https://linen.dev logo
Join Discord
Powered by
# meshing
  • l
    saw this aswell, currently putting it into my devcontainer and seeing how it goes
  • l
    another option might just be to ditch the symmetry and calcualte the full aircraft domain, rather than half of it
  • l
    shitty for optimization but if there is no other way then hey, you gotta do what you gotta do
  • f
    For me I need to make conformal background mesh, or that is the plan currently. It will give the best results
  • l
    because my aircraft wall does not actualyl touch the outer boundaries
  • o
    i think that your BL will be problematic after the dual mesh, one would expect to create them AFTER any mesh modification.
  • l
    can SHM or rather cfmesh, because I hate SHM, do poly boundary layers?
  • l
    i would think the bl generation is tied to the cell type the meshers are designed for
  • o
    also, I highly recommend for you is to run your simulation dont stay in the look of the mesh, i played a little bit with polydual before, and the meshes looked great until the simulation was runned.
  • o
    SHM for sure, cfmesh what it does is divide the last layer of cells, similar to refineWallLayer app of OF, SHM push the layer by pushing the cells. I think in case that you are bringing a 'classic' mesh from outside (by classic i mean a mesh that you have not intentionally created a super thick 1 cell layer before) it would be better SHM
  • o
    SHM does poly boundary layers as it extrudes the surface cells, generally it creates poly cells as after the snappying step there are all kind of crazy cells on surface anyways.
  • l
    hm okay good to know, but I have found that SHM does not really work for my surface
  • l
    I can barely scratch 50% coverage of the surface for the boundary layers
  • l
    it might change with this as maybe its easier now without odd cells across the surface that break the BL
  • l
    but even before it barely scratched 50% coverage with all the quality checks pretty much disabled
  • l
    So using the conversion and then using improveMeshQuality DOES lead the mesh to be calculated but as probably expected it broke the boundary layer to a point where its not uniform anymore and at that point why even bother from there
  • l
    So I guess now Im gonna try a tetra mesh without BL and then generate the boundary layers for that
  • l
    spoiler alert SHM boundary layers did not form at all, none
  • o
    strange, what is the output in the log?
  • l
    Copy code
    nSurfaceLayers 6
expansionratio 1.2
firstLayerThickness 0.001
minThickness 0.000001
nGrow 0
featureAngle 180
nRelaxIter 5
nSmoothSurfaceNormals 1;
nSmoothNormals 3;
nSmoothThickness 0;
maxFaceThicknessRatio 1;
maxFaceThicknessRatio 1;
maxThicknessToMedialRatio 1;
minMedialAxisAngle 30;
nBufferCellsNoExtrude 0;
nLayerIter 50;
  • l
    Copy code
    Quality controls only:
minVol 1e-30
minTetQuality 1e-30
maxConcave 80
  • l
    On every layer iteration its just giving me a bunch of faces with face-decomposition tet quality smaller than 1e-30 and faces with concavitiy greater than 80
  • l
    checkMesh -allTopology -allGeometry on the origin mesh on which I apply the BL gives 5 failed checks with
    Copy code
    Checking geometry...
   *Number of severely non-orthogonal (> 70 degrees) faces: 84.
  <<Writing 84 non-orthogonal faces to set nonOrthoFaces
 ***Error in face pyramids: 3 faces are incorrectly oriented.
 ***Max skewness = 10.3338, 3 highly skew faces detected which may impair the quality of the results
 ***Error in face tets: 909 faces with low quality or negative volume decomposition tets.
   *There are 4203 faces with concave angles between consecutive edges. Max concave angle = 89.9791 degrees.
 ***Concave cells (using face planes) found, number of cells: 274066
 ***Faces with small volume ratio (< 0.01) found, number of faces: 1239

Failed 5 mesh checks.
  • l
    which isnt great yeah but i should atleast get BL over most of the surface if that works and only collapse on the local bad areas
  • y
    where are the bad areas?
  • l
    some around trailing edge and a lot of the concave cells are all around the domain but I know theyre fine because I had used them to calculate without boundary layers and it was fine
  • z
    Turn off all the quality checks and send it
  • z
    See what happens
  • l
    i did, didnt want to mesh
  • l
    or, well, add layers