Hi ,
I tried to use hybrid granular molecular to simulate multisphere systems, the atoms are successfully read, and I used fix rigid infile to define the center of mass, and moments of inertia. Initially, the non-spherical bodies are uniformly distributed in the simulation box, with a volume fraction of about 0.1, and the first few thousand time steps are running fine (time step = 1e-6), however, the simulation box will increasing after several thousand of time steps, and I can see the total amount of atoms are decreasing.
I monitored the Pzz, and it's increasing rapidly while I also see the simulation box (vol) increases dramatically. Also, I checked the animations, and this problem occurs when the volume fractions of these non-spheres bodies gets larger, say 0.3 ? BTW, I've tried to reduce the time step to 2e-7, but the problem still occurs almost at the same time ( ie. number of time steps x time_step)
The following is my input script, any hint to fix this problem?
#######################################################
....
atom_style hybrid granular molecular
atom_modify map array
comunicate single vel yes
boundary p p m
read_data data.txt
....
neighbor 0.001 bin
neighbor delay 0
neighbor exclude molecular all
.....
pair_style gran model hertz tangential history
pair_coeff * *
.......
fix 1 all rigid molecular infile rigiddata.txt
.......
#################################################
Here is the section of monitoring the pzz and vol
Step Atom Pzz Volume
1 43248 307.10302 1.5e-05
...
6000 42033 28702996 0.00022867619
Any help will be highly appreciated, Thanks!
Best
richti83 | Fri, 05/08/2015 - 12:03
check your boundary
I guess for higher solidfraction there is a huge overlap between two bodies which leads to a high velocity and so the atom is moving fast in z direction. As your z boundary is m it will increase to |zmax| of this particle.
Any chance to make a primitive wall at top and bottom of your simulation domain to avaid that particles travel farther than necessary ?
cheng1988sjtu | Fri, 05/08/2015 - 16:08
Hi RICHTI83,
Hi RICHTI83,
Thanks for your reply ! I forgot to include the line that I define the top and bottom boundary, in which the primitive boundaries are already used:
############################################################
.......
fix zwalls1 all wall/gran/hertz/history primitive type 1 zplane 0.0
fix zwalls2 all wall/gran/hertz/history primitive type 1 zplane 0.3
.......
##############################################################
Do you mean that I should use 'boundary p p f' ? Or what else can we do to avoid the excessive overlap between the rigid bodies? Thanks!
cheng1988sjtu | Sat, 05/09/2015 - 03:37
sorry, When I reduce the
sorry, When I reduce the Young's modulus, the problem still occurs, except that it occurs at a much later time. Any hint to resolve this problem?
cheng1988sjtu | Fri, 05/08/2015 - 22:13
Hi
Hi
Just in case you are interested, reducing the Young's modulus seems to fix the problem, but a reasonable Young's modulus is still desirable.
richti83 | Sat, 05/09/2015 - 17:09
Young's modulus
but a reasonable Young's modulus is still desirable
What is in your term reasonable ? You should know that you can not use the real Young's modulus of your material because this would need a very very small timestep.
Normaly we use something between Y=5e6 and Y=5e7 for DEM with a timestep in a range from 1e-4 to 1e-6. We calibrate the material in the lab and make comparison simulations with different friction settings to get the same behaviour (e.g. angle of repose) in the sim with reduced Young's m.
Maybe you should read some basic DEM literature about critical timestep, reducing young's modulus, friction and material calibration.