time step vs material properties

Submitted by CARLOS AVILA on Tue, 10/12/2010 - 03:24

Hi,
Could some body please, let me know what causes the following error message:
*****************************************************************
Setting up run ...
Memory usage per processor = 10.1914 Mbytes
Step Atoms KinEng 1 Volume
1 2000 0.0047170791 0 0.0015
......
.........
37000 2000 0.0094914152 0 0.0015
38000 2000 0.0094929527 2.9568818e-05 0.0015
39000 2000 0.0097540535 0.00030106194 0.0015
40000 2000 0.0101535 0.00085611282 0.0015
41000 2000 0.012838727 0.0024416766 0.0015
ERROR on proc 0: Failed to allocate -1915839920 bytes for array neigh:binhead
*****************************************************************

I think it has something to do with the time step size and the Young modulus that I am giving as Input.
#Material properties required for new pair styles

*****************************************************************

fix m1 all property/global youngsModulus peratomtype 200e9

fix m2 all property/global poissonsRatio peratomtype 0.31
fix m3 all property/global coefficientRestitution peratomtypepair 1 0.6
fix m4 all property/global coefficientFriction peratomtypepair 1 0.5

#New pair style__________________________________________________________________
pair_style gran/hertz/history 1 0 #Hertzian without cohesion
pair_coeff * *

timestep 1e-6
*****************************************************************
I am working with steel properties and SI units. I computed this time step based on Silbert paper, it depends basically on particle diameter and gravity. But when I set E=200e9[Pa] the simulation time increases from minutes to hours, and in most of the cases gives the error message given above.

Any response will be truly appreciated.
Regards,
Carlos

Forums:

ckloss's picture

ckloss | Tue, 10/12/2010 - 08:24

>>I think it has something to do with the time step size and the Young modulus that I am giving as Input.
yes - the error means that your simulation "exploded". Whenever an explicit time integration is used, you have to care about the time step size

>>E=200e9[Pa]
This is very high, most people use lower values, but it depends on what you want to picture with your simulation.
The values in the examples are really very low, in order to be able to run the simulations quickly.

I recommend to use the fix check/timestep/gran command, as it is done for some of the examples that come with the tarball - it will warn you if the time-step is too high according to Hertz and/or Rayleigh criterion

Christoph