Hi all,
I am going to insert particles in a box as follows, but this take too much time ( about 10 days with 120 cores)!
I would be pleased if anyone knows why!
--------------------------------------------------------------------------my input----------------------------------------------------------------
# insert granular particles into channel
echo both
atom_style granular
atom_modify map array
communicate single vel yes
boundary f f f
newton off
units si
region reg block 0.0 7.0 0.0 0.37 0.0 0.02 units box
create_box 2 reg
neighbor 0.00072 bin
neigh_modify delay 0
variable yMp1 equal 70.e9
variable yMwall equal 200.e9
variable cRFp1p1 equal 0.01
variable cRFp1wall equal 0.01
variable cRFwallp1 equal 0.01
variable cRFwallwall equal 0.01
variable pRp1 equal 0.18
variable pRwall equal 0.37
variable cRp1p1 equal 0.6
variable cRp1wall equal 0.5
variable cRwallp1 equal 0.5
variable cRwallwall equal 0.6
variable cFp1p1 equal 0.6
variable cFp1wall equal 0.4
variable cFwallp1 equal 0.4
variable cFwallwall equal 0.8
#Material properties required for new pair styles
fix m1 all property/global youngsModulus peratomtype ${yMp1} ${yMwall}
fix m2 all property/global poissonsRatio peratomtype ${pRp1} ${pRwall}
fix m3 all property/global coefficientRollingFriction peratomtypepair 2 ${cRFp1p1} ${cRFp1wall} ${cRFwallp1} ${cRFwallwall}
fix m4 all property/global coefficientRestitution peratomtypepair 2 ${cRp1p1} ${cRp1wall} ${cRwallp1} ${cRwallwall}
fix m5 all property/global coefficientFriction peratomtypepair 2 ${cFp1p1} ${cFp1wall} ${cFwallp1} ${cFwallwall}
fix m6 all property/global characteristicVelocity scalar 2.
#pair style
pair_style gran model hertz tangential history rolling_friction cdt #Hertzian without cohesion
pair_coeff * *
#timestep, gravity
timestep 5e-8
#fix ts_check all check/timestep/gran 10000 0.1 0.1
fix gravi all gravity 9.81 vector 0.0 -1.0 0.0
#walls
fix xwalls1 all wall/gran model hertz tangential history primitive type 1 xplane 0.0
fix xwalls2 all wall/gran model hertz tangential history primitive type 1 xplane 4.5
fix ywalls1 all wall/gran model hertz tangential history primitive type 1 yplane 0.0
fix ywalls2 all wall/gran model hertz tangential history primitive type 1 yplane 0.37
fix zwalls1 all wall/gran model hertz tangential history primitive type 1 zplane 0.0
fix zwalls2 all wall/gran model hertz tangential history primitive type 1 zplane 0.02
#particle distributions
region ab block 0.0 4.5 0.02 0.03 0.0 0.02 units box
fix pts1 all particletemplate/sphere 1000001 atom_type 1 density constant 2650 radius constant 0.00036
fix pdd1 all particledistribution/discrete 1000001 1 pts1 1.0
fix ins_mesh1 all mesh/surface file mesh/face1.stl type 1
group nve_group1 region ab
fix ins1 nve_group1 insert/stream seed 1000001 distributiontemplate pdd1 nparticles 360000 &
vel constant 0.0 0.0 0.01 particlerate 70000 &
overlapcheck yes insertion_face ins_mesh1 extrude_length 0.005
#apply nve integration to all particles that are inserted as single particles
fix integr1 nve_group1 nve/sphere
#screen output
compute 1 all erotate/sphere
thermo_style custom step atoms ke c_1 vol
thermo 1000
thermo_modify lost ignore norm no
compute_modify thermo_temp dynamic yes
#insert the first particles so that dump is not empty
run 1
dump dmp all custom 100000 post/dump.liggghts_init id type type x y z ix iy iz vx vy vz fx fy fz omegax omegay omegaz radius
restart 10000000 restart1*.bin
run 100000000 upto
unfix ins1
run 110000000 upto
write_restart bed.restart
----------------------------------------------------------------------------------------------------------------------------------
Thanks;
Elham
cstoltz | Wed, 06/10/2015 - 12:39
I would say it is because you
I would say it is because you're running a huge number of timesteps. Your modulus is very large, which necessitates a very small timestep, and hence a huge number of timesteps.
Two other observations:
1) Your domain is very large relative to the volume expected to be occupied by the particles. You could likely take advantage of this to insert particles at a higher rate.
2) Your domain is 7.0 m long in the x-direction, but the walls are set only 4.5 m apart. Thus, you're wasting about a third of your processors on open space.
Regards,
Chris
elham.nasiri | Wed, 06/10/2015 - 20:05
Thanks Chris,
Thanks Chris,
I need a large modulus for simulating of quartz particles.
Also,
1) I created a 2D face with stl. format in x=0 - 4.5 , z=0 - 0.02 and in height 0.15m from bed and used this face for insert-stream command. With using this face, it could not possible to insert particles at a higher rate and the particles were inserted less than requested. I don't know how insert the particles at a higher rate!
2) I am gonig to insert the particles in entire domain but step by step because the measure of particles in various places is different. Even if I define the various commands for insert particles in various places, the commands run as the order in which they are defined. And this don't effect on execution time. Is it true?
Best Regards,
Elham
elham.nasiri | Fri, 06/12/2015 - 07:41
useing of partition command
Hi,
I looked in manual and found "partition" command for dividing the total area into multiple partitions, each partition is assigned a different number processor. But I could not exactly understand how? If anyone knows, please tell me how!
Regards
Elham
JoshuaP | Fri, 06/12/2015 - 09:49
As long as you are not in a
As long as you are not in a very high stress state, the rearrangement of particles defines your behaviour and you can decrease the youngsmodulus. But you should check it maybe first.
regards
Joshua
elham.nasiri | Sat, 06/13/2015 - 07:59
Hi Joshua,
Hi Joshua,
Thanks but I tried the lower youngsmodulus and the correct simulation didn't found. I don't understand the concept of "the rearrangement of particles defines your behaviour and you can decrease the youngsmodulus"! Please explain more.
Thanks
Elham
cstoltz | Tue, 06/16/2015 - 13:33
If all you're trying to do is
If all you're trying to do is insert the particle to generate a packing, you can typically reduce the modulus of the particles to something on the order of 5-10 MPa without significantly affecting the packing arrangement. This will enable you to use a much larger timestep, and thus run the simulation faster.
To insert particles faster, consider upping the initial velocity or changing the insertion region.
Regards,
Chris