[Kiffer]

I wonder the difference between the structured grids and unstructured grids on the effect of CFD accuracy. Thanks a lot.

[leflix]

Originally Posted by Kiffer

I wonder the difference between the structured grids and unstructured grids on the effect of CFD accuracy. Thanks a lot.

This is very easy to explain whith a draw and no so with words.
Each time you can refer to a cell or a grid node with two indices(in 2D) like (i,j) it is structured. When you have a node refered by (i,j) you implicitly know that its neighbours are (i+1,j) ,(i-1,j), (i,j-1) and (i,j+1).
For unstructured grids each node or cell is not refered by two indices, but only one let's say l. And the fact of knowing the indice l does not tell you which are its neighbours. You need a link table which gives you such needed information. Who is the neighbour of who, and what are the vertices which define a given cell. Typically triangle mesh in finite element method is unstructured.

Theoretically if things are correctly done, the solution should be independent on the type of grid. The discretization of diffusive flux is the key point in such stuffs especially when the cells are not orthogonal. Note that you can have structured grids which are not orthogonal and unstructured grids which are.
An other point which matters is that the matrix or the linear system obtained after discretization on a structured grid has a multidiagonal structure which is very favourable for most of the linear system solvers available. However for unstructured grids the matrix has non zero coefficients distributed "every where" and the linear system solvers are less efficient. A lost of accuracy or an increase of cpu time may be expected.
Finaly codes based on unstructured grids cost also more in memory management.

[sbaffini]

This PhD thesis has a small part related to the topic:

http://powerlab.fsb.hr/ped/kturbo/Op...JureticPhD.pdf