code_saturne User's Forum

Hello, everyone

Recently, I use code_saturne 3.0.7 to simulate backward facing step flow. The Reynold nember is about 36000 base on the height of the step. The more information can be found in NASA turbulent model website(https://turbmodels.larc.nasa.gov/backstep_val_sst.html). The turbulent model used in this case is SST turbulent mdoel. The height of the step is H=1. The distant between step and inlet is about 182H, so at the -4H location before the step, the thickness of boundary layer is about 1.5H. The picture in attaachment is the formulation to initialize k and omega, and inlet boundary condtion for turbulent model is set constant which same as the initialization. The mesh, xml file and result can be found in attachment. In x=H after step, the result isn't correct compare with the experiment and CFL3D. So, how to simulate this case correct? Can anybody help me? Thank you!

Best regard

Hello,

Could you also post your "listing" file ? Also, your y+ values are very small, so the mesh might be too fine for this model. You may also try the "scalable wall functions" advanced option.

Was the same mesh used for CFL3D ?

Regards,

Yvan

Hello Yvan,

Thank you for your reply. Follow your advice, I do some modification and upload the new result.

First, I generate a little more coarse mesh from Salome 7.8.0 by myself. The mesh from NASA website is not used for this case because its too finest. The location H=-4 is the inlet of the backward-facing step. I use a long pipe before the inlet to generate more accuracy inlet boundary condition. If I use two scale wall model, through some tests, I find give the uniform velovity proflie at the inlet of pipe(H=-182), the turbulent plate boundary layer will have 1.5H thinckness at H=-4. But the result seem improve little. As the corner eddy under the primary sperate eddy is too large at x-direction, the velocity profile at H=1 is incorrect compare with experiment values. If I switch the wall function to scale wall function, the inlet velocity profile at H=-4 is uncorrect. And change the length of pipe, I can't find the correct velocity prolife.

The new initialize formualtion for k and omega is from Menter SST AIAA paper.

Is the new mesh still too finest(I can't find any warning about the mesh)? What to do next to get the correct result. Thank you.

Best regard.

Hello,

I'm not too sure what may be causing this. It would be interesting to see if you have the same issues with Code_Saturne 4.0 or 4.3, as some turbulence model improvements may have been done (esp. regarding boundary conditions), and they are closer to the version on which model fixes can occur (if necessary).

Regards,

Yvan

Hello,

It is still not clear to me if the same mesh as your finest mesh was used for CFL3D. Your new mesh is still much too fine near the boundaries, as your y+ is much below 1, which is usually the target for low-Reynolds turbulence models. Since k-omega should work also at higher Reynolds numbers, a CFL of 1 would be a minimum.

It is hard for me to find physical property in input velocity in the CFL3D input file, but a different velocity, viscosity and/or density (i.e. Reynolds number) could also change the y+ values, so you may want to check this (in particular, the dynamic vs cinematic viscosity could be tricky).

Regards,