In my continuing work on thermal expansion for pebble beds, I've begun looking at the evolution of the average force magnitude on each pebble.
I took mean values of the per-particle force and plotted them in time. The plot came out with these periodic spikes in the force. When I looked at the data, the spikes in force were occurring on exactly every 10th snapshot. I'm fairly certain the spikes are not physical. My conclusion is that, because I dump (and print thermo data) every N steps but save a restart every 10*N, that it's logical to think the restart and dump actions are interfering with each other somehow. I included a plot showing this maybe more clearly than I'm saying it.
I read the dump and restart docs and don't see any reason how they could affect each other, but I can't come up with a better reason.
Any ideas?
| Attachment | Size |
|---|---|
 force_plot.png | 10.44 KB |
ckloss | Mon, 03/25/2013 - 15:25
Hi jtvanlew,
Hi jtvanlew,
how often were neigh lists re-built in your simulation? Potentially some operationis missing in the thermal expansion model at the moment on re-neighboring steps. If you can reproduce the issue on a small-scale case, I can have a look.
Cheers
Christoph
jtvanlew | Thu, 04/04/2013 - 23:11
Actually, I've never modified
Actually, I've never modified the "neigh_modify delay 0" from the original example file I used as a template to begin my script. To be honest, I just breezed over that setting as I assumed it was OK since the simulation was running but never really spent time to understand what its function was.
I've tried duplicating this error with small modifications of other example scripts and so far haven't been able to. Maybe the source of the error is something other than what I thought since I can't recreate it. If I run into it again, I'll try to sniff out the root.
regards,
Jon