silo problem

Submitted by adabubu on Mon, 11/18/2013 - 15:43


my simulation is about silo. Particles are poured into a silo (cuboid 0.58m * 0.875m*2m). Due to highly overlapped atoms are involved, so I used "fix nve/limit" to define the nve intergration. After particles are poured into the silo, i used "wiggled" for getting better packing, however the numerical results of pressures on walls are larger than them from Janssen equation (about 30% larger , then I tried to "unfix nve/limit" and use " fix nve/sphere" after pouring particles, while the pressures on the wall are still larger than them from Janssen equation.
And at same time I also did two other simualtions with a cubic silo (0.6m*1m*2.0m) and a cylidner with 1m diameter. I compared the nuemrical results of pressures on walls with Jassen pressures. And they showed a good match. In those simualtions, I used "fix nve/limit" and " fix mve/mesh wiggle". I felt it is very strange that these three simulations showed such difference when they are compared with Janssen, even though the only difference among them is only the difference of geometry of silo. Anyone could give some some tipps about this? Thanks a lot!


adabubu | Wed, 11/20/2013 - 14:32


I used "insert/pack". The reason I use "fix nve/limit" is that there are more than 250000 particles in my simulations. This simulation is based on an large model experiment. In order to get similar fillheight like in experiment, I tried several methods, e.g using wiggle, unfix nve/limit and fix nve/sphere, however when fillheight in simulations was satisfied comparing with experiment, the wall pressures were very unsatisfied comparing with Janssen theory, the nuemrical results of wall pressure were twice larger than them from Janssen equation. I have tried so many times and so many different silo models whith different geometries, so far for the silos with cylindrical and cubic geometries, the numerical results are basically satisfied (comparing with Janssen theory), while for the silo with 0.857m*0.58m*3m dimension(the geometry used in experiment), the nuemrical results were alway larger than them from Janssen theory(almost twice). It makes me wonder if there is problem in Liggghts. From my experiences using Ligggths, it works fine ( for simple Janssen silo problem) if only few particles are involved (for example less than 50000), once more particles are needed (fro examle my simulations), there will be some problems.
As for the friction, yes, I not only considered friction, also "rolling friction".
I don't know if anyone has ever done similar simulations, right now I am really stucked here and I don't know what's wrong with my simulations. Hopefully someone can help me.


ckloss's picture

ckloss | Tue, 11/26/2013 - 10:00

Hi Xuetao,

that is the d/D ratio for your simulation? If d/D is not <<1, the Janssen equation has to be modified


wanjianjgfeng | Sat, 04/18/2015 - 11:39

How do you calculate the pressure against the wall? I am so interested in the command you used. stress/atom ?
Manys thanks!