[SardarMech]

Hi everyone.

I am writing this post hoping that someone can guide me out of the trouble I have.

I started Fluent modelling of a stirred tank. The model specs are:

SPECIFICATIONS of the MODEL

Meshed via Fluent With Meshing, element size minimum:1 centimeter maximum 6 centimeters or so.

Tank: height 6 m - dia: 3.8 m (there is one single baffle there too)

medium: a slurry water with total density of 1.6 [kg/m3].

I am using MRF, since sliding mesh was a bit demanding to create in Fluent With Meshing, I working for a tight deadline.

There are two impellers whose velocity is 80 RPM (i.e 8.9 rad/s)

Steady/transient(?) ||||| K-e, RNG, swirling flow ||||| BC: top of the tank is pressure outlet turbulence intensity (0.001) and hydraulic diameter (3.8 m = tank diameter(?)) ||||| tank walls are set to zero shear stress to model flow passing freely over the tank wall. ||||| PISO with gradual timesteps of 0.1, 0.25 and 0.5 so far for transient and Coupled pseudo for steady ||||| pressure standard, momentum k and e set to 2nd order |||||

SIDE NOTES
Since doing a transient simulation would call for a long (at least 5 minutes) of stirring the tank, and since I hope to find flow patterns after the tank is fully stirred, a steady simulation seems to be suiting my case, however, only physically and apparently not numerically.

When doing steady, I monitor some parameters stability and I am not content with scaled residuals as suggested in other posts. But since I am having too high residuals for continuity (i.e no less than 1e-2) and other residuals, I doubt the validity of the monitored parameters, thinking to myself that this can be by nature a transient model and not a steady one. That is why I have kept with transient so far (i.e a few seconds into the simulation playing with URFs and models).

MY GOAL:
Like stated earlier, my goal is to stir the tank for 5 minutes and in case transient modeling is not numerically achievable or is expensive, then to run a credible steady state simulation where I can judge from both scaled residuals and monitored parameters (moment on shaft, velocities, force on the baffle etc) the accuracy of the results.

So my questions are:

1. Am I using too fine a mesh (average of 3-4 cm) for a tank of 68 m3; so much so that the fine mesh is causing the divergence?

2. Is it numerically logical to switch to steady solver after the fluid seems to be stable in the transient simulation (i.e transient ==> steady, and by stable in transient I mean flow patterns seem to be fully occupying the tank)? I have found the opposite to be more commonly stated in cfd posts, but not transient to steady.

And What can be possibly wrong with my model configurations stated above, like RNG etc etc.