Mesh Convergence

[lancermandc_co_za]

Hi All,

I'm very new to Inventor Nastran so please be gentle with me 🙂

How on earth do I get my mesh convergence plot to show anything but a straight horizontal line??

[John_Holtz]

Hi @lancermandc_co_za . Welcome to the Inventor Nastran forum.

You mentioned the convergence plot was a "straight horizontal line". But the image shows a single point, and the model is setup for global mesh refinement (which produces a graph with a single point). Were you using the local mesh refinement instead of global mesh refinement when you got the straight line?

Regardless, whenever the automated convergence does not work, you need to refine the mesh manually. "Mesh > Mesh Control" and specify the smaller mesh size and locations. It looks like you just need to refine the faces around the inside of the outer holes. Since your model is symmetric and has a number of repeating patterns, you can refine each hole with a different mesh size! For example, one hole with a mesh size of 10 mm (giving 2 elements through the thickness), another hole with a mesh size of 5 mm, another with a mesh size of 2.5 mm.

John

[lancermandc_co_za]

Hi John,

Firstly, thank you for taking the time to help me.

Whether I use global refinement or Local refinement, I either get a dingle point on the graph or a horizontal line (see attachment.

Could you walk me through obtaining an actual graph that represents convergence using my model?

[John_Holtz]

Sorry I was not clear. When the automatic convergence does not work, there is no way to make it work.

Instead, you turn off the automatic convergence and do a convergence study manually:

1. Run the analysis with an initial mesh size.
2. Enter the mesh size and result into Excel.
3. Add Mesh controls to the model where the stress was high or where the stress gradient (change in stress per length) is high.
4. Regenerate the mesh.
5. Run the analysis.
6. Repeat steps 2 through 5 until the graph levels off. Make sure that the location of the stress you write down is the same. In other words, you cannot use the max stress value if it changes from one location to another location in the model. (In your model, the max stress may move from one hole to another hole. Since everything is symmetric, that is not a problem as long as the position within the hole is the same.)

Note that some users are reporting that local mesh convergence worked in version 2024 but not in version 2025. I will have to do some tests. If I confirm that finding, I will report it to the development team.

John

[lancermandc_co_za]

Hi John,

Ok, that is much clearer now 🙂

Once again, thank you for your time to help. Much appreciated...

Hopefully the techies will figure out a way to program this feature to make it more user friendly....

Regards, 

[lancermandc_co_za]

I was just thinking, are you saying that Inventor can't create a convergence, but I can create one manually?

[John_Holtz]

I do not know if the "Mesh > Convergence Settings" is completely broken, or if it does not work on certain models. (I only hear about the situations where it does not work. 🙂).

Yyou can always specify mesh refinements at regions where you know you need a smaller mesh. Human intelligence (HI) may not be faster than artificial intelligence (AI), but at the moment HI is still better than AI, and better than pre-programmed logic.

"A" Convergence Settings

"B" Mesh Control

John

[lancermandc_co_za]

Hi,

Just a bit of info again, I know (we all do) that Inventor's FEA meshing and Nastran meshing are very different in complexity so, what I did find is that Inventor's FEA is able to plot the convergence graph whereas Nastran cannot....