Shrink-wrap boundaries

Submitted by tjleps on Tue, 01/16/2018 - 22:45

I'm having trouble implementing shrink wrap boundaries. I keep getting an "illegal simulation box" when I use s boundaries, but it works for m, p, and f boundaries. Does anybody have any idea what's going wrong? Here is my script:

atom_style hybrid granular dipole
boundary f f f
newton off
atom_modify sort 100 0.6

communicate single vel yes
units si
processors 1 4 1

region reg block -1.5 2 -.2 .2 0 2 units box
create_box 1 reg
neighbor 0.005 bin
neigh_modify delay 0

hard_particles no

fix m1 all property/global youngsModulus peratomtype 1.e9
fix m2 all property/global poissonsRatio peratomtype 0.278
fix m3 all property/global coefficientRestitution peratomtypepair 1 0.2
fix m4 all property/global coefficientFriction peratomtypepair 1 0.6
fix m6 all property/global cohesionEnergyDensity peratomtypepair 1 10
fix m7 all property/global coefficientRollingFriction peratomtypepair 1 0.4

pair_style gran model hertz tangential no_history cohesion sjkr rolling_friction cdt
pair_coeff * *

timestep 0.00002

fix boxwall1 all wall/gran model hertz tangential history primitive type 1 xplane -1.5
fix boxwall2 all wall/gran model hertz tangential history primitive type 1 xplane 0
fix boxwall3 all wall/gran model hertz tangential history primitive type 1 xplane 2
fix boxwall4 all wall/gran model hertz tangential history primitive type 1 yplane -.2
fix boxwall5 all wall/gran model hertz tangential history primitive type 1 yplane .2
fix boxwall6 all wall/gran model hertz tangential history primitive type 1 zplane 0

fix pts1 all particletemplate/sphere 15485863 atom_type 1 density constant 7750 radius constant 0.01
fix pdd all particledistribution/discrete 86028157 1 pts1 1.0

group nve_group region reg

region insbox block -1.49 -0.01 -0.19 0.19 0.01 1.99 units box
fix ins nve_group insert/pack seed 86028121 distributiontemplate pdd &
insert_every once overlapcheck yes volumefraction_region 0.3 region &
insbox ntry_mc 10000

fix 1 nve_group nve/sphere

fix 2 all gravity .1422 vector 0.0 0.0 -1.0
#fix mag all magnetic 0 0 0

fix ts all check/timestep/gran 2000 0.1 0.1
#compute 1 all erotate/sphere

dump dmp1 all custom 5000 post/dump*.liggghts id type x y z vx vy vz
dump dmp2 all image 5000 post/dump*.ppm type diameter view 90 90

run 400000 upto
write_restart post/boxavalanche.restart

paul | Wed, 01/17/2018 - 23:51

What is wrong with using m? Do you not have a region in which you can always expect at least one of your particles to be?

If you rely on s, I'd start debugging here because here the error is thrown:
https://github.com/CFDEMproject/LIGGGHTS-PUBLIC/blob/7c32dc4e5ef4c7d5d0a...
Modify the error messages to check which block is the cause, print some of the variables and try to find out what upsets him ;)

- Paul

tjleps | Thu, 01/18/2018 - 07:01

Paul,
The problem with m boundaries is that there's a lot of empty space that is making imbalanced core loads when running the simulation on more processors.