[rolando]

Hi,
this is not a bug but a way to overcome some trouble resulting from surfaceInterpolation::deltaCoeffs().
The problem arises, if there are some concave cells in the mesh.
(I know, that´s not fine, but they are outcome of a mesh moving process and do not affect the quality of my calculation.)

As there is:
DeltaCoeffs = 1.0/(mag(unitArea & delta) + VSMALL);
deltaCoeffs tends to infinity in the above case.
This causes the problems in my calculation.
To overcome this problem I use a limiter:
DeltaCoeffs = 1.0/mag(max(unitArea & delta, 0.087 * mag(delta)));
The parameter 0.087 could be set smaller (~0.2 should also do fine) to approximate the original implementation.
The parameter 0.087 is the output of some test calculations.

Rolando

[henry]

Your proposal looks fine and I will consider it further. How did you find that 0.087 is the optimal coefficient and how general do you think it is?

Henry

[rolando]

I could not say that 0.087 is the optimal value, I think that a somewhat smaller value would be more general. With 0.03 or 0.04 you should be save.
The value 0.087 corresponds to a maximum angle of 85° between delta and unitArea.
By choosing this value there was no more influence of the concave cells in my calculation.

Rolando

[billy]

nice picture and explanation.