ansys Mechanical and would like to have the following points clarified.

 

1) I have the Distributed checkbox checked and 12 Cores set in the ribbon Home/Solve area (12 CPUs are the maximum allowed by my licence), and the meshing time for an ~1 million elements mesh is half a hour:

• Is it too long or not? I was expecting a faster meshing with 12 CPUs... Am I, actually, using 12 CPUs also for meshing, with that setting, or not?
• If I am using 12 CPUs, are 1~1.6 million elements considered such a big mesh? Actually, after these meshes are done, the FEA calculations terminate in ~ five minutes.
• What are the typical elements numbers to judge an Ansys Mechanical mesh coarse, middle-size, big, untractably big (for my PC specs)?
  

 

2) I am experiencing rather slow graphic-window refresh rates after mouse clicks or settings in these two circumstances:

•  when I select to view mesh elements with a specific value of quality metrics (e.g., skewness)
•  when I create a cross plane, and when I switch between a cross plane and the full-model view

In the first case I may have to wait for more than ten seconds before having the view updated, while in the second case I have to wait 3~5 seconds for the switch. Is this normal for my settings and PC specs, or I can improve these refresh rates with some more settings?

 

Thank you for your time and attention,

Henny

[Govindan Nagappan]

The distributed toggle and the core setting in the Home/Solve section is for the solver and not for the mesher. If you are using the default mesh method (sweep/patch conforming tetra), these methods are not parallel. However, if you have multiple parts in the model, each part gets meshed in a separate CPU. If you have a multibody part(for conformal meshing), then the bodies in the multibody part are meshed in 1 cpu. 

For information on memory requirements, check the Mechanical APDL documentation on performance. You can see information on how much memory is needed for 1 Million degree of freedom for sparse solver, iterative solver etc

Chapter 4: Memory Usage and Performance (ansys.com)

 

If you use Multizone method, this method can use multiple CPU's . 

Without looking at the model, I am not sure what could cause the meshing to be slow. For meshing, you have two setso of input that can check - mesh controls(sizing, method, details of mesh) and geometry. If this is machine related: Is this a shared resource that you are using? Are there other process running on this machine?

[Gary_S]

The slow graphics refersh may be caused by a 3D input device such as a spaceball, trackball or 3DConnexion device. Remove the device to confirm. Updating the driver often resolves the issue. 

Also, any sort of graphics capture program runing in the background can sometimes cause such issues. 

For licenses, make sure you are not referencing any old / invalid license servers.

[bottini.henny]

Dear Govindan and Gary,

thank you for your fast replies.

OK, I was not meshing with multiple CPUs because I have only one part and the default meshing method. Actually, meshing the coarsest (0.25 million elements) mesh takes two minutes.

I have understood that ther is no other way to mesh with multpile CPUs except for Multizone method and a model with more than one part: is this correct?

My meshing settings are:

-) gloabal element size

-) two local element sizes on two different sets of surfaces (one set with 16 faces, the other with 13 faces)

-) Face meshing applied on two sets of faces (one set with one face, the other with three faces): Internal number of divisions =2

-) Span angle center = medium

-) rest of settings at Ansys defaults

The workstation I use is not shared, and there is no other process running at the same time as Ansys besides standard Windows 10 background processes.

 

I replaced my trackball with a customary mouse, but found no sensible difference. And I closed the screen capture app that was actually open, but did not notice any improvement.

I am accessing the Ansys Workstation through remote desktop from an Ubuntu PC.

I'd appreciate if you could provide me with a rule of thumb to judge the sizes of my Ansys mesh.

[bottini.henny]

Thank you fort he fast replies.

 

My mesh settings are: global element size and two local element sizes on two different sets of surfaces; Sizing/Span Angle Center is set to medium; all the rest of the settings are at Ansys defaults.

Yes, meshing the coarsest (0.25 million elements) mesh takes two minutes, but now I know that the meshing process is not multi-CPU, so I am no more in doubt about the time for the ~1 million element mesh.

I have understood that there is no other way beside multizone meshing to use multiCPUs for meshing: is that correct?

The workstation I use is not shared, and there is no other process running at the same time as Ansys besides standard Windows 10 background processes.

 

I replaced my trackball with a customary mouse, but found no sensible difference . I am accessing the Ansys Workstation through remote desktop from an Ubuntu PC.

I closed the screen capture app that was actually open, but, again, did not notice any improvement.

 

As asked before, I'd appreciate if someone could tell me a rule of thumb to judge the size of an Ansys mesh.

 

Thank you again for your time and attention