I think this is issue is related to the previous post by Zhenli, but I don't want to hijack that forum post with my own issue. I am trying to simulate a direct shear test with mesh walls and 11700 multisphere particles (log file attached). The multisphere particles are actually spherical consisting of one sphere of smaller diameter within another sphere since multisphere templates of only one sphere are not allowed. The simulation runs to completion without an error being thrown, but partway through the simulation the total number of particles suddenly drops. This drop doesn't happen gradually, it just instantly changes to a smaller number one or more times during shearing. Reviews of simulation snapshots showed that the particle assembly appears to inwardly implode in the middle of the direct shear box walls. I tried running this simulation with a time step 10 times slower, and the same thing still happened. The particle disappearance does not occur when I reduce the total number of particles in the simulation by a factor of 10 or when I use only 2 processors. When I increase the total number of particles and use 24 processors, the particle disappearance does occur. I also repeated this same simulation with regular spheres and no multisphere commands and no problems arose. In fact, the results for the multisphere and regular-sphere simulations are the exact same up until the point where the multisphere particles disappear. I'm thinking this issue is either a bug or an issue with my neighbor/skin settings. I would really appreciate any help figuring out what the issue is so I can finish my degree and graduate :-) .
| Attachment | Size |
|---|---|
 Log file in which particles randomly disappear (see changes in total number of atoms) | 556.37 KB |
| inputfile.txt | 21.83 KB |
estefan31 | Sat, 03/25/2017 - 03:33
I think I found the problem.
I think I found the problem. I had a loop with the command "run 10000 pre no post no" so I wasn't updating the neighbor lists during shearing.
kashif saeed | Thu, 05/04/2017 - 16:32
MULTISPHERE PARTICLES DISAPPEARING IN MIDDLE OF DOMAIN
Hi estefan,
can you help me, I am doing simulation for mixing of multisphere particles but during simulation my particles are lost.
estefan31 | Sat, 05/06/2017 - 02:14
Hi Kashif,
Hi Kashif,
Are you running your simulations across multiple cores? I had a problem because I put a loop that included 'run 10000 pre yes post no'. Within the 10000 cycles, particles would migrate across not one but TWO cores. Since the core it had migrated to had no idea the particle existed, the particle was simply lost. But if I ran only 1000 cycles per loop, the neighbor lists would update more frequently and the particle would no longer migrate to cores that were not tracking it. So basically if you have a loop, try running fewer cycles within that loop and make sure to check if your neighbor lists need to be updated more frequently.