I bet there is no such a thing on earth. But I might be dead wrong. If you know a software that makes hexa meshes automatically for 3D, please let me know. Is it worth pursuing such a seemingly impossible thing?

Even for a very simple geometry such as a pyramid with square base, it is impossible, topological sense.

Hope to see a miracle! Thanks.

Selina

Hello Selina,

I am sorry I can't be your hope.

But why bother with hexa. Just try tet or mixed meshes unless the accuracy of your analysis is critical.

Melvin

(1). Why not take a look at the standard mesh generation book by Joe Thompson,"Numerical Grid Generation", which is available free on Internet.

THANKS. I bet almost every one at this forum knows and read the book at least once. Do you know by any chance a auto hexa generator without any crack inside a domain. Having said that we should start to think about topology seriously. Thanks.

Selina

Yes there are quite a few algorithms to do this. This is a big topic being very actively researched but mainly by the FE structures people. Here is a link to a few online papers:

http://www.andrew.cmu.edu/user/sowen/recent1.html

Yes, it exists, but will be publicly released only in a few months as we are still improving and interfacing this new product. It has already been tested successfully on various internal and external configurations (planes, cars, Formula 1, complex valves, passenger cabin of an aircraft, highly twisted fans, ...).

Do not hesitate to contact Numeca for more information.

You must be careful about what you mean by "hexa meshes". Most commercial codes can handle a fair amount of distortion from the ideal cubical shape. There is nothing impossible about meshing any shape with a hex grid. To see this, note that any tetrahedron can be divided into four hex cells (Possibly distorted, but topologically hexahedral). When you throw in the possibility of non aligned or overlapping meshes, there are several more solutions ...

I'm also curious about why you insist on hex meshes in the first place. Most commercial codes have been drifting to the opposite extreme, of allowing various cell shapes -from tetrahedral to polyhedral - to make the meshing task easier. Although a simple hex grid might be "prettier", it is a hard luxury to justify anymore on a strict performance basis.

(1). Tet/Tri cell with line edges can not represent the actual curved shape geometry. (2). This is the reason why Phong shading and Gouraud shading were invented about 25 years ago to display smooth image from a flat facet surface model. (3). Even with the hexa cell, the finite-volume method simply does not have the capability to represent the curved geometry properly through individual cell. (4). Only the finite-difference method through the coordinate transformation (body-aligned coordinates)can properly account for the real curved shape geometry through the coordinate transformation factors. (5). If you don't think so, then you can try to use 2-D tri-cell mesh and run a simple supersonic flow over a NACA smooth surface airfoil. You will always see the expansion waves and shock waves generated from the vertices of the wall boundary. In other words, using finite-volume method, the flow will see the airfoil as a diamond-shaped airfoil.(This was verified several years ago when I was using a widely used commercial CFD code. It is still being used today.) Such solution is simply incorrect, not even an approximation. (6). This is not a problem when one use finite-difference method through coordinate transformation, where the boundary is transformed from a curve into a line through transformation factors. (finite-volume cell does not have this property. With the finite-element method, the edge shape is pre-determined also. In other words, the real geometry can not be reproduced in FV and FE method.)

Hi,

ICEM have just finished something that may be of interest, it's not 100% HEX but is close.

ICEM can now take a TET mesh and automatically convert it to as many HEX cells as possible. An example can be seen at www.icemcfd.co.uk. Click on the NEWS FLASH button in the bottom left. This will be released in 4.1.2 over the next few weeks. The example here is simple but it is an automatic process so the data can be a complex as you like.

If any body has any questions on this then please let me know.

Regards

David