[Anna Tian]

Hi,

Recently I always heard about that there are more and more percentages of CFD user using unstructured grids. And there are less and less percentages of CFD user using structured grids. The unstructured grids (including hybrid grids), which are easy to generate (the most important advantage), will be the future trend as its technology becomes more and more mature and our computer becomes more and more powerful. Is this true? I'd like to hear about your opinion.

[jchawner]

Hello Anna:

Yes, there is a trend in CFD away from structured hex grids and toward unstructured grids of all types. And one of the main motivations is the reduction in mesh generation time; in general, it's faster to generate an unstructured mesh than a structured grid. Of course, the reduction in time has to be balanced by accuracy of the CFD solution; many believe structured grids provided the basis for more accurate solutions than unstructured grids. Perhaps a secondary factor to be aware of is that a structured solver (one designed specifically for structured grids) is generally faster and uses less memory than an unstructured solver.

A very interesting compromise is the user of overset structured grids. What makes structured grid generation difficult, as you know, is the creation of a multi-block topology wherein each block is a structured grid and they all abut with point to point connectivity. This can be very challenging for complex geometry, even if your solver can tolerate singularities in the grid (e.g. a face collapsed to a line). Imagine however that instead of the blocks abutting point to point they overlap. This would allow you to generate relatively easily a structured grid around all the geometry. The trick is to then compute how the blocks overlap and how flow solver data can be interpolated from block to block.

Of course, this is just my opinion.