A student of mine improved the public normal model hysteresis because it seems not to be correct in all situations
(delta_max was not limited, so the k2 stiffnes was never used in compacting case (only in relaxing)).
See this plot for a comparison between new and old model for a two-particle-contact:
comparison_liggghts3.png
script: two.lmp
A patch can be found here:
http://www.richtisoft.de/transfer/normal_model_hooke_hysteresis.patch
There seems to be a problem in Multicore usage for original and new model, particles are jumping arround unexpectedly.
(delta_t=2.5e-5 small enaugh for D=10..22 mm in my opinion).
The master theses (german) can be found here: http://www.richtisoft.de/transfer/master_thesis_hysterese.pdf
A good source for all the theroy arround Luding and Edinburgh-Elasto-Plastic Adhesion (EEPA) contact Models is the PhD Thesis of J.P. Morrissey. It can be found here: https://www.era.lib.ed.ac.uk/bitstream/1842/8270/1/Morrissey2013.pdf
I would like to implement the EEPA (nonlinear-hysterese-model, Page 68 in Morrissey2013) normal model in LIGGGHTS, but need first to know where the difference between singlecore and multicore operation could come from.
See this videos for what I mean:
singlecore:
http://www.richtisoft.de/transfer/hysteresis_sc.mp4
multicore:
http://www.richtisoft.de/transfer/hysteresis_mc.mp4
Thanks to all the developer who made LIGGGHTS3 so much better to read and extend to own needs !
Best,
Christian.
| Attachment | Size |
|---|---|
 comparison diagramm | 52.62 KB |
 input script 2-particle-contact | 2.49 KB |
ckloss | Wed, 09/17/2014 - 10:42
Hi Christian,
Hi Christian,
thanks for sharing this! I'll try to have a look at your code in the frame of the next release!
Best wishes
Christoph
aaigner | Wed, 09/17/2014 - 12:30
Older version
Hi Christian,
thanks for sharing this, also from me.
You are totally right that k2 was only used for the un-/reloading part. That is because I used an older work from Luding (Luding, S. (2008). Cohesive, frictional powders: contact models for tension. Granular Matter, 10(4), 235–246. doi:10.1007/s10035-008-0099-x). Your version with the limited k2 for high impact velocities is a newer, improved version. I will implement your version during the next release.
The problem with the multicore simulation is somehow wired... it is on my ToDo-List now.
Best wishes
Andreas
sirop | Fri, 09/26/2014 - 04:31
definition of effective radius
string 98: double reff=ri*rj/(ri+rj);
has to be written as: double reff=cdata.is_wall ? cdata.radi : (ri*rj/(ri+rj));
aaigner | Tue, 09/30/2014 - 09:33
Thanks..
Thanks for the note! It will be fixed in the next release.
ckloss | Thu, 10/30/2014 - 18:17
Hi Christian,
Hi Christian,
for this release we didn't have enough time to pick it up, but Andreas is working on it and will probably contact you soon...
Will be in the next release (3.0.5)
Best wishes
Christoph
Rahul | Fri, 08/28/2015 - 09:49
Hi Christoph and Christian,
Hi Christoph and Christian,
Was this bug fixed with multi-core? I am using the recent update and I see this unexpected jumping around of particles.
Regards,
Rahul
benkla | Sat, 07/16/2016 - 13:35
History value
Hello,
in my opinion the multi-core problem is related to the newton flag of the history value - the maximum overlap should always be positive.
Set Line 71 to:
history_offset = hsetup->add_history_value("deltaMax", "0");
Tested with 1...4 cores.
Regards,
Benedikt
aaigner | Tue, 07/26/2016 - 10:03
good job
Hello Benedikt,
Thanks for that hint. You are absolutely right, the newtonflag is wrong. I am going to fix it in the master repo in a minute, thus this bug should be gone in the next release.
Regards,
Andreas
masood | Wed, 03/01/2017 - 09:04
Offset value
Hi,
According to Luding's [2008] model (the schematic diagram of force-displacement), there should be an offset value of (-f0) for all branches of hysteresis forces, which is missed in the algebraic equation. Lack of this offset makes the contact force independent of particles surface energies. The f0 is easily calculated based on the JKR pull-off force and another input needs to be defined to get the pairwise surface energies of particles.
Best,
Masood