Convergence study in lattice optimization process

Altair Forum User

Hi,

 

There is a question in a process lattice optimization process.

 

Is the stress of any lattice cell's bar which is optimized from any mesh size converged?   

 

As you know, we always need a convergence study in FEA. In other words, the finer mesh gives us an accurate result in case of structural analysis so that finer mesh is preferred. At the same time, the cell size of a lattice is defined by a mesh size in case of optistruct lattice optimization. So, finer mesh will be better for stress convergence but too small mesh results in small lattice cell and it's impossible to manufacture by A.M. sometimes.

 

As a result, we plan to mesh a solid by a certain size first which is already confirmed to be printable in lattice, say 5mm length cubic. Then proceeding into lattice optimization process based on the mesh assures the printable lattice. But is this stress value converged? Because we are not sure whether the stress of the lattice is converged or not. Or is any lattice configuration stress converged since it's just a bar?

 

Thanks in advance for your help.

 

 

Altair Forum User

Hi Masa,

 

I need to check about the stress convergence. But, 1.8 penalties (POROSITY set to LOW) which is called 'Natural' penalty gave a good correlation between physical lattice structure and virtual intermediate densities. 

 

Altair Forum User

Prakash,

 

Thanks for your reply.

 

So, to set 1.8 penalties yields a good balance both numerical convergence and manufacturing in general. Is that correct? 

Altair Forum User

I am checking about the convergence part (pertaining to stress), however, in certain cases, a more porous structure may be favored for considerations other than stiffness. E.g.: buckling, thermal performance, etc. In such cases, the POROSITY parameter can be set to either MED or HIGH.

Altair Forum User

Thanks for your information.

 

It's interesting that the best porosity is different in several cases

 

differently.  I think these types of knowledge are not sheared enough and need to pay attention for applying a product design. Anyway, my current interest is stress and want to know fundamental rule to model desensitization for FEA and lattice size which can be printable then.   

Altair Forum User

HI,

 

Lattice members are modeled with beam elements. So there is no mesh convergence issue as beam models are accurate for any cell size to the extent beam theory applies (plane section remains plane).

 

Again, a lot of things come in to picture like local stress concentrations, type of printing adapted, etc... So one has to take care before proceeding. 

Altair Forum User

Hi.

 

Thanks for your advise.

 

I agree with you if the lattice is composed with bar elements about mesh convergence issue,however, in case of beams such as Euler beam or Timoshenko beam basically requires enough fine mesh descritizaion despite 1D. Anyhow, I think that the point is whether the lattice part is bar or not and no mesh convergence if it's bar. 

I appreciate you advise about this point.

 

And as you pointed out, there are too many things to consider now. So, I will focus on one of them first.

Thanks.