What is the best model, if there is one, to model airflow on a openwheel car? the Spalart Allmaras seems really more efficient and stable than the k-epsilon but are the results believable? (vortexes...)

Thank you.

Dear Greg,

unfortunatly there is no easy answer to your question. Yes the SA model is very stable because it is a one equation model and probably over-predicts eddy viscosity which stabalizes the effect of the turbulence model on the RANS equation solution. The S-A model is well validated for external aerodymamics of streamlined bodies like airfoils at low to moderate angle of attack.

Another very stable model that is used a lot with Fluent is the Realizable K-epsilon model. It is a bit more expensive since it is a 2-equation model and also can tend to over-predict eddy viscosity.

Have you done a literature search of similar problems? A good place to start is with some of the validations on Ahmed's body and other test cases. It also depends on what you are interested in knowing? Overall drag? Local mean velocities? Transient info like vortex shedding frequency or noise, etc?

Often other factors like geometry details which must be neglected, domain size, mesh resolution, uncertainty in boundary conditions, discretization errors, etc. can contribute as much to the absolute error as turbulence model.

I would suggest you do the literature review and choose one (maybe 2 turbulence models) which you (or other analysts) have found to be the best suited for similar applications for the desired results. Stick with the same turbulenc model for your studies and spend some time evaluating the other factors as well.

Let us know how you make out.

Regards,

Bak_Flow

Hey,

Actualy I have to improve the downforce of a openwheel car. I just want to be sure all the vortexes which would happen in reality are present in the model.

I cannot have a better mesh resolution because of my computer memory.

My model works well when ground is moving but wheels are stationary, but when wheels rotate, with all viscosity model I have turbulent visosity since my first iteration and results diverges quickly. I have never seen a model converge after 500 iterations when it begins to diverge after 3... but maybe I am wrong.

I thind that for similar applications the Realizable K-epsilon model 2nd order is the most used.

Quite interesting paper: http://www.tech.volvo.se/ivs/courses...p_pk_jl_sp.pdf