boxes with torsional moment

[Christian Jakob]

Hi all,

I thought about extending boxes ability by including torsional moment and box-box interactions. And I want to know from you, if this a good idea.

The reason is the following: I want the model more realistic sand grains. I can create complex shapes by using clumps, but there are 2 disadvantages. First, the number of spheres is getting too high, which corresponds to high calc. times. Second, by clumping just spheres, there are no real egdes, just rounded ones. To override this I want to create clumps consisting of one sphere and one or more boxes inside the sphere. Only the edges of the box are outside the sphere, so one can create an angular particle. With one box maximal eight edges can be created.

For implementation, boxes must be extended by following abilities:

- boxes can rotate (torsional moment must be implemented)

- boxes can interact with each other (Ig2_Box_Box_ScGeom will be needed)

- all interaction types must be handled:
 * box vertex - box plane interaction
 * box vertex - box edge interaction
 * box vertex - box vertex interaction
 * box vertex - sphere interaction ----- see figure below
 * box edge - box plane interaction
 * box edge - box edge interaction
 * box edge - sphere interaction
 * box plane - box plane interaction

example box vertex - sphere interaction:

```````````````````````````******
````````````````````````*''''''''''''''''' *
`````````````````````*'''''''sphere''''''*
```````````````````*'''''''''''''''''''''''''''''''''''*
################'''''''''''''''''''''''*
___________________*''''''''''#'''''''''''''''''''''''*
_____________________*''''''#''''''''''''''''''''*
________box___________ *'#''''''''''''''' *
_________________________#'******

What do you think about this? I know it will be a lot of work, but in the end it will be a very big extension in possibilities of DEM simulations with yade (not just angular particles!).
Are there other DEM codes, where something similar it is implemented?
What about the details of implementation? Where do you see problems in coding?

Christian.

[Christian Jakob (jakob-ifgt)]

oops, the figure does not look good after submit, but you know what I mean ...

[Bruno Chareyre (bruno-chareyre)]

A very interesting work program!
Clumping spheres and boxes would be awesome.

What we have:
- boxes can rotate already.
- 6D version of box-sphere interactions is almost here since we have 6D sphere-sphere and box-sphere interactions (or let's say plane-sphere) are in the end represented as sphere-sphere contacts. It's only a matter of additional functors, but no real addition in terms of equations.

What we don't have (at all):
- box-box interactions
- box-sphere on edges or vertices

For box-box interactions, there is a serious background in 3-DEC from itasca, where arbitray polyhedral shapes are modelized. It should apply to boxes.

My only question is if it is really necessary to modelize angular vertices instead of rounded shapes. Angularity is causing major geometrical trouble, and in reality surfaces are never completely angular. Then the other option would be to have Minkowsky sum of boxes and spheres (or rounded boxes, let's say).
Somehing like this:
http://www.google.fr/imgres?hl=en&safe=off&sa=X&biw=1438&bih=801&tbm=isch&prmd=imvns&tbnid=I4ZhvDXNCA5--M:&imgrefurl=http://liris.cnrs.fr/m2disco/rapport_2007_en.html&docid=EWZqGLpwGDTYOM&imgurl=http://liris.cnrs.fr/m2disco/morpho.png&w=1252&h=596&ei=TOuHUILiJPTB0gXpyoCABA&zoom=1&iact=rc&dur=504&sig=112569331993381553393&page=1&tbnh=124&tbnw=260&start=0&ndsp=24&ved=1t:429,r:19,s:0,i:126&tx=186&ty=38

In this case you would clump spheres and rounded boxes. François Keib is currently working on something a bit similar (toward "rounded" triangulated surfaces), so it would be good to speak with him if you go for the "rounded" way.

[Bruno Chareyre (bruno-chareyre)]

I forgot to say: this is clearly not a bug report. Please use launchpad consistently.
It could have been a question, or even better a mail directly to the yade-dev list.
I'm closing the bug, please start the discussion somewhere else.

[Christian Jakob (jakob-ifgt)]

> I forgot to say: this is clearly not a bug report. Please use launchpad consistently.

I know, but I like the possibility of sending attachments, as far as I know this neither possible in launchpad questions nor in mail to yade-dev ...

I tried to understand how to calculate torsional moments for clumps. Please have a look at the pictures in the attachment. There two clumps are shown and I do not understand how their torsional moments can be calculated. In the first picture there is box clumped with a "half sphere". In the second picture you can see a box inside a sphere, like I want to model my sand grains in future. How is torsional moment calculated in this cases?
Can this be simulated with RotationEngine already?

> For box-box interactions, there is a serious background in 3-DEC from
> itasca, where arbitray polyhedral shapes are modelized. It should apply
> to boxes.

Hm, I think in the case of cubids (boxes) it should be much easier, than a general polyhedral shape.

> My only question is if it is really necessary to modelize angular vertices instead of rounded shapes.

If I want to model rounded shapes, I could clump spheres, as it is already possible in Yade ... but this is also a nice idea.

[Bruno Chareyre (bruno-chareyre)]

I'm not 100% sure but I think you can attach files to yade-dev. Give it a try please.