[sami1991]

Why in most of research papers (all papers I have encountered) used LBM simulation approach for mass transport, charge transport studies in electrode materials (a composite material of two different kind of particles and pores due to packing of these particles)(pores in the size of 50 nm-800 nm range and particle size for solids around 1 microns) instead of CFD approach ?
flow of fluid is multi component.
This problem is similar to any flow study problem in porous packing where pore size is in range of nm scale.
Can we also use CFD in such cases ?
few papers are attached here for your reference where they used LBM approach only.
http://www.sciencedirect.com/science...7877530602266X

https://link.springer.com/article/10.../e2009-01024-8

[sbaffini]

Not an LBM expert here, but it seems to be particularly well suited for multiphase and complex geometry stuff.

Still, there is also a lot of marketing around LBM in these days. This also seems to be connected to the fact that you advance the computation explicitly in time, so it basically scales almost perfectly on multicore architectures. There must be something related to the stability constraint of LBM and their resulting affordability, otherwise, I can't honestly see how is this different from explicit time advancement in classical CFD.

[flotus1]

LBM, when used to simulate fluid flows, is also CFD.
One of the benefits for complex geometries has to do with boundary treatment and meshing. LBM does not need a body-fitted grid to model complex boundaries accurately. It uses a cartesian lattice which is relatively simple to produce even for realistic geometries based on CT imaging.

[arjun]

Originally Posted by sbaffini

Not an LBM expert here, but it seems to be particularly well suited for multiphase and complex geometry stuff.

Still, there is also a lot of marketing around LBM in these days. This also seems to be connected to the fact that you advance the computation explicitly in time, so it basically scales almost perfectly on multicore architectures. There must be something related to the stability constraint of LBM and their resulting affordability, otherwise, I can't honestly see how is this different from explicit time advancement in classical CFD.

The difference with classical CFD is:

The viscosity depends on the explicit time step paramenters. (user can not directly specify viscosity as in classical CFD). Also the type of LBM will decide the viscosity and stability constraints. For example finite volume version demands much smaller time steps compared to classical LBM for the same viscosity and grid sizes.

[sbaffini]

Originally Posted by arjun

The viscosity depends on the explicit time step paramenters. (user can not directly specify viscosity as in classical CFD).

Is this just me, or this really doesn't sound as such a great feature?

[FMDenaro]

My opinion is that LBM is a method more suitable for problems of small dimension or for rarefied flows. When we reach the limit of the continuum, the PDE can be substituted by particle method.
I do not believe the LBM can be superior in the range of validity of the continuum hypothesis, it should provide similar solution of DNS.

[flotus1]

user can not directly specify viscosity as in classical CFD

You make it sound like it was impossible to simulate a fluid with a specific viscosity. Which of course is not true. Since the shear viscosity is clearly defined by the relaxation parameter, it is not that difficult to specify the viscosity directly and let the code choose the relaxation parameter accordingly.

I do not believe the LBM can be superior in the range of validity of the continuum hypothesis, it should provide similar solution of DNS.

Which it does. Where did you read otherwise?

[sbaffini]

Originally Posted by flotus1

One of the benefits for complex geometries has to do with boundary treatment and meshing. LBM does not need a body-fitted grid to model complex boundaries accurately. It uses a cartesian lattice which is relatively simple to produce even for realistic geometries based on CT imaging.

However, it is fair to add that the complex geometry handling is more marketing than anything else. Most commercial LBM codes even market themselves as meshless. Nonetheless, classical CFD immersed boundary codes just use the exact same type of mesh.

[arjun]

Originally Posted by flotus1

You make it sound like it was impossible to simulate a fluid with a specific viscosity. Which of course is not true. Since the shear viscosity is clearly defined by the relaxation parameter, it is not that difficult to specify the viscosity directly and let the code choose the relaxation parameter accordingly.


Which it does. Where did you read otherwise?

Umm, I have written codes related to both types of LBMs (classic and finite volume ones) and I would say that technically you are correct that one can simulate any viscosity he wants but practically it is not possible.

This is infact the biggest issue with LBM and currently too no good method exists. EXA powerflow example had good success but they keep it under the wraps and what tricks they are doing to fight this are not open.

Spend some time with papers from him http://researchmap.jp/read0015025/?lang=english
He spent lots of efforts with this problem.

[arjun]

Originally Posted by sbaffini

Is this just me, or this really doesn't sound as such a great feature?

No it is not. Its pain in A.

[arjun]

Here is a paper for example proposing a solution to that viscosity issue (introduces negative viscosity).

http://www.lib.kobe-u.ac.jp/repository/90000996.pdf


"
Abstract
A new scheme for the finite difference lattice Boltzmann method is proposed, in which negative viscosity term is introduced to reduce the viscosity and the calculation time can be remarkably reduced for high Reynolds number
flows. A model with additional internal degree of freedom is also presented for diatomic gases such as air, in which an additional distribution function is
introduced. Direct simulations of aero-acoustics by using the proposed model
and scheme are presented. Speed of sound is correctly recovered. As typical
examples, the Aeolian tone emitted by a circular cylinder is successfully
simulated even very low Mach number flow. Full three-dimensional sound
emission is also given"

[sbaffini]

The end quote from here:

http://www.scholarpedia.org/article/...ltzmann_Method

"Realizing the full potential of the method, up to engineering standards of accuracy, however, still raises many challenges ahead."

Seems to be quite definitive. Is there any concept as verification in LBM? For example, can we solve a steady poiseuille flow and use some control parameter to drive down the error to the desired accuracy? Is a steady flow actually reachable at all?

[sami1991]

seems like whole discussion diverted from the original query and diverted into LBM vs CFD. I would like to make it more clear and provide my electrode electrode image.(attached)

Researchers have given following arguments while using LBM over CFD in electrodes:-
"application of CFD models to sub-micron length scales is problematic because of non-continuum effects, multiple length scales, complex geometries, heterogeneous reactions at the pore scale and a large number of diffusing species."
In my case there are 5 components : H2, H2O, CO, CO2, CH4

So now, can we use CFD fluent solver to solve such problem or LBM is the only way ?
Does LBM has any commercial (free or paid) solver like fluent for CFD?

[Blanco]

The advantage of LMB in submicron lengthscale is present, you have to assess in your application if the knudsen number is still acceptable for a CFD approach, otherwise you have to go on with LBM only.

As far as commercial LBM SW are concerned I would cite PowerFLOW and XFlow, even if the two are not using the same identical approach. If I remember well PowerFlow approach brings to something similar to a VLES while XFlow should bring to a LES, bit I'm not an expert on this.

Sent from my HUAWEI TAG-L01 using CFD Online Forum mobile app

[Blanco]

I would add I don't see other points cited by the researchers as something that leads to LBM instead of CFD...you can solve heterogeneous reactions, multiple length scales in complex geometries in CFD as well, including various diffusing species (in my combustion analyses I have up to 150 species...)

Sent from my HUAWEI TAG-L01 using CFD Online Forum mobile app

[FMDenaro]

Originally Posted by Blanco

The advantage of LMB in submicron lengthscale is present, you have to assess in your application if the knudsen number is still acceptable for a CFD approach, otherwise you have to go on with LBM only.

As far as commercial LBM SW are concerned I would cite PowerFLOW and XFlow, even if the two are not using the same identical approach. If I remember well PowerFlow approach brings to something similar to a VLES while XFlow should bring to a LES, bit I'm not an expert on this.

Sent from my HUAWEI TAG-L01 using CFD Online Forum mobile app

I agree...As I wrote above, the problem is not in CFD but in the PDE written for the continuos model. Assuming the mean free path of order 10^-9 m or similar, you have to compare the characteristic scale of your problem

[sami1991]

That was very helpful, thank you for the answer.
I have explored more into this, and found Knudsen number in the range of 0.6 to 5 for my system. Do you think, CFD can be done for such system ?
One research paper have mentioned this kind of flow in the transitional regime.

Originally Posted by Blanco

The advantage of LMB in submicron lengthscale is present, you have to assess in your application if the knudsen number is still acceptable for a CFD approach, otherwise you have to go on with LBM only.

As far as commercial LBM SW are concerned I would cite PowerFLOW and XFlow, even if the two are not using the same identical approach. If I remember well PowerFlow approach brings to something similar to a VLES while XFlow should bring to a LES, bit I'm not an expert on this.

Sent from my HUAWEI TAG-L01 using CFD Online Forum mobile app

[Blanco]

Originally Posted by sami1991

That was very helpful, thank you for the answer.
I have explored more into this, and found Knudsen number in the range of 0.6 to 5 for my system. Do you think, CFD can be done for such system ?
One research paper have mentioned this kind of flow in the transitional regime.

You're welcome!
Yes transitional regime is characterized by Knudsen number in the range 0.1 - 10 https://www.cfd-online.com/Wiki/Flow_regimes

As you can see from the link, the continuum assumption holds for very low Knudsen number (<0.001). You are therefore in a "mid-land" where different approaches are actually used. I'm not an expert of these, but to cite few of them there are:
- high order slip flow methods
- diffusion modified NS eq
- Burnett eq
- molecular dynamics
- Boltmann methods

Considering this, I would suggest to move on with LBM or at least try with diffusion modified NS equations (if I'm not wrong in the continuity equation there is an additional source term which accounts for mass diffusion).

Hope this could help