about skin distance and import stl

Submitted by tju_coastlab_mq on Sun, 06/08/2014 - 10:55

Hi, everyone here.
i have several quite basic questions:

I) what is the skin distance for and how i set skin's value
I have read the manual and searched in FORUM for information about 'skin distance' (cutoff distance = force cutoff distance + skin distance).
i am wondering, when just simulating 'granular' type, what is the skin distance or cutoff for? (i know it's for neighbour list building to calculate non-bonded forces when Molecular Dynamics). because in my opinion, the granular particles, not like atoms or moleculars, will just interact with each other when they contact (collision), and when they donot contact, there will be no force at all, am i right? i cannot figure out the meaning of the skin distance in granular simulation, can anyone tell me?
The skin distance have any contact with the particle radius? how can i set the appropriate value before a simulation?

II) import stl
i need a quite complex wall in my simulation, so i have to use an stl file.
i'm wondering the mesh size of the stl file have any contact with the radius? because when i save the stl from AutoCad, there seems no options for me to choose the size of triangle mesh.

Regards,
Mq

Forums:

tju_coastlab_mq | Sun, 06/08/2014 - 21:47

And when using 'pair_style gran/...', what value will 'force cutoff distance' be set?

Regards,
Mq

msandli | Wed, 06/11/2014 - 17:59

I am also interested in what the skin distance and force cutoff distance refer to. Are there any books/articles about this? Information on characteristics of discrete element method simulations is hard to come by.

aaigner's picture

aaigner | Thu, 06/12/2014 - 09:37

Mq pointed out correctly that the cutoff distance = force cutoff + skin and that it is important for the neighbour list build. So let's have a closer look:

  • The force cutoff: For the granular style -- since their are no non-contact forces (except special cohsion models)-- it is just the sum of the two particle radii ... cut_force = r_i + r_j.
  • The skin cutoff enlarges the cell size for the neighbour list build and this is good. Therefore, the particles may move a little bit and the neighbour list is still correct. Without the skin factor we would have to rebuild the neighbour list each timestep, which is quite expensive. If the skin factor is too large, you save the neighbour list build, but you have to check many possible contact partners each time step, which is quite expensive too.
  • In my case the best pratice is to choose the skin factor of the order of the particle radius. For polydisperse simulations this is more tricky. :-/

Hopefully I could answer your questions.

Bests
Andreas

tju_coastlab_mq | Thu, 06/12/2014 - 20:23

Hi, Andreas. Thank you SO MUCH about the explanation : )

And i have another question: are there any principles about choosing proper STL mesh element size ?(may be associated with particle radius?)
Because when i search information here, it seems unnecessary to set dense grid, which is quite expensive.

regards,
Mq

msandli | Thu, 06/12/2014 - 19:04

Andreas,

Thanks you VERY much for that explanation! Because of that, I was able to change my skin distances and bin size, and my (although small) simulation cut it's time down by a factor of nearly 6