code_saturne User's Forum

Hello Team,

I am trying to simulate the large eddy simulation for the vertical pipe flow case with the dimension of (diameter of the pipe = 0.25m) and length (0.465m) for the Re = 4000. The liquid metal (LBE) is the working fluid (density = 10465kg/m3; kinematic viscosity = 2.29*10^-7) as per the Reynolds number calculation the velocity is around = 0.0037m/s. The inlet boundary condition has been applied at the pipe inlet with the velocity value of 0.0037m/s. The free inlet/outlet at the outlet of the pipe and wall condition at the pipe surface. The structured mesh that has been used for the simulation (please find the attached file_mesh_topview). In the results I found the mesh artefacts throughout the pipe (please see the attached file; velocity_liquid_metal).
I have tried the same case (Re = 4000) with air as a working fluid (density = 1.225 kg/m3, kinematic viscosity = 1.5*10^-5). The inlet velocity (u = 0.25m/s) has been calculated based on the Reynolds number calculations. The boundary conditions at the surface of outlet and pipe surface remain same (i.e free inlet/outlet at the outlet and wall at the pipe surface). The interesting aspect has been noticed that in that case there are no such mesh artefacts (please find the attached file velocity_air).
I have tried different gradient calculation method (for instance least square; iterative method of handling non-orthogonality), still there are artefacts in the liquid metal case.

Hello,

The artefacts seem much reduced compared to the mesh you had with a mix of tetrahedra and hexahedra, though there seems to be a small increase where mesh non-orthogonality is the strongest.

Is the velocity field an instant value ? Do the artifacts appear on LES time-averaged quantities ?

Best regards,

Yvan

Hi,

Yes the mesh artefacts are appeared on Les time averaged quantities as well.

Hi,
In the present case we used the large eddy simulation for Re = 4000 vertical pipe flow with liquid metal as a working fluid. A pipe flow case has the mesh artefacts. For that solution we used different gradient calculation method (for instance least square; iterative method of handling non-orthogonality), Reduce the pressure relaxation factor (0.3), increase the sweep reconstruction (10 for pressure) and so many factors. however, in each condition there are some artefacts in the results. Do we have other options to try in order to resolve the issue?

Hello,

No, I do not have more ideas f you have already tried all LES variants (Constant/Dynamic Smagorinsky or WALE), except explicit filtering, (which would require a user-defined function to recompute turbulent viscosity, and which has not been tried on our side to my knowledge).

If you do not have stability issues, I would avoid relaxation of pressure increase, as I am not sure what effect it has on the time scheme (I doubt the scheme is still 2ndorder in time with that option). So I would set that to 1.

Best regards,

Yvan

Hello,

I just thought of an additional test : could you deactivate the "improved pressure interpolation" option ?

Best regards,

Yvan

Hello,

Thank you for your constant support, In my case, I did not activate the pressure interpolation option.

Hello,

A colleague had issues on an LES case which might be related to your artefacts, though I ma not sure. He is also testing a few options to see if they have any impact.

I don't expect any news before the end of next week, but I'll keep you updated.

Best regards,

Yvan

Hello,

Another idea suggested by some colleagues : add a slope test and 2% upwind.

Best regards,

Yvan

Hi,

Slope test for velocity and 2% upwind (blending factor 0.98).