CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Open Source Meshers: Gmsh, Netgen, CGNS, ... (http://www.cfd-online.com/Forums/openfoam-meshing-open/)
-   -   gmshToFoam - insufficient geometrical tolerance? (http://www.cfd-online.com/Forums/openfoam-meshing-open/86223-gmshtofoam-insufficient-geometrical-tolerance.html)

will.logie March 17, 2011 07:54

gmshToFoam - insufficient geometrical tolerance?
 
1 Attachment(s)
Hallo fellow Gmshers,

I've recently stumbled across a warning in checkMesh and I'm not sure whether the problem indicated is leading to the problems I'm having in my simulations or not.

The gmsh I've created (see pic in attachments) is supposed to be a 2 wedge from a cylindrical water tank, and the holes represent a coil heat exchanger. My original thinking was that structured-hexahedra are preferable but with the simulation irregularities, I'm beginning to question this. In any case, I have parameterised the gmsh file for later optimisation.

But to return to the point, when I run checkMesh I get the following failure:
Quote:

Wedge front with angle 1.00004 degrees
***Wedge patch front not planar. Point (0.260332 0.0457038 0.00454412) is not in patch plane by 3.73728e-08 meter.
Searching for the original point in the gmsh.msh file I find this likely suspect -> 8920 0.2603322947962474 0.04570379734961246 0.004544117108580378.

Is this warning to be taken seriously and if so, why doesn't gmshToFoam write a points file accurate enough?

If anyone can shed light otherwise, I'd be most thankful.
Cheers,
Will.

styleworker November 3, 2014 12:58

Maybe you have already solved your problem in the meantime.

For the sake of completeness: change writeFormat to binary in your controlDict before running gmshToFoam

rudolf.hellmuth February 2, 2016 09:10

gmsh saves the point positions with 16 digits of precision. gmsToFoam is truncating these big numbers, and thus making tiny shifts in the normal vectors of patch faces. Then the solver sees these tilts as non-planar. This problem doesn't occurs if the conversion is to binary data instead of ASCII.

I don't think these tiny point position shifts would impair your simulation.

I guess this problem could be solved in the source code of gmshToFoam with some sort of regularization of these truncation events.


All times are GMT -4. The time now is 09:47.