Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

Sign In Apply for Membership

wrongOrientedFaces
  • yeahright March 2010
    Hi,
    I'm trying to mesh a fan/propeller. If I create the prismatic layer, the checkMesh utility complains about :

    ***Error in face pyramids: 14 faces are incorrectly oriented.
    <<Writing 12 faces with incorrect orientation to set wrongOrientedFaces

    This sounds alarming, OF does however compute something resembling a flow before divergence ... What does that mean ?

    Can I avoid this by adjusting the boundary layer parameters, any suggestions ?

    Kind regards
    Peter
  • Oliver April 2010
    Hi Peter,

    the problem is known but difficult to avoid. I have prepared a small illustration to show what those wrongly oriented faces actually are; Mattijs Janssen from OpenCFD explained it to me.

    [attachment=0]FacePyramids.png[/attachment]

    It is easier to explain in 2D, hence you have to assume that the quad cell in the left picture is a prismatic cell. The left cell, however, is perfectly healthy and the four triangles which are formed by the centre and the edges (corresponds to faces in 3D) are the face pyramids. They all have a positive area and checkMesh would be happy. On the right side you see a distorted version of the cell and the centre is outside the actual finite volume cell; hence two face pyramids become negative (the two red triangles). The cell is still fine for most finite volume applications but it might, obviously, have an impact on accuracy and stability. According to Mattijs it would definitely make any Lagrangian solver fall over. Those cells are very difficult to avoid in certain regions (e.g. the connection of a sharp trailing edge and a symmetry plane). My advice, however, would be to try to use the grid you have (under-relaxation, start with 1st order, the usual tricks).

    I have a few ideas on how to avoid the problems completely, but they are not easy to implement. Realistically that might be done in co-operation with a university (e.g. master thesis) and thus I wouldn't expect that to be available soonish. As far as I know, however, commercial meshers will produce similar problems).

    I hope this helped a little!

    Regards,
    Oliver
  • yeahright April 2010
    Oliver,
    Thank you for your answer. As you told, I've been computing the case by using the usual tricks, 1st-order upwind, relaxation and
    lower Re etc at the start and it seems that it might converge to a meaningful result.

    Let me take this opportunity to thank you for Engrid mesher, it has some some really fine qualities, when compared to snappyHexMesh,
    for instance.

    I wish that it would be possible in the future versions to control the the boundary layer height in different areas/surfaces.
    The wish list of course is infinite but first of all I would like to be able to do the following
    sudo apt-get install engrid
    as the efforts to compile Engrid with all the correct dependencies has really driven me grazy :D .
    Maybe in the next Ubuntu 10.04 LTS ?

    Best regards
    Peter
  • Mike April 2010

    "yeahright":

    The wish list of course is infinite but first of all I would like to be able to do the following
    sudo apt-get install engrid
    as the efforts to compile Engrid with all the correct dependencies has really driven me grazy :D .
    Maybe in the next Ubuntu 10.04 LTS ?


    You can already do that by adding this PPA to your sources list: https://launchpad.net/~zohn-joidberg/+archive/engrid
    Unfortunately, you'll get an old version for now.
    (Don't worry, it might get updated soon. ;) )

    Getting engrid into official Debian or Ubuntu repositories might take some more time and is unlikely for Ubuntu 10.04. But who knows? Maybe it will work.

    Until then, use our install script: viewtopic.php?f=2&t=62
  • chossfeld June 2010
    Hello,

    I understand the issue with the wrong orientated faces. I got the same problem now and I can not manage the simulation with the normal stuff. I am trying to simulate a heat transfer case with chtMultiRegionSimpleFoam. I have found out so far that if I use the basic mesh with I have done with Salome (netgen) I can do the simulation with laminar flow. In this case checkMesh does not complain very much. When I important the fluid part of the mesh into engrid, all cells (Tetras) are gone. When I do new netgen only mesh with engrid everything works fine (and checkMesh does not complain too much) except that I am limited to laminar flow. When I try to introduce a prismatic boundary layer for turbulent calculation (I have tried several settings), I get wrong orientated faces and I am not able to get a stable solution. So the issue should be located somewhere within the prismatic boundary layer construction. I am using the latest version of engrid (1.2.0) and since the last post to this topic is rather old I wonder if there is a solution.

    Any hind is highly appreciated

    Christoph
  • Oliver June 2010
    Hello -- the 1.2.0 should be fairly robust. Certain geometries, however, never work out. Could you plot the region with the wrongly oriented faces and post it, please? Maybe there is a simple trick to make it work for your case. I assume that you have tried the usual tricks of under-relaxation and so on ... -- Regards, Oliver
  • atypicalguy December 2012
    I am having the same problem. I was able to get all the faces oriented correctly once, but somehow am unable to repeat that. checkMesh is giving me thousands of wrongOrientedFaces. They seem to be along the leading edge of the wing. It is a wing and a symmetry plane. Not sure what I am doing wrong; seems like a random problem that is there sometimes but then sometimes not.
  • wyldckat December 2012
    Greetings atypicalguy,

    Try exporting the surface mesh and then run in OpenFOAM this command:
    surfaceCheck -help

    After you've checked the options, at least try once the command with the exported geometry:
    surfaceCheck surface.stl


    My guess is that along the leading edge the refinement is somewhat damaged, or that faces are mixed on it.

    Best regards,
    Bruno