Yade

damping in ViscElMat

Asked by feda

Hello,
I am simulating a periodic isotropic compression of well graded material (Cu>3) to reach the stable loosest state (target porosity).
I use ViscElMat to define the material (Kn,Ks,cn,cs,frictionangle,density):
-The friction angle is set to 0.7 (the porosity will increase by increasing the friction)
-cn and cs are in functions of a predefined critical damping ratio, kn and the mean radius of particles.
-The damping in Newton Integrator is set to 0. The energy dissipation is then introduced in the system by means of a viscous damping defined at contacts.
This set of parameters allows obtaining porosity smaller than the target porosity (a more dense state) and if I increase the values of cn and cs by increasing the critical damping ratio, I obtain the same value of porosity.

Now, if I set cn and cs to 0 and I use a global numerical damping instead, I can reach the target porosity (obtain a loosest state).
How we can interpret this difference, and which method is better to use with viscElMat?
Thank you
feda

Question information

Language:
English Edit question
Status:
Solved
For:
Yade Edit question
Assignee:
No assignee Edit question
Solved by:
Jérôme Duriez
Solved:
Last query:
Last reply:
Revision history for this message
Best Jérôme Duriez (jduriez) said : 		#1

Hi,

I think you should choose what is the "best" method according to your goal only.

If your goal is to have the loosest possible state, and numerical damping allows that and not contact viscous damping, then numerical damping is the best method. Period.

I'm not surprised you get different porosities with these two cases: your isotropic compression is probably not quasistatic (like for other users, me at least) and all the oscillations you get during this phase will tend to densify your packing.

You observed that numerical damping is a better trick to damp these oscillations and not densify your packing, then just stick with numerical damping.

Which numerical method you're using to construct your packing is not really important in my opinion, what matters is the state of the packing after generation, and what parameters you'll use for the actual simulation.
(For this second phase, I'd understand you prefer include damping from a physically based visco-elastic contact law rather than using non-existing in nature numerical damping forces)

Revision history for this message
Bruno Chareyre (bruno-chareyre) said : 		#2

The porosity at equilibrium is not single valued, it is a range of possible porosities. Which one is reached along a particular loading path depends on every possible factor in principle, including numerical damping.
As Jerôme suggests, you could introduce numerical daming only in the preparation phase if it helps reaching a particular porosity.
Alternatively, for the same purpose, decrease the upper bound of wall velocity, wallMaxVel iirc (or the growth factor if internalCompaction=True); this will compact more gently then give looser states.
Bruno

Revision history for this message
Anton Gladky (gladky-anton) said : 		#3

Another option is to use restituion coefficient instead
of direct setting of cn and cs parameters.

Best regards

Anton

Revision history for this message
feda (feda.s) said : 		#4

Thanks Jérôme Duriez, that solved my question.

Revision history for this message
feda (feda.s) said : 		#5

Thank you for your suggestions and comments:
-using the restitution coefficient instead of cn and cs doesn't change the result.
-decreasing the maximum strain rate of periodic cell doesn't change enough the result.

Using global damping instead of viscous damping seems the better solution to reach the target porosity in my case!