Personal tools
You are here: Home → Forum → Use → Meshing problem - Piston
 

Meshing problem - Piston

Up to Use

  • Meshing problem - Piston

    Posted by Collette at August 09. 2006
    Hello,

    I am trying to do the tutorial from CAELinux related to the piston with salome 3.2.1
    Despites the differences between salome-2 and  salome-3, I have succeeded in all the steps but, I have met a problem related to the meshing: When I use the recommended sequence of mesher (Wire discretisation - Triangle(mefisto) - Tetrahedron (netgen), salome pops up an error message:

    Mesh Computation Failed

    Here are the messages from the console:

    th. 3019982528 - Trace /dn22/SALOME/series3x/SMESH/Mandriva/SMESH_SRC/src/SMESH_I/SMESH_Gen_i.cxx [239] : SMESH_Gen_i::SMESH_Gen_i : standard constructor
    th. 3019982528 - Trace /dn22/SALOME/series3x/SMESH/Mandriva/SMESH_SRC/src/SMESH_I/SMESH_Gen_i.cxx [1967] : SMESH_Gen_i::Load
    th. 3019982528 - Trace /dn22/SALOME/series3x/SMESH/Mandriva/SMESH_SRC/src/SMESH_I/SMESH_Mesh_i.cxx [92] : SMESH_Mesh_i
    th. 3019982528 - Trace /dn22/SALOME/series3x/SMESH/Mandriva/SMESH_SRC/src/SMESH/SMESH_Mesh.cxx [89] : SMESH_Mesh::SMESH_Mesh(int localId)
    th. 3019982528 - Trace /dn22/SALOME/series3x/SMESH/Mandriva/SMESH_SRC/src/SMESH_I/SMESH_Gen_i.cxx [2768] : SMESH_Gen_i::Load completed
    Loading required GL library /usr/lib/libGL.so.1
    th. 3019982528 - Trace /dn22/SALOME/series3x/SMESH/Mandriva/SMESH_SRC/src/SMESH_I/SMESH_Mesh_i.cxx [92] : SMESH_Mesh_i
    th. 3019982528 - Trace /dn22/SALOME/series3x/SMESH/Mandriva/SMESH_SRC/src/SMESH/SMESH_Mesh.cxx [89] : SMESH_Mesh::SMESH_Mesh(int localId)
     CalcLocalH: 1403 Points 0 Elements 2818 Surface Elements
     WARNING: Intersecting elements 1661 and 1757
    el1 = np = 3 714 715 433
    el2 = np = 3 716 379 382
    layer1 = 1
    layer2 = 1
     WARNING: Intersecting elements 1662 and 1757
    el1 = np = 3 434 433 715
    el2 = np = 3 716 379 382
    layer1 = 1
    layer2 = 1
     WARNING: Intersecting elements 1662 and 1838
    el1 = np = 3 434 433 715
    el2 = np = 3 379 716 381
    layer1 = 1
    layer2 = 1
     WARNING: Intersecting elements 1757 and 1661
    el1 = np = 3 716 379 382
    el2 = np = 3 714 715 433
    layer1 = 1
    layer2 = 1
     WARNING: Intersecting elements 1757 and 1662
    el1 = np = 3 716 379 382
    el2 = np = 3 434 433 715
    layer1 = 1
    layer2 = 1
     WARNING: Intersecting elements 1838 and 1662
    el1 = np = 3 379 716 381
    el2 = np = 3 434 433 715
    layer1 = 1
    layer2 = 1


    If i do the meshing using netgen 1D-2D-3D, there are no problems. I am able to launch an aster computation ...

    Your sincerely,

    Yann COLLETTE

    • Re: Meshing problem - Piston

      Posted by Igor Nazarov at August 10. 2006

      Most likely you have problems with 3D meshing.

      1. First of all check your shape in Geometry module. Shape must be valid (But it seems to shape is valid since you have built 3D mesh “using netgen 1D-2D-3D”. 

      2. Check dimensions (bounding box to see sizes of your shape) in order to right choose of Algorithm parameters.

      3. Define Hypothesis: Wire discretisation - Triangle(mefisto) - Tetrahedron ((netgen) or (netgen-1D-2D)) +Tetrahedron (netgen)

      4. Define Algorithms and their parameters depending on sizes of your shape and its features

      5. Compute mesh


      Your mesh computation failed because of Intersecting elements. But 2D mesh has to be built.

      1. Check 2D mesh with controls – free edges, face orientation, mesh quality etc.

      2. Change Algorithms and their parameters to improve mesh

      3. Repeat steps 5-7

       
      Mesh has to be computed for valid shape . Be careful selecting Algorithms and their parameters. Submesh creation can be useful for some regions with features if steps 1-8 lead to failure.


      Igor

      • Re: Meshing problem - Piston

        Posted by Collette at August 10. 2006
        Thanks for the hints. But, how do you explain that the meshing works fine using netgen 1D-2D-3D ? The mesh is fine enough to allow a computation via code-aster. Your sincerely, Yann COLLETTE

        • Re: Meshing problem - Piston

          Posted by Collette at August 10. 2006
          Hello,

          You were right, there was a problem with the geometry (so, now, I don't understand why netgen 1D-2D-3D has worked) :)
          I've checked the geometry of the piston. I've tested all the "check" method and every method gives a false result. Except one: Measures->Check and improve blocks compounds.
          The error is "Extra hedge" and gives me the incriminate sub-shape Solid_1. I've checked and I found nothing labelled Solid_1 in the exploded piston !
          How can I locate the incriminated extra hedge ?

          Your sincerely,

          Yann Collette

          Yann Collette

          • Re: Meshing problem - Piston

            Posted by Collette at August 10. 2006
            Hello,

            I've tested the same process than with the piston tutorial but with a file from the opencascade website: a lever harm.
            I have the same error than with the piston (extra edge). But the meshing phase works fine with tetrahedron(netgen) - max volume = 10000 / triange(mefisto) / wireframe - average length = 5.

            Yann Collette

            • Re: Meshing problem - Piston

              Posted by Christophe Bourcier at August 11. 2006
              Hello

              The function "Check and improve blocks compounds" tells if your geometry can be meshed by the hexahedron algorithm. An extra edge shouldn't be a problem for the tetrahedron algorithm.

              The combination of algorithms Wire-Mefisto-Netgen is very sensitive about their parameters. Try something simplier:

              - 1D: Wire discretisation, Average length = 1
              - 2D: Mefisto, Length from edges
              - 3D: Netgen, no hypothesis

              Of course, you must adapt the average length to a value closer to your shape.

              If the computation succeed, you will see Netgen1D-2D-3D gives a much better result than Wire-Mefisto-Netgen.

              Cheers,

              Christophe
Powered by Ploneboard
Document Actions