New Features

Submitted by cgrohs on Wed, 05/19/2010 - 14:58

Hi Christoph!

Is there an estimation when the new features (e.g. non-spherical particles, ...) will be available for public?

Best regards,
Christian

Forums:

ckloss's picture

ckloss | Wed, 05/19/2010 - 16:22

Hi Christian,

we are quite far with these features. Generally, anybody who is interested in those new features such as non-spherical particles, the 6dof solver etc.,.. is invited that we join forces for validation and industrial application
Christoph

Knappi | Fri, 11/26/2010 - 17:26

Hi Christoph,

I'm interested in using non-pherical paricles. I want to model wheat kernels and they can be modelled well with two combinded spheres. When will this feature be released?

Greetings Thorsten

alexander.polson | Mon, 11/29/2010 - 17:37

I would also be interested in the use of "clumps".

I understand a clump to be a set of overlapping spheres, with pre-defined inital relative center positions and fixed arbitrary radii, that are constrained relative to each other but able to move together as a unit under laws of Newtonian physics. Is that right?

The use of such clumps would then also bring in the need for supporting functionality, such as:
- a method to "fix/pour/dev" these particles into simulation space
- templates to have definitions for individual shapes that make up a family of different shapes that are called from the particle generator according to user requirements
- a way to account for overlapping volumes in the accurate determination of actual particle masses and moments of inertia
- there might be more requirements

If enough groups want this functionality, can we try to share the cost for these developments? Of course, the more groups, the lower the cost per group.

Is this suggestion meaningful to any other users? Please reply with any ideas.

It also depends on availability of current LIGGGHTS developers for this work. I dont have enough experience to do this, but would be willing to share costs with others, if possible.

Kind regards

Alexander

ckloss's picture

ckloss | Mon, 11/29/2010 - 17:43

>>The use of such clumps would then also bring in the need for supporting functionality, such as:
I agree with all your points and there is one more that will be important:
- true parallelization of integration

>>It also depends on availability of current LIGGGHTS developers for this work.
That's true... we have a couple of aspiring students, but of course it needs some funding before someone can start working on that topic honestly.

>>If enough groups want this functionality, can we try to share the cost for these developments?
Sure - that's how open source works!

I suggest we discuss further details via email

Best regards,
Christoph

htm | Sat, 12/04/2010 - 00:12

Hi Christoph
I have worked with non-spherical particles on a code that me and my master´s adviser worked on a few years ago. You can check the details on the following reference

Non-iterative and exact method for constraining particles in a linear geometry
Horacio Tapia-McClung and Niels Grønbech-Jensen
Journal of Polymer Science Part B: Polymer Physics
Volume 43, Issue 8
15 April 2005
(pages 911–916)
DOI: 10.1002/polb.20383
http://onlinelibrary.wiley.com/doi/10.1002/polb.20383/abstract

Basically we calculate the lagrangian coefficientes needed to constrain the particles on a straight line and we can model long particles of different shapes. Our method is exact (compared to others on which an approximation is obtained for the constrain forces) and efficient as it only involves inverting a finite matrix which can be easily implemented. I would very much be interested in testing this method on LIGGGHTS for a possible implementation, please let me know if you may be interested on further exploring this option.
Sincerely,
Horacio

ckloss's picture

ckloss | Mon, 12/06/2010 - 10:34

>>I would very much be interested in testing this method on LIGGGHTS for a possible implementation
Sure - that sounds interesting

>>only involves inverting a finite matrix which can be easily implemented
I am not sure if this is exactly what you need, but there is a simple Gauss algorithm, MathExtra::mldivide3 and an iterative Richardson method already implemented, FixRigid::richardson.

If you have a question, feel free to contact me

Cheers,
Christoph