'stamp' simulation project by cbabua


#1

I created a new simulation project called 'stamp':

stamp exercise


More of my public projects can be found here.


#2

Nice deep drawing project @cbabua!
I just noticed that you defined a linear elastic material for the aluminum sheet. This could of course be done on purpose to get a first impression of the simulation settings without having all nonlinearities activated from the first try.

If, on the other hand, you need some realistic elasto-plastic material properties for aluminum, you can have a look at this project of a stamping simulation where I used elasto-plastic material properties for the aluminum sheet.

If you have any further questions, I’m happy to assist!

Best,
Richard


#3

@rszoeke

Do you have an idea, why the calculation takes so long?


#4

Hi @cbabua,
actually there are several reasons:

  1. Generally the mesh size is rather large for a highly nonlinear analysis (almost 1 Million DOFs), this results in a high solution time per iteration
    -> Use a manually refined mesh for the curved surfaces and the sheet metal and a very coarse mesh for the stamp to optimize mesh efficiency. Using a hexahedral mesh would be another option, but this is currently not possible to generate on SimScale, you would have to create it locally and upload the mesh.

  2. You have A LOT of slave contact nodes. The number of contact nodes on the slave side are the main factor that influence the speed for the contact resolution.
    -> Usually we can split the slave surfaces to and only assign the relevant sub-faces as slaves, but in your case this seems to be limited to possibly a part of your upper sheet surface.

  3. Your convergence seems to be very slow. This can have multiple reasons.
    -> Enable contact smoothing to smoothen your surface normals to not have sudden changes of normals between elements.
    -> For a first test you could try to soften the contact and use a smaller penalty coefficient. This might result in significant penetration, but might at least make a first approximate result possible.

  4. Your numerics settings might not be optimal.
    -> Choosing only 8 cores for the parallel calculation on a 32 core machine might improve the speed (doing so you would run the problem on 8 MPI processes each having 4 core shared memory parallelization).

I hope this enables you to get some initial results! I would really like to see them.

Best,
Richard


#5

Hey @cbabau!

As suggested by @rszoeke! just follow the above points and hopefully you would get good results in fairly less time. On top of that what you can also do is to take only quarter of your model since it’s symmetric along all directions and than apply symmetry boundary condition. You can have a look at these projects for more info.:

Similar as your case:

[Radial Stamp Deep Drawing] (https://www.simscale.com/projects/rszoeke/radial_stamp_2/)

For smaller model:

[Impact of an elastoplastic bar] (https://www.simscale.com/projects/ahmedhussain18/impact_of_an_elastoplastic_bar/)

Hope this helps. If you have any question/s, feel free to ask. Happy Simulating!

Good luck!


#6

Yeah!!! It works! :wink:
Thanks!


#7

Glad to hear! :smile:


#8

Is the metal sheet plastical deformed? 5e+08 Pa > Rp 0,2 of Al or is the sheet broken? :smiley: