[pilot.john]

Hello everyone

I am currently investigating the effect of slant angle changes on the drag coefficient of an Ahmed Body using Simcenter Star-CCM+.

At this stage I have to validate my sims with the original paper on the Ahmed Body (linked below).
My problem is that my Drag cd values are quite high compared to the experimental value of 0.290 for a slant angle of 25deg according to Ahmed’s paper.
I get for k-e 0.377-0.368 and 0.38 for K-omega. I aim to get to 10% error.

I am a CFD beginner and I also somewhat limited by PC’s 4 cores, so I only use a model split in half.
One of the problems I have is that parameters such as temp are not stated in the paper, so I assume 1.225 kg/m^3 and 1.79E-5 Pa-s for dynamic viscosity in order to match the paper’s Reynold’s number of 4.29 million. I use the same wind tunnel dimensions as Ahmed’s paper, using volumetric controls I have made a finer mesh around the AB where most of the interesting features happen with a 4mm custom size.
I have included my sims and some photos of the mesh (using hexahedron mesh).
https://drive.google.com/drive/u/2/f...CxQIDoC7gX8uT1
If you could help me or give me some suggestions I would be very thankful.


Some Salient Features of the Time -Averaged Ground Vehicle Wake
https://www.jstor.org/stable/44434262?seq=1

[Skylar]

I just saw this post for the first time. I'd suggest checking out previous CFD papers like "Assessment of hybrid RANS-LES formulations for flow simulation around the Ahmed body" by Guilmineau et al. (2018).

If you're limited to only four cores, simulating the flow over an Ahmed body will be difficult if you want an accurate solution. First, it's been well documented that RANS models like k-epsilon and k-omega SST overpredict the separation on the rear slant for bodies with slant angles below the critical angle of 30°; Guilmineau has another paper where this is shown for a 25° Ahmed body in particular. You'll need to use something like LES or DES for a more accurate prediction, and these are much more computationally expensive due to their grid requirements.

Second, the Ahmed body's wake is inherently 3D, so you need to simulate the flow over the whole body. Using symmetry along the centerline restricts the spanwise velocity - this is not physically accurate, especially if you have to use a model like LES.

Third, make sure your mesh is sized appropriately. In particular, from the paper I named earlier, the Taylor microscale and viscous sublayer thickness are two important parameters for this. Make sure your timestep is appropriately selected as well. The theory guide and user's guide for Star-CCM+ should have guidelines for different turbulence models - check those out.