[vkrastev]

Hi all,
can someone explain briefly what is the difference between the cellLmited and the cellMDLimited options for the gradSchemes in OF?

Thank you in advance

V.

[vkrastev]

Any replies?

[cnsidero]

Here's a general explanation. Standard gradient limiting, i.e. cellLimited, clips each component of the gradient equally (remember it's a vector). cellMDLimited is a "M"ulti-"D"imensional limiter whereby the gradient is clipped in the direction normal the cell faces.

cellMDLimited should be less dissipative.

A little extra info ... there are two types of limiting: cell and face. Cell limiting determines the limited gradient along a line connecting adjacent cell centers. Face limiting determines the limited gradient on the face itself.

Cell limiting should be less dissipative.

Of course, reducing dissipation has the potential benefit of better accuracy but increases the risk of instability. The limited gradient schemes in OpenFOAM listed from least to most dissipation:

cellMDLimited
cellLimited
faceMDLimited
faceLimited

Disclaimer: My C++ interpretation skills can be suspect at times so if I've stated something incorrect, I welcome anyone else's comments.

If you wanna look yourself, the source code for the gradient schemes can be found in the following directory:

$WM_PROJECT_DIR/src/finiteVolume/finiteVolume/gradSchemes/

[vkrastev]

Thank you for the explanation, this "summary" was exactly what I was looking for.

Regards

V.

[akidess]

Chris, I think your post has some really valuable information, so I took the liberty to put it onto the openfoam-extend wiki:
http://openfoamwiki.net/index.php/Op...guide/Limiters

- Anton

[makaveli_lcf]

Hi all!

I am a little bit confused about face/cellLimited syntax. Let say we have cellLimited scheme with leastSquare scheme:

gradSchemes
{
default cellLimited leastSquares 1.0;
}

What does exactly the number "1.0" mean here? Is it non-orthogonal correction limiting or is it some conformance parameter?

Thank you in advance and have a nice day!

Alexander

[cnsidero]

Originally Posted by makaveli_lcf

Hi all!

I am a little bit confused about face/cellLimited syntax. Let say we have cellLimited scheme with leastSquare scheme:

gradSchemes
{
default cellLimited leastSquares 1.0;
}

What does exactly the number "1.0" mean here? Is it non-orthogonal correction limiting or is it some conformance parameter?

Thank you in advance and have a nice day!

Alexander

The 1.0 is like a switch for the limiter, which can vary between 0 and 1. Setting it to 0 effectively turns off the limiter (i.e. no limiting), setting it to 1 is fully on and anything in between it's partially on. Setting it to 0 will give the best accuracy (because you're not limiting) and setting it to one will give the best stability.

[makaveli_lcf]

Thank you Chris for clarifying my doubts! I also found it from my previous request))) Sorry for double-posting! But now I am sure about this point.
Do you have some experience if *MD* limiting really changes the convergence behavior or improves calculation performance?

[cnsidero]

Originally Posted by makaveli_lcf

Thank you Chris for clarifying my doubts! I also found it from my previous request))) Sorry for double-posting! But now I am sure about this point.

No problem.

Originally Posted by makaveli_lcf

Do you have some experience if *MD* limiting really changes the convergence behavior or improves calculation performance?

I have not performed a careful study investigating the differences between the regular limited and MD limited schemes. Although, now that you mention it, I think it would be a valuable exercise.

My initial feeling would be that it might be tough to find an appropriate test case(s) to look at the differences. For simple problems (i.e. simple, nice meshes), I suspect there differences will be negligible. And more complex problems introduce other uncertainties (turbulence model, etc).

The effect of a limiter shows up most when there are high gradients (e.g. shocks) so a problem involving high gradients would be the best. Perhaps a free surface problem (VOF)?

[makaveli_lcf]

My current problem (solidification modelling) also includes high gradients, so I'll try to get some information comparing 2 calculations with MD or without and report here...

[makaveli_lcf]

hmmm... first investigations are confusing... Just now I got the 1st order solution to be more or less converged to semi-steady state, thus switching to the 2nd order. linearUpwind and Gamma schemes for the convection term in momentum equation lead to the divergence of the solution (simulation crash). Tried to change from cellLimited to the faceLimited (more diffusive???) for the gradients: even with the 1st order solution diverges...

Continuing my tries... Will report later on.

[makaveli_lcf]

Additional question: I didn't saw any other schemes apart from CD for the diffusive term in the momentum equation:

div((nuEff*dev(grad(U).T())))     Gauss linear;

... and laplacian schemes...

Are there any special reasons for that?

Thank you!

[immortality]

hello Dr.Alexander
how's the results of your investigations?
I like to know for selecting a good scheme for my high gradients shock problem.

[fumiya]

Hi,

I'm summarizing the limited version of gradient schemes, e.g. cell(MD)Limited and face(MD)Limited schemes.
As of now, only cellLimited scheme is covered in this slide but I will update it as soon as possible.

http://www.slideshare.net/fumiyanoza...es-in-openfoam

Hope this helps,
Fumiya

[ziad]

Thanks for sharing this Fumiya. Your English is pretty good actually...

Originally Posted by fumiya

Hi,

I'm summarizing the limited version of gradient schemes, e.g. cell(MD)Limited and face(MD)Limited schemes.
As of now, only cellLimited scheme is covered in this slide but I will update it as soon as possible.

http://www.slideshare.net/fumiyanoza...es-in-openfoam

Hope this helps,
Fumiya