Refinement on the whole sample was working correctly before, but as soon as I did the refinement on each contact part, the simulation did not work correctly.
Now I will add amplitude, I want to make sure if I understood correctly or not, column A is the displacement (BC) and here instead of the number 1, I should enter the maximum displacement?
Enter the data as it is. Then reference the amplitude in the boundary condition definition (with ~ 6 mm magnitude like before). Amplitude value will be multiplied by BC magnitude at each time step.
I did the same but it still doesnât work. For example, Iâve only done mesh refinement on the edge of the clip (No 1)
As soon as I remove the mesh refinement it works correctly (if I reduce the size of all the elements it works correctly, but when I add local mesh refinement it doesnât work!!!) (No 2)
Are you sure that the amplitude name is referenced in the boundary condition definition ? Can you show the window with this BC ?
Sorry, I forgot to reference the domain name in the boundary condition definition. I referred it now and the answer was as follows:
In practice, you will probably need much more refinement to capture the response properly (especially with those sharp edges that are tricky for contact). Thatâs why a model reduced to the joint itself can be a good idea.
As far as I know mesh density can mess the convergence if the slave mesh ends up coarser than the master.
Does that responds the answer?
1 Like
Thank you. I re-entered the following values as the domain and got a better answer,
But I donât know why the elements in contact area at the end have interference.
Trying to simulate with accuracy the âpost clippingâ moment (when the parts pass the maximun height to the minimun again) is very very very complex, it need insannely small time step and lot of iterations. For knowing the max stress during clipping is enough to reach the maximun displacement and stay with that results.
If you want to know the unclipping force, just start your simulation from the clipped postition.
1 Like
As I said before, this mesh is way too coarse and the contact conditions are nasty (edge to edge/surface). Sometimes adding node to surface contact helps but you should focus on the mesh first (once again - try to reduce the model to the minimum).
1 Like
I didnât know this until now, but I tested it now and it didnât work.
Thanks a lot to you and SergioP1975, I will definitely check these points and let you know the result.
Itâs a general common rule in FEA. For example, from CalculiX Userâs Manual:
Finally a few useful rules if you experience convergence problems:
⢠Deactivate NLGEOM (i.e. do not use NLGEOM on the *STEP card).
⢠Try SMALL SLIDING first, and then large sliding, if applicable.
⢠Try a linear pressure-overclosure relationship first (instead of exponential), with a stiffness constant about 5 to 50 times Youngâs modulus of the adjacent materials.
⢠Define your slave surface based on faces, not on nodes. This can be especially helpful if you use quadratic elements.
⢠Make sure that the mesh density on the slave side is at least as fine as on the master side, preferably finer.
⢠Deactivate friction first.
2 Likes
That small corner I posted can help to achieve it in a reasonable time step.
1 Like
And significantly finer meshâŚ
Whenever you have a time, please share the second simulation file.
Thatâs the one:
snap fit mod.pmx (2.6 MB)
This is the correct design for snap fit. first step is to follow best practice or design guidelines for snaps. after wards define the purpose of the simulation. in most cases the only concern is the beam maximum stresses under maximum deflection.