limited precision of checkMesh?
 

Hi foamers!

I'm working with "natural"(smooth) shaped junctions in micro-scale (at least I try to). My problem already occurs by scaling down my mesh from meter to micrometer (with OFs "transformPoints-tool").
I ran checkMesh before scaling and afterwards: The results on meter-scale were OK. The shape caused a moderate nonOrthogonality to my pure hex-mesh.
But at a certain scaling point (where the cell size drops below 1µm) checkMesh detected some "wrong oriented faces"

But an analysis of this so-called-damaged mesh in postPro(VisIt) showed nothing of such problems, just a medium quality mesh.

So should I expect an impact on my calculation or are there some limits in checkMesh's pecision?

Regards,
Robert

I've made another attempt: this time with just a simple box and equidistant hex-cells. I produced the mesh size on three different ways:
- scaled by "transformPoints"-utility
- scaled by my own written scaling script
- and build the mesh immediately in the right µm size (in Salome)

Everytime the same result: checkMesh detecting weird things as soon as the cell size drops below 1µm.
Have you ever seen ideal hexahedrons with a skewness of 160 ?

I'm completely confused and can't even see a reason for this error.
Please could anyone give me an advise.


PS: I'm working with OpenFOAM 1.5

Regards,
Robert

Your findings are very helpfull for everybody simulating small biological structures! This should be in a bug report.

Is it only a checkMesh issue? Or do the solvers also have problems with these scaled cases (crashing, giving unphysical results, ...)?

Hi Robert

I have seen similar issues, however they originated from the value of writePrecision in controlDict. Try, whether or not checkMesh behaves well if you use e.g. writePrecision 14 during the transformation process.
writePrecision controls the precision of all written data, i.e. also your transformed mesh.

Hope it helps,

Niels

Thank you Niels,

this advise is quiet helpful, but in this caes it not solved the problem.
I set write precision to "17" and there was an influence but the skewness of my equdistant test cube still occurs in checkMesh (at 10µm cell size).
There must be solution: I saw cases in this forum with even smaller cells like in this thead: "interfoam behavior in micro-dimensions".

Thanks for your help
Regards, Robert

Don't forget - in double precision you have only 14 significant digits; the rest will be garbage.

Hrv

May be the source of my problem comes from preProcessing.
I used Salome to build the mesh. Salome gots a precision of 1e-7. The purpose of my scaling actions is to increase this accuracy with the scaling factor.
Precision(1e-7) * scaling(1e-6) should be accurate enough or do I misinterpreting something?