|June 22, 2015, 09:21||
What is wrong with my meshing method ?
Join Date: Jul 2012
Posts: 26Rep Power: 6
I'm currently meshing a pressure swirl injector with ICEM CFD. I tried to use hexaedral elements but, unfortunatelly, the blocking strategy is so complex for me that I decided to give up this solution. Consequently, I'm using tetrahedral elements and prisms for the boundary layers. Despite all my efforts, the mesh quality is not as good as I want.
I can't be more specific on the geometry of the injector due to confidentiality reasons but I can explain to you my meshing method in order to improve it with, hopefully, all your advices... For your information, I also followed the tips and tricks of Simon Pereira which were very helpful but now, I have reached the limits of my weak skills. Thank you for any help.
Notice : quality chart hereunder refers to the quality of tetrahedral elements only.
a) I'm building the topology (mesh_01.jpg)
b) I'm creating a Body named D01_FLUID.
c) I'm renaming the different parts of my geometry. All the points are put together in a new part called D02_POINTS and all the curves are put in a new part called D03_CURVES. Then, the different surfaces are named regarding their real function (P01_INLET, P02_EXTERNAL_TANK, P03_INTERNAL_TANK, etc..)
2) First volume mesh (Octree)
a) Global mesh parameters : curvature/proximity refinements is enabled with "ignore wall thickness" ticked (Mesh_02.jpg).
b) Volume mesh parameters : Octree with thin cuts for walls that need it (Mesh_03.jpg).
c) Part mesh setup : maximum size is defined for the different surfaces and value of 0 is put for all the curves. Some walls are defined as "internal wall" according to my geometry (Mesh_04.jpg).
d) This first volume is then computed.
e) The mesh is checked and some minor problems are detected. It's only "multiple edges" and "single edges", which seems consistent with the geometry of my injector.
3) Surface mesh
a) All the tetrahedral elements are removed in order to keep only the triangles.
b) Triangles on surfaces are smoothed. Smoothing operations begin with Laplace option up to Quality at 0.3-0.4. Then, smoothing is finished with classic option because final triangle Quality is better (Mesh_05.jpg & Mesh_06.jpg).
4) Second volume mesh (Delaunay)
a) Based on the previous mesh only made with triangles, a new volume mesh is generated but this time it's Delaunay (Mesh_07.jpg).
b) The mesh is checked : only the same "multiple edges" and "single edges" are detected.
b) The mesh is smoothed and the quality is quite good (Mesh_08.jpg).
5) Boundary layer (prisms)
a) The boundary layer is inflated part by part. The parameters are given hereunder (Mesh_09.jpg) and the "Auto Reduction" is active.
b) After the creation of the boundary layer on two different walls, I noticed that the quality of the tetra was significantly decreasing (Mesh_10.jpg).
c) And after adding the last boundary layer, the tetra quality was very bad (Mesh_11.jpg).
d) The mesh is checked : only the same "multiple edges" and "single edges" are detected.
|June 22, 2015, 09:57||
Join Date: Jul 2012
Posts: 26Rep Power: 6
6) Third volume mesh (Delaunay)
a) All the tetrahedral elements are removed but surface mesh, prisms and pyramids are kept.
b) Then a new tetrahedral Delaunay mesh is computed in order to fill the remaining volume. This operation helpt me to improve the tetra quality a little bit.
c) Then the mesh is smoothed (prisms frozen) and some cell nodes are moved one by one in order to obtain the final quality (Mesh_12.jpg).
|Thread||Thread Starter||Forum||Replies||Last Post|
|Meshing software comparison||Wikie||Mesh Generation & Pre-Processing||5||September 29, 2014 06:09|
|Lattice Boltzmann method vs Finite Element Method and Finite Volume Method||solemnpriest||Main CFD Forum||3||August 12, 2013 11:00|
|[ANSYS Meshing] Ansys meshing with extended meshing||jsm||ANSYS Meshing & Geometry||6||January 10, 2011 13:09|
|Panel method meshing programs||Charles||Main CFD Forum||3||March 24, 2007 14:54|
|Which meshing method is better?||Melisa||Main CFD Forum||0||January 14, 2004 12:38|