Conversion of a mesh to polyhedra only applies to 3D meshes that contain tetrahedral and/or wedge cells.
To begin the conversion process, ANSYS Fluent automatically decomposes each non-hexahedral cell into multiple sub-volumes called “duals” (the shaded regions seen in the 2D example in Figure 5.60: Connection of Edge Centroids with Face Centroids). Each dual is associated with one of the original nodes of the cell. These duals are then agglomerated into polyhedral cells around the original nodes. Therefore, the collection of duals from all cells sharing a particular node makes up each polyhedral cell (see Figure 5.61: A Polyhedral Cell). The node that is now within the polyhedral cell is no longer needed and is removed.
To better understand how duals are formed, you can consider the straightforward case of a tetrahedral mesh. Each of the cells are decomposed in the following manner: first, new edges are created on each face between the face centroid and the centroids of the edges of that face. Then, new faces are created within the cell by connecting the cell centroid to the new edges on each face. These interior faces establish the boundaries between the duals of a cell, and divide the cell into 4 sub-volumes. These dividing faces may be adjusted and merged with neighboring faces during the agglomeration process, in order to minimize the number of faces on the resultant polyhedral cell.
Note: By default, ANSYS Fluent checks the aspect ratio of boundary
layer cells and if the aspect ratios are high, ANSYS Fluent asks whether
you want to preserve cells at the boundary layer (which could be useful
for keeping cell counts low when retaining boundary features). You
can override this default with the text command: mesh/polyhedra/options/preserve-boundary-layer? (for additional information about these options, see
mesh/
in the Fluent Text Command List).
Important: Hexahedral cells are not converted to polyhedra when the domain is converted, except when they border non-hexahedral cells. When the neighboring cell is reconfigured as polyhedra, the shared face of the hexahedral cell is decomposed into multiple faces as well, resulting in a polyhedral cell. In such a case the shape of the original hexahedral cell is preserved (that is the overall dimensions of the cell stay the same), but the converted cell has more than the original 6 faces (see Figure 5.62: A Converted Polyhedral Cell with Preserved Hexahedral Cell Shape).
Conversion proceeds in a slightly different manner in boundary layers that are modeled using thin wedge cells. These cells are decomposed in the plane of the boundary surface, but not in the direction normal to the surface. The resulting polyhedra will therefore preserve the thickness of the original wedge cells (Figure 5.63: Treatment of Wedge Boundary Layers). In most cases, the cell count in the new polyhedral boundary layer will be lower than the original boundary layer.
To convert the entire domain of your mesh, click in the Setting Up Domain ribbon tab (Mesh group box).
Setting Up Domain → Mesh → Make Polyhedra
The following is an example of the resulting message printed in the console:
Setup conversion to polyhedra.
Converting domain to polyhedra...
Creating polyhedra zones.
Processing face zones...........
Processing cell zones...
Building polyhedra mesh...................
Optimizing polyhedra mesh.....
>> Reordering domain using Reverse Cuthill-McKee method:
zones, cells, faces, done.
Bandwidth reduction = 1796/247 = 7.27
Done.
Figure 5.64: The Original Tetrahedral Mesh, the original tetrahedral mesh of a section of a manifold, is compared to Figure 5.65: The Converted Polyhedral Mesh, which is the resulting mesh after the entire domain is converted to a polyhedra.
Note that by default, the surfaces (that is, manifold zones of type interior) will be lost during the conversion to polyhedra. If you would like to preserve any of these zones (in order to utilize them for postprocessing, for example), use the following text command prior to the conversion:
mesh → polyhedra → options → preserve-interior-zones
You will be prompted to enter a string of characters, and only those interior surfaces with a name that includes the string you specify will be preserved.
In parallel, mesh migration and reordering is automatically performed after the conversion is performed. If this process takes a long time and the solver will not be run immediately in the same Fluent session, you can disable it using the following text command:
mesh → polyhedra → options → migrate-and-reorder?
If migration/reordering is disabled, the new mesh partitioning will be applied when the mesh is read into a new Fluent session and it is recommended that you manually reorder at that time.
Some limitations you will find with polyhedral meshes that you generally do not experience with other cell types include:
Meshes that already contain polyhedral cells cannot be converted.
Meshes with hanging nodes/edges will not be converted. This includes meshes that have undergone hanging node adaption (see Hanging Node Adaption in the Theory Guide), as well as CutCell meshes (generated by Workbench or the meshing mode of Fluent, or resulting from the CutCell zone remeshing method in the solution mode of Fluent) and hexcore meshes (generated when using the Hex Core meshing scheme in GAMBIT or the Hexcore menu option in the meshing mode of Fluent).
The following mesh manipulation tools are not available for polyhedral meshes:
the mesh/modify-zones/extrude-face-zone-delta text command
the mesh/modify-zones/extrude-face-zone-para text command
skewness smoothing
swapping
Meshes in which the domain has been converted to polyhedral cells are not eligible for adaption with the default hanging node method, though they can be refined with the polyhedral unstructured mesh adaption (PUMA) method. For more information about adaption, see Adapting the Mesh.
A mesh that is composed entirely of polyhedral cells has limitations when used in a dynamic mesh problem: it cannot undergo dynamic layering, and the only remeshing method available is CutCell zone remeshing. Smoothing is allowed for polyhedral cells, and diffusion-based smoothing is recommended over spring-based smoothing (see Diffusion-Based Smoothing for details). The linearly elastic solid smoothing method is not compatible with polyhedral cells.
Since polyhedral meshes can have a much higher ratio of nodes to cells then tet or hex meshes, the use of the node-based gradient method may result in a significant increase in memory consumption compared with other gradient methods.