[Gmsh] Errors when increasing NNodesByEdge on a small edge

Fri Apr 12 15:06:07 CEST 2019

Dear Christophe,

Thank you! That looks like it has done the trick!

Kind Regards,

[FLOW]

[VUB]
        Nathan Ricks | PhD Student

Department of Mechanical Engineering
Thermo and Fluid dynamics (FLOW)

Brussels Faculty of Engineering (BRUFACE<http://bruface.eu/>)
Vrije Universiteit Brussel (VUB<http://www.vub.ac.be>)

________________________________
From: Christophe Geuzaine <cgeuzaine at uliege.be>
Sent: Thursday, 11 April 2019 9:37 PM
To: Nathan RICKS
Cc: gmsh at onelab.info
Subject: Re: [Gmsh] Errors when increasing NNodesByEdge on a small edge

> On 11 Apr 2019, at 18:22, Nathan RICKS <Nathan.James.Ricks at vub.be> wrote:
>
> Sorry for the repeated messages, just trying to sort out the issue and coming across new things/information that might help resolve the issue. In the .geo file attached, it demonstrates what "I think" is the cause of my errors - if you generate the 2D mesh and look at the trailing edge, there are cells/nodes crossing over the lines/edges where they obviously shouldn't be. This occurs on both Mac and Linux for me at least. Have you seen this issue before by any chance?
>

It seems that you ran into a case where the small random perturbation we use in our 2D initial triangulation is too large. A new version of the algorithm without any such perturbation will be available in a future release; but in the meantime you can set e.g. "gmsh -rand 1e-12" (or set Mesh.RandomFactor = 1e-12 in your script), which seems to solve the issue.

Christophe

> Thanks again,
>
> Nathan Ricks
> VUB
>
>
>
>
>
> From: Nathan RICKS
> Sent: Thursday, 11 April 2019 11:41 AM
> To: Christophe Geuzaine
> Cc: gmsh at onelab.info
> Subject: Re: [Gmsh] Errors when increasing NNodesByEdge on a small edge
>
> I'm not sure if this helps determine the issue, but it seems if I generate the mesh using the Delaunay algorithm, there are cells that break through boundaries of the surfaces, and if I use Frontal-Delaunay (which I was) these correspond to the broken cells shown in the first screenshots. I've attached two screenshots showing this, if you compare between the two you can see "correspondence of errors".
>
> From: Nathan RICKS
> Sent: Thursday, 11 April 2019 9:00 AM
> To: Christophe Geuzaine
> Cc: gmsh at onelab.info
> Subject: Re: [Gmsh] Errors when increasing NNodesByEdge on a small edge
>
> Dear Prof. Geuzaine,
>
> Thanks for the response. Indeed, I just tested this using Gmsh on my Mac, and the resultant mesh runs fine.
>
> So it seems the problem is restricted to the workstation, which is running Linux 64-bit (red hat). I just tried instead compiling from the source code and the error persists there also unfortunately. Are you aware of anything that may cause this? It's for an optimisation routine, so thousands of meshes are generated, so using my Mac isn't feasible unfortunately.
>
> In regards to the boundary layer, I was originally using the BL field directly, however ran into issues as I couldn't have a growth rate "around" the airfoil spline (to have narrower cells at leading and trailing edges), as well as perpendicular to the airfoil. Also, at the trailing edge using the fan node resulted in too small cells causing numerical errors, and without it the cells were too bad quality, also causing errors. Using the transfinite surfaces gave me flexibility in the boundary layers, allowing two different growth rates perpendicular to the surface. With that it was running very well, until now on trying to include the blunt TE.
>
> Thank you again,
> Kind Regards,
>
> Nathan Ricks.
>
>
>
>
> From: Christophe Geuzaine <cgeuzaine at uliege.be>
> Sent: Thursday, 11 April 2019 12:09 AM
> To: Nathan RICKS
> Cc: gmsh at onelab.info
> Subject: Re: [Gmsh] Errors when increasing NNodesByEdge on a small edge
>
>
>
> > On 10 Apr 2019, at 21:57, Nathan RICKS <Nathan.James.Ricks at vub.be> wrote:
> >
> > Hi all,
> >
> > I have quite a particular problem, but I'm hoping someone can help. I'm meshing an airfoil using transfinite surfaces in the boundary layer.
> >
> > I've just had to add a blunt trailing edge (mesh was working well with sharp trailing edge) to the airfoil, and now I have this one error stopping me, and I just can't figure it out...
> >
> > When I increase the NNodesByEdge on Field[11] from 6 to 15, the mesh ends up bad quality, with cell corners not matching other cells at the boundary of surfaces, and it causes SU2 to diverge instantly. It runs fine with 6, but crashes at 15. I've attached two images showing that it is a mesh issue causing this (nothing else is being changed in the setup).
> >
>
> I cannot seem to reproduce the issue on my Mac. Attached is the mesh generated on your .geo file:
>
>
> > Other things such as reducing lcMin on Field[12] can also make it crash for example, however the above has visible errors in the mesh.
> >
>
> Idem... What OS are you running on?
>
> PS: instead of generating the BL by hand, why not using the boundary layer field directly? Cf. e.g.
>
> https://gitlab.onelab.info/gmsh/gmsh/blob/master/benchmarks/2d/naca12BL.geo
> https://gitlab.onelab.info/gmsh/gmsh/blob/master/benchmarks/2d/naca12_2d_trailing.geo
>
> > Does anyone have an idea on why I can't use the Distance/Threshold field as normal on this edge?
> >
> > If you can provide any guidance that would be greatly appreciated, it's so close...! I've already tested with the latest release with the same results.
> >
> > Thanks a lot,
> > Kind Regards,
> >
> > Nathan Ricks
> >
> >
> >
> >
> >
> > <NNodes_15.png><NNodes_6.png><airfoil_level1.geo>_______________________________________________
> > gmsh mailing list
> > gmsh at onelab.info
> > http://onelab.info/mailman/listinfo/gmsh
>
> —
> Prof. Christophe Geuzaine
> University of Liege, Electrical Engineering and Computer Science
> http://www.montefiore.ulg.ac.be/~geuzaine
>
>
>
> <airfoil_level1_edit.geo>_______________________________________________
> gmsh mailing list
> gmsh at onelab.info
> http://onelab.info/mailman/listinfo/gmsh

—
Prof. Christophe Geuzaine
University of Liege, Electrical Engineering and Computer Science
http://www.montefiore.ulg.ac.be/~geuzaine

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://onelab.info/pipermail/gmsh/attachments/20190412/b67e0e5f/attachment.html>