Hi All !

As far a i know, there are two kind of CFD-solver : - based on FEM-discretization - based on FVM-discretization

The leading Software are based on FVM.

Is there a reason for this ? Which are the disavantage of FEM-solver for CFD ?

Thanks ! JFS

This question was discussed hundreds times in this Forum. Look archive.

I guess one of the main reasons FVM is the most successful and preferred is its rigorous conservation principle, even for course computational meshes. It is also conceptually much simpler to understand than FEM.

Hi JFS,

There is also the element based control volume method.

Robin

hi i also read that FEM based methodology is more reliable in case of stress related problems and diffusion related problem, but not that good in case of covection problems, maybe that is one of the points

Hi Robin,

Do you know any reference about the "Finite Element Based Control Volume Method" (Book, article, site)? I think CFX uses this technique, doesn't it?

Thanks a lot.

I doubt the strong conservation laws being satisfied locally in FV.

to my understanding, they are implicitly satisfied. I am not sure. I will carefully look through the whole process. an implicitly satisfication deos not mean bad, but could.

now here is a story about projection finite difference method when pressure poisson equation being used.

if the intermediate velocities ( explicitly got from a portion of mometum equations) is not divergence free on boundary, I claim the projection FD loses one of its good properties: strong form.

therefore, many FD algorithm which are pretty popular are not in strong form at all. those weak and non locally conseravtive FD perform as bad as FEM for incompressible N-S, which is a weak form and pays no respect to local conseravtions.

so far there exist two types of locally conservative FEM.

one is Control-Volume FEM whose having been investigated dated back late 1970s I guess.

the other is dicontinuous FEM ( so called Discontinuous Galerkin, which is not a good name, if I have a discontinous Petrov Galerkin, so what? ). here I regard spectral element method as an advanced version of FEM, and correspondingly I regard discontinuous spectral element method as an advanced version of discontinuous FEM.

In the future, I'll show how local conservation is important for solving fluid problems. I have logically sound arguments and compelling evidences.

Does any one have commerical FEM experience solving fluid dynamics? Or any comparisons with those packages? Thanks!