Hi everyone. I'm trying to simulate a stirred tank with a rushton turbine (H=0.5m Diam=0.5m). I adapted the mesh many times, and now it's composed by almost 2M cells. I'm using steady MRF simulation with k-eps-rng turbolence model, with two zones, one cylindrical and moving, surrounding the turbine, and the outer one stationary. I'm using SIMPLE algorithm for pressure-velocity coupling, presto! for pressure discretization, and 2nd order upwind for the rest, with the default underelaxation values. I started off with 100 rpm and I'm trying to increase the rotational speed.

However I'm experiencing difficulties in continuity convergence. The residuals get stuck on about 5e-02 after 11000 iteration while velocity components k and epsilon drop well under 1e-03. The flow field looks very good though. Also the velocity magnitudes are consistent with the 100 rpm.

What should I do?. Help plz.

In my experience, adapting the mesh many times is not a good idea. I always got problems similar to what you describe (and more) after such operations.

I would recommend to improve your mesh in preprocessing. and then do minimal adaptation in the solution.

Thanks for the advice. I was wondering if, given the grid, there was something to change on the solvers that can help me out.

Hey Nycera, usually if your residuals are hovering at a certain point, it is because you have transient flow structures that can not be resolved with a steady state case.

The residuals are there to monitor your simulation. At the end of the day, it is your results that matter most. What I would suggest is to compare some quantitative results from a steady state case and a transient case using sliding mesh, and see how they differ. You will more often then not get more accurate results with a sliding mesh setup. But the question is how bad (or good!) your current results are with MRF compared to sliding mesh? If they are quite close, then you can be happy with your MRF results, otherwise, you might need to change to sliding mesh to improve your results.

Hope that helps.

Ok thanks for the tip. The thing is I'm simulating with high rotational speeds, so it's not exactly quick to get a result from a sliding mesh setup starting from stationary field. I'm running out of time and I just wanted to get an overview on mrf steady setup.

Another question: if I wanted to advance in time in order to get past the "starting transient" with sliding mesh and reach a steady state of the flow inside the tank, I should use a larger timestep right? Is that possible? How should I set up the solver in order to get convergence within 20 iterations per timestep? I read on fluent manual that the PISO scheme for velocity-pressure coupling is suitable when large timesteps are used.

Is there a limit to the inrease of the timestep? The impeller tip radial velocity at 600rpm is about 9.9 m/s and the grid is made of about 1.5M cells. Is there an empirical method to determine the maximum rotation per timestep of the impeller?

I mean, I know there's a limit, I can't use a 1 second timestep. :-P

[dj_croog]

I'm doing a similar project involving a stirred tank and have had similar problems... About 1.5 million cells...

I got my continuity to converge by reducing the underrelaxation of momentum from the default value (slowly)