Coarse-graining

hunger's picture
Submitted by hunger on Wed, 11/04/2015 - 14:06

Hi all!

I was looking into literature to find out more about coarse-graining in granular simulations. I have found the paper "Three-dimensional discrete element models for the granular statics and dynamics of powders in cavity filling - C. Bierwisch Ã, T. Kraft, H. Riedel, M. Moseler (2009)" where the dimensional analysis is explained.

My question is: do I have to adjust the material properties used by the gran model hertz (Youngs modulus, Poisson ratio, coeff of restitution, coeff of friction) if I scale the particles by a certain factor?

In the paper it is stated that the coefficient of restituion should be preserved. I also understood that the Youngs modulus and the Poisson ratio should remain unchanged. This is the reason why I thought that maybe it is possible to do coarse-graining without bigger changes in the LIGGGHTS input script. But I am afraid to miss something important at this point, so I would appreciate any help!

Best regards
Harald

Daniel Queteschiner | Wed, 11/04/2015 - 16:11

Typically, you use the same material parameters. However, that way you do introduce some errors, for example by not considering the collisions between the particles that make up a coarse-grained particle. There are approaches to compensate theses errors, though. To get an idea about the potential errors I suggest to read
Benyahia and Galvin, "Estimation of Numerical Errors Related to Some Basic Assumptions in Discrete Particle Methods", Ind. Eng. Chem. Res. 2010, 49, 10588-10605

Cheers,
Daniel

nasato | Wed, 11/04/2015 - 16:50

Hi Harald,

In principle the parameters should remain the same. However you should pay attention to the regime you are working. This coarse grain model (CGM) presented by Bierwisch is valid in the quasi static regime (have a look at Sebastian Chialvo, Jin Sun, and Sankaran Sundaresan. Bridging the rheology of granular flows in three regimes. PHYSICAL REVIEW E, 85:021305/1–021305/8, 2012 for references in the different flow regimes).

Stefan Radl (S. Radl, C. Radeke, J. G. Khinast, and S. Sundaresan. Parcel-based approach for the simulation of gas-particle flows. In Proceedings of the 8th International Conference on CFD in Oil & Gas,
Metallurgical and Process Industries, pages 124/1–124/10, 2011.) proposes a correction for the CGM in the inertial flow. Basically you can differentiate the flow regimes based on the solid fraction.

I would say the paper from Benyahia and Galvin pointed by Daniel is very interesting, however they use a low solid fraction (30%) which put their flow in the inertial regime and no correction is proposed. In this inertial regime you have basically single particle collisions and Radl proposes that a relaxation in the velocity of the parcels is necessary to account for collisions of particles making up the parcels. In the inertial flow you have multiple collisions and this inter parcel collision effect is not so strong.

Cheers,

Nasato

nasato | Wed, 11/04/2015 - 16:56

sorry, in the last sentence where you read "In the inertial flow you have multiple..." should be "In the quasi static flow you have multiple..."

hunger's picture

hunger | Thu, 11/05/2015 - 09:34

Thanks to both of you for your literature suggestions, I found the papers very interesting! I am dealing with a very dense (quasi-static) flow, so I guess the scaling acc. to Bierwisch should be appropriate for my case. Since my solid flow is mainly driven by gravity and the velocities are in general low and the particle contact is very dissipative (low coefficient of restitution), I guess no further correction is necessary.

@Nasato: Benyahia and Galvin suggested a simple Improvement to the Coarse-Particle Model by adjusting the coefficient of restitution to account for the dissipation of granular energy due to particle collisions within a parcel (Part 3 of the paper).

Another point that comes up with coarse-graining is the influence on the conductive heat transfer between the solid particles and the influence on coupling to CFD (especially convection).

1) Conduction
I did not find a paper so far dealing with it. Maybe here also the phenomena compensate in a certain way? The contact area will increase tue to coarse-graining which could somehow account for the particle contacts within a parcel?!

2) CFD coupling
Benyahia and Galvin are discussiong the momentum coupling. For the interphase momentum exchange they consider rather the particle diameter than the diameter of the parcel. I think the approach here is to catch the general phenomenon due to the particle diameter, but apply the outcome on the surface or volume of the parcel. Would you agree if I say a similar approach could be chosen for convective heat transfer? That means the Reynolds and Nusselt numbers should be calculated according to the particle diameter, but the heat transfer with the parcel surface.

Or maybe the reason I did not find so much literature is that it is not common to use coarse-graining if heat transfer applies?!

Best regards
Harald

hunger's picture

hunger | Mon, 01/25/2016 - 09:07

Hi everyone!

I wanted to come back to this topic again, because I have recently found some source code in LIGGGHTS and CFDEM dealing with coarse-graining. Unfortunately it looks like this feature is not enabled in the default version, although it seems a lot of models are considering the coarse-graining factor (or at least display an error/warning message if they do not).
After modifying 'input.cpp' and 'input.h' of LIGGGHTS a bit I managed to get it running and it looks quite promising. Since this feature is not enabled by default, I wanted to ask if there are certain reasons for that? Is it still being tested for consistency? Other reasons?

Best regards
Harald

ckloss's picture

ckloss | Mon, 01/25/2016 - 23:02

Hi Harald,

you seem to have a tendency to come across things which are under active development :-)
the "coarse-graining" functionality is under development and is thus not fully "promoted" in the release. As you obviously found out, it's easy to "activate" coarsegraining with just adding a few lines, but most models are simply not yet ready for it - this is an area of active research... that's why we leave it rather to experts to try/test it and not fully promote it as feature

Best wishes
Christoph

hunger's picture

hunger | Tue, 01/26/2016 - 09:10

Hi Christoph,

thanks for clarification, I am going to use it with caution and go through the source code of the models I use!

Best regards
Harald

dventuri | Tue, 05/08/2018 - 18:58

Hi all,

I would like to bring this back up to check if the coarse-grain implementation is already part of the default implementation or is it still under the hood?

I did find some mentions of coarse-graining in the online documentation for LIGGGHTS but no dedicated page teaching how to use it.

Thanks.

masahiroT | Wed, 10/30/2019 - 10:33

I want to use coarse graining as well.

Now, input.cpp & input.h were changed so that coarsegraining command can be used.
I tried to calculate the example models, but the result was almost the same as the original calculation. The usage may be wrong.
Please tell me how to use