CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   Main CFD Forum (https://www.cfd-online.com/Forums/main/)
-   -   structured mesh (https://www.cfd-online.com/Forums/main/8497-structured-mesh.html)

rajesh December 30, 2004 05:30

structured mesh
 
hai all

can anybody give the difference between structured mesh and unstructured mesh

thanx in advance

rajesh

Hrvoje Jasak December 30, 2004 05:52

Re: structured mesh
 
On a structured mesh you can move with a compass: every cell has got East West, North, South, Top and Bottom and they are always ordered: East to West, Top to Bottom etc. Also, cells are numbered such that from the current cell index you can calculate the neighbouring cell indices.

On an unstructured mesh, the compass does not work - in order to find out the neighbouring cell index you need to use a look-up table.

Hrv

rajesh December 30, 2004 06:00

Re: structured mesh
 
i am beginner, what is a lookup table

i am using ICEM hexa CFD for Meshing and StarCD for analysis


steve December 30, 2004 10:52

Re: structured mesh
 
Because of the inherent order of a structured mesh, a cfd solver can be written using less memory and probably a bit faster than one written for unstructured meshes. With an unstructured mesh, the solver needs additional arrays (a lookup table) to keep track of who is connected who. Most modern commercial solvers are written for unstructured meshes so even if you feed them a structured mesh, they can't take much benefit from that anyway. Unstructured meshes are popular because they result in fewer restrictions on the mesh so they usually make the meshing process easier which, in most people's minds, is worth the use of a little extra memory a little extra time during the solution.

noName January 2, 2005 12:40

Re: structured mesh
 
I recall from a CAD class many years ago ...

A structured messh is one in which grid lines from orthogonal edges are exactly orthogonal. They may be subsequently mapped to circular / oblique edges to generate oblique or skew lines.

Any other mesh is unstructured.

Please correct me if I am wrong.

steve January 4, 2005 07:18

Re: structured mesh
 
Structured has nothing to do with the shape of cells - it has everything to do with topology - how they are connected. You can have all sorts of lousy shapes but as long as they are connected in basically one big brick, (ie all hex, all cells have exactly 6 neighbors,...) its a structured mesh. The brick may be distorted but it is still a brick. In the earliest schemes all cells where exactly rectangular and you had to model a curve with a series of stair steps. Eventually body-fitted structured meshes were developed which allowed curves to be followed, but the topology was still that of 1 big brick.

Chris Bailey January 19, 2005 11:53

Re: structured mesh
 
I think structured means a regular pattern of cells such that you can move from cell to cell in the same way for all cells. So it's topology, how they are connected, yes. But it's also shape to the extent that some shapes can't be regularly connected. Rectangles (in 2d) or rectangular solids (in 3d) can be regularly connected and can form structured meshes. But I think structured meshes also include, for example, triangles (2d) as long as they are topologically regularly ordered.

As a generality, the atomic structure in crystals is an example of a structured mesh, and anything this can be warped into without breaking or tearing it is also a structured mesh. And of course there are lots of different crystal patterns.

Did I get any of this wrong?

Also, rajesh, a lookup table in the context above is something used inside computer programs to relate things to each other. Some "column" in the table can be regularly ordered, and another "column" can point to whatever you want, such as a neighbor of the cell you looked up in the first column. It'd be a way to keep track of how cells are organized even if they are in an unstructured mesh in which neighbors occur unpredictably.


All times are GMT -4. The time now is 14:59.