Turbulence Model, truck aerodynamics
 

I am running simulations for external aerodynamics for trucks. Until now I have run RANS simulations using the realizable k-epsilon turbulence model. I am currently looking in more detail into the flow field around the cab side extenders and the roof deflector, also what is happening downstream if these are removed. I am thinking of trying the k-omega SST model. My thought is that this model perhaps would be less "viscous" and would not tend to smudge the flow field as much as the k-eplison models. What are your experiences regarding this dilemma? Can anyone please give me some advice about an appropriate turbulence model for this application?

There's a conceptual difference in how k-w approximates the boundary layer if I remember well (using wall functions or not). In my experience, k-e converges much better. Since you're not interested in the boundary layer, I'd run it with k-e first.

Anyway, simulating these highly separated flows in CFD will never give you a proper result, sorry! And if it does it's a pure coincidence. No matter what is being said on this forum by CFD believers. (It will give you good resutls in neat attached flow nevertheless)

You will have to run windtunnel measurements to see the difference or start driving with both options and measure your fuel use. A long and tedious task (both of them actually)

There's a great book on that subject you're working on:
"Aerodynamics of Road Vehicles: From Fluid Mechanics to Vehicle Engineering" from Wolf-Heinrich Hucho. Reading that can safe you a lot of work in options that are destined for underperformance.

Thanks for your reply, gerritgroot!

Yes I am aware of the limitations of turbulence modeling. Comparing to wind tunnel tests or road tests are of course a good alternative to confirm the results obtained by CFD. Thanks a lot for your recommendations and your time.

The SST k-omega model tends to be less diffusive than the Realizable k-eps model (Shih's model as implented in Fluent I assume you mean). Which model works best is dependent on the application. The SST k-omega model is more separation prone and predicts more secondary flow with less overall diffused results. This can be good in some cases, but it can also mean unrealisticly large wakes and separations that occur much too early. You need validation for your application to select between these two models. Both are good but both have weaknesses.