What options are available for relaxing a multisphere system? My favorite fix option is nve/limit, but this only works for regular spheres. And I'm running simulations in parallel, so inserting over 100,000 multisphere particles has been very difficult if I use the 'overlap_check yes' option. I'd also prefer to keep my domain as compact as possible, so I want to avoid inserting particles from a very tall height. If there is nothing freely available, does anyone have any recommended code modifications I can implement? I'm not sure how difficult implementing a fix multisphere/limit would be since that may involve fixing 2 different multisphere integration fixes, and I can't unfix one to replace it with another. Although I am considering having to implement a fix multisphere/limit, using write_restart, and then restarting a new simulation with a standard fix multisphere.
Also, I already tried damping the hell out of the system with fix viscous using an extremely high damping factor. That didn't work too well for me.
ckloss | Sun, 04/23/2017 - 12:13
Hi Estefan,
Hi Estefan,
relaxation for MS simulations is really, really tricky. For some (actually many) shapes and most particle-wall overlaps, you will end up with a local energy minimum that shows interlocking of two multi-sphere particles or a ms particle and the wall.
If you want to give it a try, you can modify the integration part in the fix_multisphere.cpp file.
Best wishes
Christoph
estefan31 | Tue, 04/25/2017 - 19:30
I had to get a little
I had to get a little creative for this one. In case anyone else is trying to work through the same thing, my solution is to use a very small contact stiffness (Young's modulus) when the particles are relaxing. This works well for me because I initially do not use gravity and have large overlaps present when I first generate the particles. So if I use a Young's modulus on the order of 1e6 and maybe apply some viscous damping then the particles slowly rearrange without flying out of the simulation domain. Then when the particle overlaps are small enough I apply the final contact properties needed for the rest of the simulation.