Onshape Simulation - Slow / Not working

graeme_allen846

I'm wondering if anyone else is able to enjoy Onshape's simulation feature the way it was seamlessly demo'd and functioning at nearly real-time speeds.  If I'm lucky it will actually manage to work through it's calculations but it will take HOURS.  If I click on Method>Mates>Show Connections, the wheel just spins forever.  Is there something in the way I'm building my assembly that kills the efficiency?  I'm using fasten mates and then choosing the inner hole faces as "simulation connection".  Looking for any insight on how you build your assemblies or if you just use the inferred connections from the import.

I've made an example of the type of model I'm working with:
https://cad.onshape.com/documents/49bd0953be3f91faf35538e1/w/92ef1e42e444031c776ddbd2/e/ed6d666a778a69f54ee9224d

Cheers

S1mon

Are you on a Mac? "Show Connections" has horrible performance on my M1Max Mac. It seems to be a known issue. Very very simple simulations will let me "show connections" but anything complex will hang or kill WebGL within Chrome.

billy2

Your model looks good and everything is fastened. I'm suspecting "thins". Sheet metal parts should be modeled with shell elements. No one knows what OS is doing but I suspect that it's meshing the parts. With sheet metal, the element size, at a minimum is the thickness of the sheet metal which leads to mega meshes. I've also noticed that they seem to mesh refine automatically, so a proper sheet metal mesh should be more than 1 element thickness. Now you're at multi-mega mesh.

I've never accepted simulation results without seeing a mesh. You just don't know what's going.

Start with a small sheet metal part and grow from there. 

In any simulation, Start simple and add complexity. Never start with the assembly and try to work backwards trying to find the errors. None of them work that way, not even abacus.

Can you analysis an "L" bracket and obtain meaningful results? I suspect it won't take much more to kill things because of the thins.

billy2

do you see the diamond patterns in the results?



These are usually created at the mesh boundaries. When there are large transitions between the boundaries, that's not good and you should refine the mesh in those areas. You see the red going to green in the upper left hand corner? That's a stress riser due to the fixed constraint. It's hard to get rid of that type of stuff. It does show you the approximate mesh size though.

I could be wrong, maybe they're using a meshless solver. I doubt it, meshless solvers don't have results looking like the above image. Meshless solvers have their own problems.

I use OS simulation a lot, but I've been doing simulations for a long time and know what to look out for.

The best way to solve your problem, ask for shell elements. They're coming because the world's problems are more easily solved with shell elements.

graeme_allen846

Dell Precision 3530

graeme_allen846

@billy2,  Thank you for the look, I've since given up on the OS simulation and trying my luck with Onscale.  Its interesting information and experience you present, as a newcomer to FEA I'm not able to make use of it but I'm sure someone else will find your insights helpful!