Numerical dissipation/difusion in LES
 

Dear All,

Can any body talk about numerical dissipation or diffusion in CFD? what's the difference between these two? What's the numerical scheme used in FLUENT 5 for LES? is it highly dissipative?

And if finite-volume method and second-order temporal and upwind spatial discretization are employed, how to deteremine the numerical dissipation?

In large-eddy simulation, how to avoid the subgrid stress being overwhelmed by numerical dissipation/difusion?

Thanks in advance.

Ray

Re: Numerical dissipation/difusion in LES
 

Hi Ray,

Dissipation and Diffusion are the same thing. Error that smears out sharp gradients. Dissipation is therefore associated with even order truncation terms usually i.e. a 2nd order derivative in a first order upwind scheme.

I don't know about FLUENT's implementation but LES should be used without any dissipative error - i.e. a central type scheme should be used. These schemes are dispersive i.e. error is spread in a wave like fashion which is evident as spurilous oscilations in the vicinity of steep gradients. These schems have a odd-order leading truncation term.

Error terms can be determined by writing the numerical scheme out as a taylors series expansion and looking at the largest term that has been truncated. You will have to do this both space and time.

Another option is to perform a grid and time step convergence study.

All the best, Steve

Re: Numerical dissipation/difusion in LES
 

Many thanks! Steve.