CFD Online Discussion Forums

CFD Online Discussion Forums (
-   Main CFD Forum (
-   -   ICEMCFD HEXA (

sarat Maharana December 11, 2000 07:47

Hi, I have a replay file and a tetin file.I am working on smoothing of mesh generated by running my replay file. The Problem: What is the way to get very good quality of grid when some domains are having elements with negative determinants.These elements are not smoothed at all. I have multi-block structured grid file. I need to know what exactly is blocking stucture.

Could u please explain me ?



John C. Chien December 11, 2000 15:12

(1). If you are not familiar with the code and the process of the particular mesh generation, then it is a good idea to do at least two things: (a). attend a training class, (b). follow closely the tutorial samples. (2). This code is not for a newcomer, and even with my background, I had to get help from the support engineers on the daily basis when using the code. (3). The code is one of the widely used mesh generation code, but the code has many options and is relatively hard to use. (or should I say, requires a steep learning experience ) (4). I had a very complex duct to model, and with their help, the final blocking number is 80 blocks, in order to get rid of the negative volumes.(the automatically generated bad volume mesh) (5). The code uses a multi-block structure,which the user must create first like quilt or clothing. This cloth is then projected onto (wear) the geometry model (connecting the edges of the block or cloth to the curves of the geometry model). The surface mesh and the volume mesh are then generated and projected onto the geometry. (it is like trying to cover a person with a fish net. ) (6). I had a lot of hard time in avoiding the negative volumes. Even with smaller blocks (creating the blocking structure with many more blocks so that you can fit the cloth smoother on the body), I didn't find an easy solution to it. (7). I would strongly suggest that: you send the file (by e-mail, if possible) to the support engineers and get them to fix it. You can spend your life time learning the mesh generation and still have problems in generating the acceptable mesh if the geometry is complicated. (8). Mesh generation is ART not CAD.

Sarat Maharana December 11, 2000 23:58

Hi, Thanks a lot for your response on blocking structure. I need to know how to decide on meshing laws for a geometry.Is there any specific rule to apply to fix law for entry and exit?. How grid distribution and grid clustering help to generate good quality grid ?

Thanks, Sarat

John C. Chien December 12, 2000 03:13

(1). I don't think, I can answer your question in general. (2). This is the reason why there are many other mathods invented, all trying to ease the hard work of mesh generation. (3). All I can say is, the hexa mesh is ideal for a rectangular duct. The surface mesh and volume mesh will be rectangular, and you always have positive volume cells. (4). Like kids toy "lego", you can attach additional blocks to create multi-block topological structure. So, this part of block generation is very similar to that of "Lego". (5). The need to use clustering comes from the physics of the flow solution. If you have a wall, and your soluiton and method demand the fine mesh near the wall, then you have to put more mesh points or cells there. (along the edge of the block to control the distribution) And the distribution is a function of your flow field solution, so it is hard to know ahead of the time. (6). Once the blocks are twisted and distorted, something like "Lego" made of sponge material to fit the complex geometry, then, it can be visiualized easily the vanishing volume of cells around the bends. And because of the approximations used in the algorithm, edges may cross over one another and the volume will become negative because of high skewness. (7). So, if your geometry is like a "Lego" structure, or a "Space Station" structure, then the Hexa, multi-block topology will give you good mesh automatically. Otherwise, the mesh will be studied on the one-to-one basis, related to its geoemtry and complexity. (8). There is a good book on mesh generation by Thompson, free on Internet. You should try to download it and read it. (There are professors, who are spending their life time doing one type of mesh generation. So, the task is not trivial.)

Axel Zinser December 12, 2000 06:41

[...] I would strongly suggest that: you send the file (by e-mail, if possible) to the support engineers [...]

Just curious.. why do you suggest to use e-mail instead of ftp?

John C. Chien December 12, 2000 08:34

(1). It ended up that way. And e-mail proved to be easy on the user's side. (2). It has something to do with the average user's working environment. (3). So, e-mail was used to send the files back and forth, with no problems for the user's side. (4). If you like to use ftp, and can use ftp with no problem from user's side, then, you are free to use it. (connecting one computer to another outside computer is always a no no for me. Ideally, one should not even send the file to other company at all. Anyone sending the file to outside company must remember to get the internal approval. But that is something else.) So, it is a matter of convenience. That's all.

J Keays. December 12, 2000 10:37

Hi John,

I wonder if you could tell me the internet address of the site where Thompsons book is located? Is it mentioned in the section on Literature in this site?/ Thanks..

Brian Gribben December 12, 2000 13:25

The url for the book is

A link to it is provided under "Mesh Generation" on this site.

The book is excellent, although Thompson's more recent publications are perhaps an easier read. However, putting the entire book where everyone can see it for free has to be applauded!

John C. Chien December 12, 2000 15:40

(1). Yes, I agree with you. Only the master has such vision. (2). The book used to be hard to get and was rather expensive.

All times are GMT -4. The time now is 14:14.