Currently i m working on a project for my Ph D at Faculty of Aerospace Engineering and i've come accros a problem related to a CFD modeling. The porpose of my project is to make a comprehensive study of air flow in transonic and supersonic aircrafts compressors. So i was told that finite element based CFD softwares can not provide accurate results at high Reynolds numbers. Is it true?

Hello Radu,

Yes, what you heard is true! Despite of the fact I started studying CFD with the Finite Element Method for flow simulations, I gave up after the tests with transport scalar. To begin with I developed some CFD programs (Euler 2D/3D) using the Finite Difference Method and latter I used the Finite Volume (Navier Stokes 3D). Right now I am a fan of the Multidimesional Upwind Residual Distribution schemes. Between the disadvantages of "classical" FEM I would mention:

- the difficulties to stabilize the scheme for advection dominant problems, - the high cost (i.e. computer time), - the requirements in memory (for the large grids you want to have to resolve the important details of your turbulent flow transonic/supersonic - shear layers, shock waves, mixing layers, etc).

If you still want to give it a try, start with the DG (Discontinuous-Galerkin) method. It is "cheaper" than the "traditional" FEM method and more accurate (with shape functions of sufficient high order) than the classical second order Finite Volume methods. If you need further informations please send me e-mail at dc@mail.vok.lth.se.

Best regards,

Doru

(the exact translation in Romanian) Salut Radu,

Da, ce ai auzit tu auzisem si eu! Desi am inceput CFD cu metoda elementului finit pt curgeri, am renuntat dupa testele cu transport scalari.Pt inceput am dezvoltat niste programe cu diferente finite (Euler 2D/3D) si mai apoi cu volume finite (NS 3D) (acum sunt un fan al metodelor de upwind multidimensionala). Printre dezavantaje pt FEM as vedea dificultatile de a stabiliza schema pt curgeri dominant convective , costul de calcul si necesarul de memorie interna, daca vrei cu adevarat sa ai nivelul de rezolutie al curgerii care sa-ti "captureze" ce este mai important in curgere (zonele cu gradienti mari-straturi de forfecare(strat-limita, zona de amestecare), unde de soc,etc.). Daca vrei sa incerci, studiaza metoda DG(discontinuous-Galerkin). Este mai rapida ca FEM, si este in general mai precisa (dar mai costisitoare) decat FVM. Daca vrei mai multe informatii, scrie-mi la dc@mail.vok.lth.se Toate cele bune, Doru

(PS. transmite salutari d-nilor profesori: Berbente, Stanciu, Galetuse, Danaila, etc. Dansii mi-au fost profesori si imi face placere sa mentionez asta)

No, it's not, but getting accurate results for the problem you've mentioned with any kind of CFD software would probably make a nice PhD diddertation in itself.