[vachan]

Dear all,

I am using flexi (https://www.flexi-project.org/) for a supersonic turbulent flow simulation. In my case, it turned out that most cells lying on the walls required limiting and hence were converted to FV subcells.

When I plot the wall friction results and compare with other simulations, I find my results slightly off at some locations (pic attached). Also the variation shows a staircase-like behaviour, with non-smooth transition between elements. Pressure variation matches well (pic attached).

I am suspecting this is because of the wall cells being converted to FV subcells. Is this really the case? Or does flexi give the correct wall friction/heat transfer results even with wall cells being converted to FV subcells? Is there a preferred and correct way to compute these wall quantities (that depend on solution gradients) in case wall cells require limiting?

Thanking in anticipation
Vachan

[mblin]

Hi Vachan,

since FLEXI is a DG code you will almost always see small oscillations in the boundary layer indicating that you are a bit under resolved.

However, having FV cells directly next to a wall is not that preferable since the quality of the solution is not as good as for DG elements. Thus, I would suggest that you try to tune your indicator to remove the need for FV elements in the boundary layer.

You can compute the wall quantities by either using the posti_visu tool or by using the BL_Plane in the record point tool (see NACA tutorial).

Hope this helps.

Best
Marcel

[vachan]

Dear Marcel,

Many thanks for responding!

since FLEXI is a DG code you will almost always see small oscillations in the boundary layer indicating that you are a bit under resolved.

I agree! I am currently trying to run a finer grid simulation using the coarse converged solution as the initial condition. This I am planning to achieve through posti_swapmesh tool. I am facing some problems here too, which are documented in this thread Usage of flexi's posti_swapmesh tool.

However, having FV cells directly next to a wall is not that preferable since the quality of the solution is not as good as for DG elements. Thus, I would suggest that you try to tune your indicator to remove the need for FV elements in the boundary layer.

Again, you are correct. I currently use Persson's indicator with density variable. When I tried to use pressure the simulation was facing starting issues. I also tried to restart the converged solution with using pressure for indicator, but that still didn't make the boundary cells fully DG. I guess this issue cannot be solved separately and is linked to the resolution of the boundary layer. Do you have any other suggestions in this regard?

You can compute the wall quantities by either using the posti_visu tool or by using the BL_Plane in the record point tool (see NACA tutorial).

Noted. I am currently using posti_visu and exporting the data in a csv file from ParaView.

Thanks again!
Vachan