Hello, I'm looking for references on the implantation strategies of turbulence model (k-eps) in finite elements. Thanks for help ! Ben

Where do you start from ?

the code NSC2KE is a 2d NS code using k-e w/ finite elements. you can find a link to the code and documentation at http://capella.Colorado.EDU/~laney/softaero.htm

http://www.featflow.de ?

Hi, I know the finite elements : I work with a code for viscous steady and non-steady flows I know about k-eps, and I want to introduce it in my code. I'm looking for programation strategies, algorithm, practical know-how Thanks for help, Ben

Dear Ben,

I managed to program a 1D resolution (no simple algorithm) of the incompressible k-eps with finite differences for linearized custom equations with additionals sources. I was rather confident: I had yet solved analytical first order turbulence model (and spent weeks to relax it).

However, k-eps convergence appeared as a more relevant tricky business: my code still not converges. So, I wonder who can help you: programation strategies have been argued to depend on many conditions to me, herein. I could see various tricks within open sources codes, but these could not help me enough.

For finite elements, if read a preprint (Lee and Leone, 1991: "Numerical Modeling of Turbulent Dispersion Around Structures Using a Particle-in Cell Method", available as a pdf from the Lawrence Livermore National Lab.)

Paper refered to some 'clipping' method (used by the FIDAP code ? ), whose purpose is to avoid negative values for k and epsilon, and not too large nu_t values, but this is rather poorly documented, as are the various others tricks (e.g.: nu_t = C_mu*k^2/(tiny+abs(eps)), instead of: nu_t = C_mu*k^2/eps) involved within CFD solvers. I implemented clipping, as well as various unsuccessfulls relaxations schemes...

I planned to summarize tricks for k-eps, some day, so as more and more people can develop open source codes, within buying some more book. If you plan to go this way, please, do not embed pre- and post processing, it would be easier to understand. You may find links to postprocessing software within the forum.

For pre-, post-, as well as 'proto' processing purposes, high level interpreted languages toolboxes (e.g.: Scilab may incorporate finite elements utilities) may help, at computing time cost, however. For research purposes this shouldn't hurt (much).

Own opinion(heard this as open source developpers policy): consider starting from an existing software provided by some of the links below. This way, you may spare pain re-using a maximum of availables (if suited enough) parts.

Hope some Doctor will let us know more