I'm not sure if this is intended behavior, but it seems that the coefficient of restitution ("e") specified in the input file is *not* the same as the actual measured ratio of the velocity of a particle after/before a collision.
The restitution coefficient for a collision can be defined as:
e = vel_{after} / vel_{before}
But after running some simple tests with the LIGGGHTS normal model hertz, I have found the "output" restitution coefficient (measured by the velocity ratio) to be greater than the "input" restitution coefficient in the input file. Although the two values do converge at e = 1.0, which is expected.
Could the developers provide some insight into this?
I ran into this issue because we are looking into adding a kind of viscous damping (similar to the one in normal model hooke), but with a different parameterization.
| Attachment | Size |
|---|---|
 in.test_.txt | 2.68 KB |
 plot_ein_vs_eout_hertz.png | 23.7 KB |
ckloss | Mon, 04/25/2016 - 21:48
Hi Sam,
Hi Sam,
thanks a lot for posting this! Typically the simulated coeff of restitution should not deviate more than a couple % from the expected one, at least for e > 0.2
Did you check the time-step size with fix check/timestep/gran? Anyway,I 'll have a quick look at your case before the next release!
best wishes
Christoph
sbateman | Tue, 04/26/2016 - 18:50
Hi Christoph,
Hi Christoph,
I didn't check the timestep with fix check/timestep/gran, but the input file has a timestep of 1e-8 sec, which I would expect to be small enough to run a simulation with 0.5 mm sand grains.
ckloss | Wed, 05/11/2016 - 22:10
Hi Sam,
Hi Sam,
the test you posted is not done properly. You're not integrating the 2nd particle (so it does not move), but the mass of the particle is finite - so this is a contradiction. You should either use a fix freeze (which will assign infinite mass to the particle not moving) or apply the negative velocity to particle 2 and integrate it as well
For the test as you posted it, the output is
e specified: 0.8
e obtained: 0.85something
I modified it and assigned the negative velocity to the 2nd particle and obtained:
e specified: 0.8
e obtained: 0.800355054438655
I am sure for the other cases in your curve it will be similar. Just be aware that the correlation gets worse for smaller coefficients of restitution, so you might see a little deviation there
Best wishes
Christoph