curvature at the interface (interFOAM)
Hi all,
my question ist about the curvature calculation in interFoam. The magic happens in interfaceProperties.C and was also discussed often in the Forum. Anyhow, I still have an open question. The curvature is calculated by K_ = fvc::div(nHatf_); in the simple expression which is ok for me. Before that, the gradient of alpha is determined on the cell faces by interpolation from the cellgradient of alpha. Afterwards, the faceunit interface normal flux is calculated (nHatf_ = nHatfv & Sf;) and used for the calculation of the curvature as written above. But... I couldn't find out, how this divergence is realized. The Doxygen linked me to an explanation which is not helpful for me. I also set up a dummy hex 3x3 case an tried to understand the calculated value for the curvature in the centralcell, but did not succeed. I'll be glad for any advices. Greetings Lindstroem 
Hi Lindstroem, respect to the code of fvc::div, in gaussConvectionScheme.C we have:
Code:
00110 template<class Type> Code:
00042 template<class Type> Regards. 
Hi Santiago,
thanks for your help. But doesn't the GaussConvectionScheme need two parameters (faceFlux and vf)? in the interfaceProperties it is only called with one argument (nHatfv_). The reason why I am digging through that, is that I wanted to calculate the sum of the curvature over a circle which should analytically be 1/r. If I sum up the K_ in the interfaceProperties it results in sth. about 10e6 where the analytical solution would be 6.6. Do you know the reason for that? Greetings 
Right. You are calling fvc::div. You can see it also calls fvc::surfaceIntegrate.

Hi Anton,
thanks, I think I got it :) Greetings 
Hi again,
I would like to specify the question posted above. I have a 2D case with a bubble. If I sum up the curvature in interfaceProperties.C Code:
K_ = fvc::div(nHatf_); Has anyone did the same and can tell me what I did wrong? Thanks! Lindstroem 
How did you initialize the shape?

setFields with cylinderToCell
Code:
cylinderToCell 
From what I know it's better to start with a nonequilibrium shape (e.g. a box), relax the shape and then evaluate the resulting curvature.

Thanks for your comment. Just tried that:
The sum of the curvature starts with 6.802736e11 and ends with 9.166001e12. Should be something about 4. //edit: seems to be known: http://www.openfoam.org/mantisbt/pri...php?bug_id=158 
I don't understand why you want to check the accuracy of computed curvature in that way. Actually, at the interface of the droplet, at every point the curvature should be 1/r. I do not think summation of the interface curvature will equal 1/r. It should be n/r where n is the number of points in the summation.
Regarding to the curvature, you should note that the curvature vary quite a lot in interfoam. If you take the average of the curvature at the isosurface of alpha1 = 0.5, you will get a value close to the value 1/r. Since the interface is smeared over 34 cells approximately, at cells with alpha1 = 0.9 or 0.1, you still have curvature different than 0. Therefore summing the curvature of cells belonging to the interface also can not give a proper result. Hope it is clear for you. 
Hello Duong,
i got your point  absolutely right that summing up would result in sth like n/r. Thanks for your comment! Lindstroem 
Hi again,
I'd like to ask one more detail to my initial question concerning fvc::div(): Is it true, that we calculate the second derivative (the divergence) only at the centroid point of the cells using the first derivative (the gradients) from the faces? 
To me divergence and second derivative refer to two different things...

ok, sorry, the divergence of the gradient...

All times are GMT 4. The time now is 16:25. 