k-omega or k-epsilon ?
 

I am simulating a fsae car. First I have simulated it with the realizable K-Epsilon Two-Layer with exact wall distance and two-layer all y+ wall treatment. Second i have simulated the car with the K-Omega SST Turbulence with exact wall distance and all y+ wall treatment.

My y+ goes from 1 to about 20 and i took the exact same mesh for both simulations.

With the k-epsilon i got the following values: Cl=-0.575 (Fl=-191.65N) and Cd=0.372 (Fd=124.05N)

And with k-omega i got: Cl=-0.524 (Fl=-174.49N) and Cd=0.358 (Fd=119.35N)

To that i have to say the the convergence of the k-epsilon was better it converged after about 1000 iterations and the k-omega doesn't really converged after 4200 iterations.

I have read, about different models and every it says that the k-omega is better for seperated flows and so better for getting the right lift and drag values.

I don't know which model should know be right or better..

Thank you

Hi Dennis,

You are right, k-omega model is better for separated flows, should be used at high velocity, needs more number of prism layers for near wall resolution.

However I recommend you to use k-e with all wall treatment model for fsae applications since fsae cars don't traverse at very high speeds.

Ok and what is the best range for the y+ with k-epsilon all wall treatment model to get lift and drag values that are near the real values in a windtunnel.
I have read different ranges for y+ with this model...
And how would you set the prism layers and mesh size for best results?

Thank you

Hello Dennis,


Regarding Y+ values, I suggest you to read this thread:

http://www.cfd-online.com/Forums/sta...-question.html

Regarding Mesh size, I cannot really tell you until I understand how your CFD domain is, what features you are looking to resolve so on and so forth.

I am simulating an external flow and want to get the lift and drag values. Moreover I need to know where the centre of the pressure force is. Also I need to know when the flow separates of a wing and I need the pressure around the car.

I think because of the lift and drag values I have to set a fine mesh and a y+ around 1. Or is that wrong?

I would recommend k-omega SST over k-epsilon for external aero applications like this. Separation prediction is better. For important surfaces (wings, tires if you can afford the mesh, floor plates) use y+=1. Otherwise stick to 30<=y+<=100 or so.

On important surfaces you will want at least 10 prisms or so, more if you can afford it. On less important things you can stick with 5 or so.

FSAE cars are somewhat small, but a mesh of ~20M cells or so will do okay. More if you can afford it is better, obviously. Aero simulations are very expensive, so you need to do what you can with what computing power you have.