[lzgwhy]

Dear friends,
Here I have a problem. I want to create a mesh for a 3D domain surrounded by a torus, which is similar to a tyre. It is very easy to create a mesh in which all elements (or cells) are hexahedron, i.e. the mesh is structured. However, I want to create a mesh in which all the elements (or cells) are triangular prism or other kinds of prism, i.e. on a cross section of the domain, I create a 2D mesh in which all elements are triangle or some other shapes and then I sweep this 2D mesh along the center circle of the 3D domain to create the 3D mesh for the domain. I tried two methods. In the first method, I first created the 3D blocks which together are topologically equivalent to the domain and then I changed the mesh type of one cross section face of one block to "all tri" and make it as the swept face. But this method always led to some errors. In the second method, I fisrtly created a 2D block on one cross section of the 3D domain and then I make the 2D block free and changed its type as all tri; finally I swept the 2D block along the center circle of the domain to make 3D blocks which are almost the same as those in the first method. This method did not resulted in errors, but the mesh can not be created successfully since only shell mesh were generated and no volume mesh were generated. Could you tell me how I can fix this problem? Do you have any other better method to create such a mesh for the 3D domain? Thank you very much and I am looking farwards to your reply.

Liu

[PSYMN]

My guess is that no one considered making a swept block do this (start and end at the same face), probably everyone prefers a hexa mesh to a swept triangle mesh...

But if you want to get this done, how about generating a a 2D Mesh, converting to unstructured mesh and then using the Extrude command to rotate the mesh (instead of blocking) around your central axis...

[lzgwhy]

Thank you very much. In some cases, the prism mesh may be better than hexa mesh, for example, when the dynamic mesh method is used. I tried your method and it was successful. However, I have a new problem. If the domain is just topologically equivalent to a domain surroundded by a torus, which means the cross sections may not be the same and the center circle line of the domain may not be on a plane, how can I create a prism mesh on it?

I have a another question here. I firstly create a block and do the association between it and some parts of the whole geometric model. And then I create a new block and do the association between it and the left parts of the whole geometric model. During the process, maybe I should hide the old block. Then I have two blocks and they may be separated without any geometric connection, though some faces of one block have the same spatial positions as some faces of the other block. In this cases, after the mesh is created, there are two sub-meshes based on the two cooresponding blocks. Then could I adjust the mesh nodes on the common surfaces of the two sub-meshes to make the whole mesh as a real one? Thank you again.

Originally Posted by PSYMN

My guess is that no one considered making a swept block do this (start and end at the same face), probably everyone prefers a hexa mesh to a swept triangle mesh...

But if you want to get this done, how about generating a a 2D Mesh, converting to unstructured mesh and then using the Extrude command to rotate the mesh (instead of blocking) around your central axis...

[PSYMN]

If the domain is just topologically equivalent to a domain surroundded by a torus, which means the cross sections may not be the same and the center circle line of the domain may not be on a plane, how can I create a prism mesh on it?

I would just block it (to get that topology, just do a simple Ogrid block with faces on the top and bottom, then delete the middle block, then run an Ogrid thru the remaining ring of blocks, no faces.) Then associate the blocking with your particlar geometry with its various cross sections.

That would get me a hexa mesh. You can't sweep a circle like that.

I have not heard this idea that prisms are better than hexas for anything in the solver.

[Rana shaharyar]

is it possible to create a prism meshing in ansys fluent module. in other words without using ICEM.

[PSYMN]

Yes, Fluent meshing includes a wide range of capabilities including prism...