Meshing Frustrations

[lkrenzler]

I'm new to Nastran but have used Inventor's simulation for years.  I'm wondering if there is any way to reliably arrive at a mesh that works?  The default settings always fail but no feedback on why is given.  Sometimes I am able to get things to mesh by adjusting the tolerance (in a blind trial and error sort of way) and sometimes not.  I have assemblies that have the exact same part (see screen shot) and one meshes and the other fails!  I'm wondering how that's possible?  The standard Inventor simulation environment has no problem with it.

The documentation is very little help.  Does anyone know of a procedure that would help arrive at a working mesh or is it just shooting at a black cat in a dark alley that may or may not exist? 

These are the same part.

[lkrenzler]

BTW, after installing the latest patch, even some parts that were OK are now failing, so even worse.  No amount of playing with the tolerance seems to fix it now.

[John_Holtz]

Hi @lkrenzler

Sorry to hear about the problems that you are having with the meshing. Mathematics can be strange in that way: two items that should be identical can have subtle differences. (I do not know the details of the meshing algorithm, but my observation is that meshing begins at certain locations that can be different even when you have identical parts.) It looks like your part may be a mirror image (depending on how the model was created), so that could be another wrinkle in the work.

If you are working with solids and not working with surfaces/shells, I suggest that you uncheck the "Continuous Meshing" box. That option probably does not do what you think it does, and it could be related to some of the meshing problems.

A general tip that I used to diagnose meshing problems is to change the Idealization from solid to shell, and try to mesh the model using triangular mesh. If the shells do not mesh, you might be able to see a face that does not mesh. That will at least focus your attention to a particular area. Sometimes splitting a large face into smaller areas can help with the meshing.

Feel free to attach your model if you are able too. (Please zip the part files and assembly files, .ipt and .iam.)

[lkrenzler]

I've attached the model.  This is just an extremely simple example.  The part that fails is the same part (just copied and constrained differently) not a mirrored part.

Unchecking "Continuous Meshing" does nothing.

Before I updated to 2019.1.1 I was able to get it to work by changing the tolerance to .0002 (just a blind guess as there seems to be little info on how this really works and what a valid range of values would be).  Now that I updated, this won't even work.

This used to work, but not now.

If it has trouble meshing a simple HSS cut at a 45 then there is a problem in my opinion.  I can get it to work by beveling the edge of the cut but if I have to manually modify parts just to get them to work then it's just not practical at all for us.  It really needs to be a bit more robust than that!  Making a working mesh should be pretty much automated at this point.

No it's not OK.  Could we at least get a hint as to what's wrong?

The default element size seems to make no sense.  It starts out at about 5" for an assembly that has 1.5" HSS in it.  If I use the slider set mid way it says 40".  Seriously?

I've tried adjusting mesh size, refinement ratio, min and max angle, growth ratio, suppress short features, min feature angle and midside nodes to every combination I can think of to no avail and since there is no feedback I don't know if I'm getting any closer or not.  Obviously, if we want to use this we can't spend all day playing with meshes.

The assembly below has identical parts that are even oriented the same direction but some fail and some don't.

If there is any way to consistently get a mesh to work, I'd really appreciate hearing it.  Thanks.

[John_Holtz · Accepted solution]

Hi @lkrenzler

I was misunderstanding your description that the parts were "copied". I assumed that you meant they were copied in the assembly, or multiple instances of the same part were added to the assembly. They are actually different part files that have similar (but different) commands used to make them, so they are different parts.

• part 000074154.ipt will mesh
• part 00074155.ipt will not mesh. When I try meshing this part as shell elements, the white edges (see image below) shows where the faces are not knitted together properly. There is no particular reason why; that's just the way math works.

Problem mesh area.

So you open the part that does not mesh (00074155.ipt) and rebuild it ("Manage > Rebuild All", and I usually do "Manage > Update Mass" for good measure). Then in the assembly you can do "Manage > Update" to make sure the assembly has all of the changes. Then the model will mesh in In-CAD.

I assume the same rebuilding of the parts will fix the more complex model, too.

______________________________________________________________

If my post answers your question, please click the "Accept as Solution" button. This helps everyone find answers more quickly!

[lkrenzler]

Thank you very much John, that did it!  I never would have guessed that.

And sorry, that was my mistake, in that example they were not the same part file.  In the more complex one they were.

[john.millsKPPEJ]

Even more bizarre.   Model meshes with controls in Linear Static.  Added modal analysis (had to rebuild mesh controls, perhaps I am missing something), and now one stupid small part refuses to mesh.  Exact same model.  Triple checked controls.  Do I have to even go in and make sure the parts are ordered the same in the list??? Its a combination solid shell model, so need continuous meshing.  So far, nothing works.

[john.millsKPPEJ]

Ok, how is this for messed up?  Every single part in the assembly has a specific mesh control.  HOWEVER, when I changed the overall mesh size (which should have no effect on anything) from 2 in to 3 like the previous, the one small stupid part now meshes!