Hi!

Can anybody point a basic reference on Overset Grids?

Thanks in advance!

Anton

Actually, I am also interested in the overset grid. I heard that it enjoys freedom of unstructured grid while having some of advantages from structured grid.

I have a book "Handbook of grid generation" by Thompson. Check the chapter 11. "Composite overset structured grid"

-Selina

Two of the more basic references that I have used in the past are

"On the Use of Composite Grid Schemes in Computational Aerodynamics", Steger & Benek, Computer Methods in Applied Mechanics and Engineering, Nos. 1-3, 301-320, 1987.

"On the Spatial and Temporal Accuracy of Overset Grid Methods for Moving Body Problems", Meakin, AIAA-94-1925.

Most of the textbooks in CFD that I have don't really spend a great deal of time covering overset methods and some of the subtleties involved. There are usually a number of papers presented at the AIAA conferences that utilize overset grids in various applications, and there is an overset grid symposium that meets bi-annually. The most recent one was held this past October - a Google search on 'overset grid symposium' will return a host of results such as

http://www.arl.hpc.mil/Overset2002/

Thanks Selina, Greg!

What I want to know about the overset grid methods are their advantages and disadvantages for the moving body problems. For example, what methods are used for interpolation and what are the computational expences comparing with rearrangement of the traditional "one piece" grid?

This is probably the best place to start:

http://www.llnl.gov/casc/Overture/

The best reference is probably the J.Comp.Phys. paper from 1990 discussing interpolation.

For moving bodies, overset grids make it very easy to introduce relative motion between bodies in flow. The individual bodies have their own sets of grids which are moved along with the body and no grid deformation or regridding needs to be considered (although it could be included within the overall framework - some work is currently being done on modeling aeroelastic problems using overset grids). So you grid once at the start and that's all the grid generation that is required. The price you pay is in the grid assembly process, where the individual grids are pieced togeether into a composite grid system and the grid connectivity information is established, and in the possible degradation of information passage across interpolation boundaries. The grid assembly process is not automatic, regardless of what anyone says. It has been made much better of late with codes like PEGSUS 5, but it does require a certain amount of experience to obtain good composite grids. Most of this experience really comes from knowing what kinds of grids to generate initially that will give good composite characteristics when fed into the assembly process. It's not difficult but it can be tedious and there is a learning curve. Interpolation is typically trilinear (at least for structured grids, which is primarily where overset grids are used currently. Some work is being done to extend the methods to overset grids). Because of this there has been an ongoing debate about loss of conservation at overset grid interfaces. Most of these difficulties can be avoided if the grid regions that overlap have similar cell sizes, but that again requires a little experience to be able to do well. There is also an issue with the propagation of signals across grids with disparate cell sizes, and problems of transferring gradients across interpolation boundaries, i.e. interpolating information from boundary layers. In short, overset grid methods are very robust methods for handling moving bodies if the practitioner is aware of some of the issues involved and can account for them.