I am trying to carry out a 3D simulation with a Reynolds number of 5000 and a Prandtl number of 10.
I am using a double distribution function approach, and I couple the thermal and velocity lattice.
I also use the Smagorinsky model for turbulence modeling.
The limit on the lattice velocity and lattice Mach number for considerably refined meshes, derive in small fluid relaxation times.
However, the situation gets worse as the thermal relaxation time gets very close to 0.5 due to the large Prandtl number. This causes large instabilities.
I am using a D3Q27 and a D3Q7, and I have tried with MRT and BGK.
Has anyone experienced this issue and found a solution to these types of simulations?
You can use the UnitConverterFromResolutionAndLatticeVelocity with lattice velocity between 0,1 and 0,15. That will give you the highest possible relaxation time in the chosen resolution. Then look at SmagorinskyBoussinesqCouplingPostProcessor3D. You can stabilize your simulation there with the turbulent Prandtl number.
Thank you for your reply on this.
I am doing exactly what you recommend, keeping the lattice u at 0.1 and adding an extra tau_advection as a function of tau_fluid and turbulent Pr number. I have also tried refining the mesh of course.
I feel like the double distribution formulation hits a limit at a sufficiently high Reynolds number. For example, at a Re 10,000 it is very hard to simulate Prandtl numbers around 10. The thermal field gets too unstable.
from my experience the turbulent Prandtl number must be between 0,3 and 0,7. By coarse resolutions even smaller than 0,3. If you use this thermal coupling postprocessor the standard BGK collision must be also enough. Dous your dynamics use the relaxation frequency from lattice cells or the global from parameters?