Very large penetration in contact DISCRETE in dynamic simulation

mf

Hello,
first of all, until a few weeks ago, I almost always used contact CONTINUE, so I am not quite familiar with contact DISCRETE. However, I was quite surprised how well it works in dynamics, runtime is very good. But, I just saw, that it is possible to have a very penetration, although everything converges and all parts move as they should. Have a look at this:

The interpenetration is very large and visible with DEPL scale = 1! How is this possible?
The contacts are defined like this:

contact = DEFI_CONTACT(identifier='31:1',
FORMULATION='DISCRETE',
FROTTEMENT='COULOMB',
LISSAGE='OUI',
MODELE=model,
ZONE=(_F(ALGO_CONT='PENALISATION',
COULOMB=coeff_friction,
E_N=e_n,
E_T=e_t,
GROUP_MA_ESCL=('PISTON_CONT_PIN1', ),
GROUP_MA_MAIT=('PIN_CONT_PISTON1', ),
NORMALE='MAIT',
TYPE_PROJECTION='ORTHOGONALE'),

I am quite sure, I am missing something. All surfaces are oriented with ORIE_PEAU, intially all surfaces are at 0 tolerance in the step. Is my mesh too coarse? Any ideas?
Thanks,
Mario.

mf

[unknown]
Hello,
you shouldn't post copies or excerpts of standards, because it is a violation of copyright.
Mario.

jacob

Hello,
well, since you use PENALISATION you only have to increase E_N. Convergence does not mean low penetration and vice versa.
Or this doesn't help?

mf

jacob
Hello,

thanks, yes I know. My point is, that I have never encountered that with CONTINUE, where I always use PENALISATION also. My setup seems to be very sensitive to E_N, when I increase it, I tend to get non convergence in a later stage...

Maybe there's a setting I do not know about, which I forgot,

thanks,

Mario.

mf

[unknown]
Hello,

well thank you, I read Heinrich Hertz's original paper from 1881 many years ago https://home.uni-leipzig.de/pwm/web/download/Hertz1881.pdf. I don't see how this is of any help here.

This is an actual screenshot from a minute ago: the 10% took nearly 24h. That's what I am dealing with. Because I can't measure the penetration, I compare two nodes that are on opposing bodies directly next to each other. I don't even know if that works, because both bodies are allowed to rotate (there could be a DZ from rotation):

Do you recognize those 1mm downward bumps in the curve? I don't even know what is going on there, ideally this curve should somehow resemble a sine curve (100% roughly two cycles). I also tried CONTINUE, that's the red runcases.

Mario

mf

[unknown]
What do you mean? It says to control visually that there are no interpenetrations.

Which I did. In the first post. And there are interpenetrations. A lot.

Or do you mean REAC_GEOM = CONTROLE? This wouldn't help because it is already AUTOMATIQUE.

Mario.

mf

[unknown]
I still don't know what you mean. This is already shown in the first post, it is not clearly visible, but if you look closely, you see the overlapping (=interpenetrating) bodies. The graphics looks blotchy in the overlap area:

This is clearly interpenetration.

Mario.

mf

[unknown]
How does you having no penetration help me?

I can't use a much better mesh, otherwise the simulation will take forever. The forecast time is already 10 days.

Increasing E_N does not help much, because I lose convergence.

Your inner part seems to have a lot of clearance, mine doesn't even have clearance. It's a close fit at 0.0 everywhere.

Mario.

mf

I have a mail attached to this forum, I don't know how and if it works, but sometimes I get mails from users of this forum.
I don't see it in your case.
Does that work?

I can open your files in your link. I'll try to modify mine accordingly.
Thanks,
Mario.

mf

[unknown]
Yes I already did that.

Thanks, your E_N seems high tough, such a value only works with CONTINUE from my experience. But we'll see...

See the culprit below, the liner is making all the problems, the piston tumbles etc.

Mario.

mf

[unknown]
Because I was lazy and only wanted to know if I could do it. I know how pistons are designed, crankshafts not so much. Also the simulation would be even larger.

I only want to know if the piston does what it should.

I a few hours I will know if your parameters are any better.

Mario.

mf

[unknown]
Don't stay up late because of me. I hope I don't forget to answer.

Mario.

mf

[unknown]
Well, I had to restart because of an Error with the Graphics (the monitor going to sleep interferes with SM).

Now at t=6e-4s I needed 41 iterations in total, that's almost double the number of iterations (23 before), but my DIY-'penetration' value (SUIVI_DDL_1 minus SUIVI_DDL_2) is lower.

I'll wait until tomorrow, maybe it gets better when the load on the piston is gone (the 'ignition'),

thanks anyway,

Mario.

mf

[unknown]
It is still running, it is at 40%, so Monday evening it will be finished. It's very slow compared to the previous run, but I won't cancel it now. Already, a lot of time was invested,

Mario.