hi everyone. my question is about Ahmed Body. have you ever used this 3D problem for your validation cases. my problem is the material properties of air. my drag coefficient is %30-40 higher than the experimental value. I used ro as 1.14075 kg/m3 (@ h=426m T=20Celcius). the height is the geographical altitude of the Gottingen city that the original experiment conducted. but there is no info about material properties. does anyone interested in this topic?

Hi,

The Ahmed test body is a very complicated flow and it is not easy to get accurate drag results for it. Depending which body geometry you use you need either RANS or LES/DES to get an accurate solution.

Regards, Glenn

I'm using a 3D model with approx 1million cells. I tried different back angles. I'm using fluent and very accurate results are found with fluent in 2002 with SST and realizable k-e model. the experimental drag coefficient is 10 degree Cd=0.230 I found 0.280 RKE y+>30 standard WF 25 degree Cd=0.285 I found 0.316 RKE y+>30 standard WF 30 degree Cd=0.378 I found 0.345 RKE y+>30 standard WF 35 degree Cd=0.260 I found 0.350 RKE y+>30 standard WF

bound conditions

velocity inlet with 60m/s, Turb In. 0.5%, 3.077m Hydraulic Dia pressure outlet with 0Pa gauge pressure (modeling open test section) , a symmetry plane and the rest is wall of course. The engineering working in fluent distibutor told me there isn't any problem in my model so I tought that the problem is in the material properties or smth else.

I know the physics of ahmed body very well but neither of the RANS models can capture the separating and reattaching of the flow for 25 degree back angle case. LES is too computationally expensive I don't know what to do

there is a pdf document about accurate fluent results that I mentioned above.

http://rapidshare.de/files/14797064/...GM-02.rar.html pages 34-35

We have doubt about CFX Reynolds Stress Model anyway. We try a couples of test cases and RSM gives the worst prediction. Don't spend too much time to try RSM unless time is not your concern.

For an SST model your Y+ should be a lot less than 30, approx 2-3.

I often found that steady-flow solution becomes very difficult to converge if you set Y+ too low. I'll suggest a mesh dependency test to check the mesh sensitivity.

but if your y+ is too big then you will not be running an SST model, more like a K-E model.

Hi.... Highly refined mesh leads to numerical instability as stated in user manual. I don't think SST & KE are the same....I run some test cases with SST & KE (average Y+ around 49) and found that solution obtained from these two models are very different.

From my understanding, the advice given by user manual is to put at least 15 nodes in the boundary layer if kw based model is used. Under "Guidelines for Mesh Generation", it stated that "kw models do accept coarser meshes, due to automatic near wall treatment for these models".

Yes. The general requirement for low-Re model is that Y+<=2 but this is often difficult to apply for 3D body.

A mesh dependency test would be a good way to check the solution sensitivity.

Hi,

TB is correct:

Highly refined meshes do cause numerical instability, but you are talking about y+<0.01 and that should not be a problem here.

SST uses a k-w like formulation in the boundary layer, transitioning to a k-e formulation in the main flow. SST also contains additional terms for curvature correction so you would expect it to produce different results to k-e.

For a flow like the Ahmed body where the boundary layer, separations and reattachments are important then you will need a mesh with a low y+ (1<y+<5 at a guess, check with mesh independance) AND at least 15 nodes in the boundary layer AND a reasonable mesh transition from the prism layers to the tet mesh.

Glenn Horrocks