Curvilinear boundaries
 

Hello, I write my Fortran-program for the 3D incompressible Navier-Stokes equations on the multiprocessor computer (cluster). Different ways of dealing with curvilinear boundaries are known: 1) the generalized curvilinear coordinateses; 2) non-uniform cartesian grids; 3) multiblock uniform cartesian grids which take more finer near boundary. What way is better? What grids (structured or unstructured) it is better to prefer? Thank.

Lew.

Re: Curvilinear boundaries
 

hi that depends on your application, also time available , your understanding of the variety of methods ! Obviously the unstructured grid methods are most flexible in handling geometries but the coding in 3-d is most complex. Don't go for orthogonal grids which refine near boundaries.Non-orthogonal curvilinear grids (multi block)are the best compromise but u have to see whether your geometry is mappable or not.Still this is the best possible approach for most of the people.