Compute pair/gran/local and dump

Submitted by tshan on Thu, 12/16/2010 - 10:57

Hi, all

Could someone please tell me how to use the compute pair/gran/local and the dump local command to output pair forces in detail? I do not know how to use them. Actually, after dump the pair forces, it will be possible for me to extend the pizza.py to realize them in Paraview.

Could I add these two lines in the input file?
.....
compute 2 all pair/gran/local force
.....
dump dmpf all local 1000 dump.force c_2 #or c_2[1] c_2[2] c_2[3]
................

Does c_2 mean the force value between two particles? Further, how can I get contact normal?

Thank you.

Best,
Tong

Forums:

JF's picture

JF | Thu, 12/16/2010 - 12:03

Hi tong,

I can try to help you. In your example c_2 is a table with 3 columns (fx,fy,fz) and N lines (N= number of interactions).

Then when you dump c_2[1] you obtain the vector (fx).

To understand this, find this line : if(cpl && !addflag) cpl->add_pair(i,j,fx,fy,fz,tor1,tor2,tor3,shear); in pair_gran_hooke_history.cpp

and look at this function (ComputePairGranLocal::add_pair) in the compute_pair_gran_local.cpp

So dump_local does not save the normal force.

Regards
JF

tshan | Thu, 12/16/2010 - 13:53

Hi, JF

Thanks for your reply.

Then the dump local could only output the value of contact forces(fx fy fz), but how do I know the positions of these contact forces (the x y z of start points and end points)?

I mean if I know the positions of particles which interact with each other on each line, it will be possible to calculate the contact normals while convert the dump to vtk/vtp using python.

B.t.w., could you please let me know how to be into the code? Is there a manual or something that could tell me the whole structure and variables of liggghts or lammps?

Best,
Tong

ckloss's picture

ckloss | Thu, 12/16/2010 - 14:04

Hi Tong, please have a look at the doc for compute pair/gran/local:

compute pair/gran/local command

Syntax:

compute ID group-ID pair/local

ID, group-ID are documented in compute command
pair/gran/local = style name of this compute command
zero or more keywords may be appended

keyword =

pos = positions of particles in contact (6 values)
id = Ids of particles in contact (2 values)
force = contact force (3 values)
torque = torque divided by particle diameter (3 values)
history = contact history (# depends on pair style, e.g. 3 shear history values)

So you can output positions of the two particles, particle ids, components of contact force, and history values

if you use the standard form of the command

compute 2 all pair/gran/local

all these values are output by defaut

>>Actually, after dump the pair forces, it will be possible for me to extend the pizza.py to realize them in Paraview.
That would be great, because this link is currently missing!

Cheers,
Christoph

tshan | Thu, 12/16/2010 - 14:32

Hi, Christoph

Thanks for your reply. I see...

However, I met an error when I use liggghts 1.2.2 beta and there's no problem when I used 1.1. It was like the following:

...
910000 8355 6.4146856e-05 6.75022e-05 0.001
920000 8355 6.5172034e-05 0.00012546572 0.001
930000 8355 8.1417103e-05 0.00014026541 0.001
ERROR on proc 0: Too many neighbor bins
ERROR on proc 1: Too many neighbor bins
application called MPI_Abort(MPI_COMM_WORLD, 1) - process 0
application called MPI_Abort(MPI_COMM_WORLD, 1) - process 1
rank 1 in job 3 tong-laptop_32914 caused collective abort of all ranks
exit status of rank 1: return code 1

It is OK when I was using 1.1 at 930000 but it broke at 1140001, and I was always meeting this kind of error, could you tell me where is wrong with this kind of error exactly?

Too many neighbor bins
This is likely due to an immense simulation box that has blown up to a large size.
The answer in lammps is too simple...

Thanks again.

Best,
Tong

ckloss's picture

ckloss | Thu, 12/16/2010 - 14:40

>>This is likely due to an immense simulation box that has blown up to a large size.
>>The answer in lammps is too simple...
No, this answer is correct! The question is why that happened...

>>It is OK when I was using 1.1 at 930000 but it broke at 1140001, and I was always meeting this kind of error
Probably due to bad dyanmics, check with shorter dt.

Christoph

tshan | Thu, 12/16/2010 - 16:47

Hi, Christoph

The identical example was generated by liggghts 1.2.2 and liggghts 1.1. The problem was that when 1.2.2 was used, it was broken down with "too many neighbor bins", while there was no problem when 1.1 was used. I want to use 1.2.2 because the pair forces are needed to realize force-chain.

So Why the identical example finished in 1.1 could not be finished in 1.2.2?

Thanks!

Best,
Tong

prasantud | Mon, 10/17/2011 - 13:11

Hi Jean,
In calculating contact stress sigma(alpha,beta)

sigma(alpha,beta)=1/V*[sum(i) sum(i neq j)*[r(ij)*fij(beta)/2]]

fij(beta)=fij(beta)_tangential+fij(beta)_normal

If normal force is neglected, will it give valid result for contact stress.
regards,
Samantaray