April 19, 2019 at 2:33 pm #4255
I am currently studying gas-liquid flow patterns using the lattice-Boltzmann method. Since both phases are expected to attain turbulent regime (Re around 1e5) in a range of situations, I am using the LES-Smagorinsky model, coupled with an MRT collision operator, for both liquid and gas phases.
The code works, but stability is a major issue. I have the impression this is mostly because relaxation frequencies are way too close to 2, but high velocities (>0.2) occur as well.
My latest ideas, which I will try soon, are:
– Using an equilibrium distribution function up to o(Ma^3) or o(Ma^4), to reduce trouble with high velocities.
– Using regularized-MRT dynamics.
I would greatly appreciate any suggestions on how to increase stability. Any useful references or papers on turbulent flows with LBM are welcome, too.
Thank you for your replies!April 25, 2019 at 9:38 am #4269
I just can guess, because I don’t know your simulation setup. If you want to simulate such a high Reynolds number, you should respect your grid spacing.
How do you estimate your grid spacing?
Do you have a wallbounded flow, if this is the fact what is your y+?
If you grid is to coarse you have to deal with a small tau, which could lead to instabilities.
Do you know where the instabilities occur? (at the boundaries, at the gas liquid Interface…)
You lattice velocitiy is too high, you should avoid lattice velocities higher than 0.1.
-increase your grid size to get a stable simulation
-estimate your smallest scales and adapt your grid spacing again to get physical meaningful results
Best MarcMay 2, 2019 at 10:37 am #4298
Thank you very much for your reply. Mi simulation setup is the following:
– Geometry: straight, cylindrical pipe in 3D.
– Boundary conditions:
– Inlet: plug velocity profiles for liquid and gas, which enter separated (liquid in the lower-half of the inlet, and gas in the upper half). I use a ‘velocity-BB’ rule like the one implemented in OpenLB.
– Outlet: constant pressure, zero gradient for velocity, using extrapolation rules.
– Walls: no-slip condition via full-way bounce-back.
Resolution and time steps are estimated so that velocity at the inlet is lower than 0.01, and so that the smallest tau of the two phases is higher than a certain value, admittedly very close to 0.5. Increasing grid size is of course always convenient for stability, but I have not found any significant improvement upon using very high resolutions while respecting the stability restrictions.
Contrary to what I thought last week, turbulence does not seem to be the main problem since the model works well with high Reynolds numbers in single-phase flows, even with such small taus.
In fact, following your suggestions I have found that instabilities occur at the interface. Given that I use a free-surface VOF algorithm that is somewhat cumbersome, I will search for possible sources of instability there.
Again, thank you for your answer. Your idea of knowing where exactly instabilities occur has been very helpful.
You must be logged in to reply to this topic.