coarsegraining command¶
Syntax¶
coarsegraining cg_factors args
- cg_factor = coarse-graining factor to be applied (dimensionless)
- args = model_check
model_check = error, warn or off
Examples¶
coarsegraining 2.0
coarsegraining 4. 2.
coarsegraining 3 model_check warn
Description¶
Coarsegraining is a methodology to reduce the computational effort of a DEM calculation by scaling up the particle size by a coarse-graining factor (cg_factor). Thereby, the model size decreases by factor cg-factor^3.
However, changing particle size will change the physics of the model. On the particle scale, this change is obviously inevitable. However, in many cases model constants can be adapted so that on the bulk scale the change of the physics caused by the coarse-graining model is acceptable. See e.g. (Radl) and (Bierwisch) for details.
If more than one coarse-graining factor is entered, type-specific coarse- graining is activated. So every atom-type in the simulation may have a different coarse-graining factor.
Some of the models in LIGGGHTS(R)-INL support scaling of the model constants to counteract this change in physics on the bulk scale. Moreover, some commands (such as insertion commands or the neighbor command) re-scale some of their length scales. Using model_check = error will throw an error if a model/command does not yield constistent results with coarse-graining. Using model_check = warn will issue a warning if a model/command does not yield constistent results with coarse-graining. model_check = off will suppress all warnings and is only intended for long CFDEMcoupling simulations.
Model overview:
This table provides an overview of how different models behave when coarsegraining is used. Some models scale lengthscales, other models scale model parameters to make results more coarse-graining consistent. Models that are not mentioned here are not affected by using the coarsegraining command
| create_atoms | not supported | |
| fix addforce | not supported / inconsistent | |
| fix drag | not supported / inconsistent | |
| fix dragforce | downscaling of A, Re; upscaling of dragforce | |
| fix efield | not supported / inconsistent | |
| fix heat/gran/conduction | not supported / inconsistent | |
fix mesh/surface/stress |
Finnie wear model | inherently compatible |
| fix particletemplate/sphere | radius -> cg_factor*radius | |
| fix particletemplate/multisphere | radius -> cg_factor*radius | |
| fix spring | not supported / inconsistent | |
| fix spring/rg | not supported / inconsistent | |
| fix spring/self | not supported / inconsistent | |
| fix viscous | not supported / inconsistent | |
| neighbor | skin -> cg_factor*skin | |
| pair gran/* | any cohesion or rolling friction model | not supported / inconsistent |
| pair gran/hertz/history | base model | inherently compatible |
| pair gran/hooke/history base model | base model | inherently compatible |
| pair gran/hooke base model | base model | inherently compatible |
| pair gran/hertz/history/stiffness | not supported / inconsistent | |
| pair gran/hooke/history/stiffness | base model with damp_massflag=0 | kn -> cg_factor*kn, gamman -> cg_factor^2*gamman |
| pair gran/hooke/history/stiffness | base model with damp_massflag=1 | kn -> cg_factor*kn, gamman -> gamman/cg_factor |
| pair gran/hooke/stiffness | base model with damp_massflag=0 | kn -> cg_factor*kn, gamman -> cg_factor^2*gamman |
| pair gran/hooke/stiffness | base model with damp_massflag=1 | kn -> cg_factor*kn, gamman -> gamman/cg_factor |
| set | diameter -> cg_factor*diameter |
Warning
Even if a model does not support proper coarse graining, in many cases it will still be possible to calibrate the model constant manually to capture the bulk scale physics with a coarse-grained model.
Restrictions¶
General warning: Coarse graining is an approximative method and will _never_ deliver exactly the same results than a fine-grained simulation.
One natural restriction is that the ratio of geometry length scale to particle length scale should be >> 1 when coarse graining is applied.
See description above the fact that not all LIGGGHTS(R)-INL models will deliver consistent results with coarse-graining ratios > 1.
The coarsegraining command needs to be at the beginning of an input-script. It needs to be defined before any fixes or the definition of a simulation box.
Default¶
cg_factor = 1 (no coarse-graining applied)
(Radl) S. Radl et. al., PARCEL-BASED APPROACH FOR THE SIMULATION OF GAS-PARTICLE FLOWS, Proc. 8th International Conference on CFD in Oil & Gas, Metallurgical and Process Industries, 2011
(Bierwisch) Journal of the Mechanics and Physics of Solids, Volume 57, Issue 1, January 2009, Pages 10ff