Mmg remove boundary and deform surface?

Hello Agliane,
hope you’re fine,

unfortunately while using mmg 5.4.0, I encountered a new problem of mesh deformation and disappearance of surface using mmg, I do not know if it is in relation with mmg or it comes from the main mesh generated by FreeFem ++, I join you the mesh, the code FreeFem ++ and the two images before and after adaptation where we can see that the front of the top bay has been deformed and the surfaces around the bay up and down have disappeared, I also put the message to you that I got when generating the adapted mesh :

Warning: MMG5_chkBdryTria: 44 extra boundaries provided. Ignored

and in the more complex geometry I obtained these warning :

Warning: MMG5_chkBdryTria: 26434 extra boundaries provided. Ignored

Warning: MMG5_interpreg_ani: at least 1 impossible metric interpolation.

Warning: MMG5_movbdyregpt_ani: unable to compute optimal position for at least 1 point.

Thanks in advance,
problem_adaptmesh_mmg.edp (970 Bytes) toto.0.mesh (25.0 KB)

Hi Georges,

Fine, thank you. And you?

You are facing different problems:

  • the front of the top bay has been deformed because the angle of 3Pi/4 between the 2 surfaces has not been detected as a sharp angle. By default, an angle is seen as “sharp” if the angle between the surface normals is greater than 45° (i.e. the angle between the 2 surfaces is smaller than 3Pi/4) so I suppose that we miss the detection at the epsilon machine: I try to remesh using the -ar 40 option of Mmg and the surface is not deformed anymore;

  • the removed boundaries are due to the fact that Mmg preserve only “boundary” triangles (triangles between two domain with different references or triangle at external boundary). Mmg warns you that it remove some of the provided triangles here :
    ## Warning: MMG5_chkBdryTria: 44 extra boundaries provided. Ignored
    You can choose to preserve non-boundary triangles (in the sense of Mmg) using the -opnbdy option (but it slows down the remesher).

  • The 2 other warnings means that Mmg fails to compute, interpolate or intersect the anisotropic metric on 1 node at least and fails to move one point during the optimization step. It is not “normal” but happens frequently with 3D aniso size maps : you don’t have anything to do, just to check the quality of the output mesh (but it is always better to do that even if Mmg doesn’t send any warnings).

Please, let me know if you need more help,