For the initialization is ok. But if you want to capture time dependent phenomena, it is better to apply a mesh motion.

Franco

It really depends on what you are trying to measure and what the geometry is. As mentioned above, you will inherently be unable to capture any transient effects. It's known that in 1-stirrer vessels there are macro-instabilities that may affect mixing, and these may have a stronger effect in cases with multiple impellers where strong compartmentalization occurs (we're working on a publication on this). You will not capture these effects with MRF. Still, especially in 1-vessel tanks, the mixing time and behavior can be reasonably captured.

If you want to include reaction kinetics and check some long-term developments, using MRF may be the only solution realistically. Take into account that some features may not be captured, and you can at least get a crude picture of development of substrate gradients.

Some questions you may wish to ask:

- Rotor/baffle interaction (if strong, use slidng mesh)

- significant macro-instabilities in single phase? And are accurate mixing profiles required or more order-of-magnitude estimations? If the latter, MRF is fine.

- With aeration, are temporal instabilities significant? In my experience they are not very strong in stirred tanks, but again the relevance depends on your goal, e.g. accurate mixing profiles (sliding mesh) or order-of-magnitude substrate gradients (MRF).

Thank you both for you answers.

My objective is to do a gas-liquid simulation of dual impeller STR and get a crude picture (MRF) of mass transfer.

Whereas i understand MRF to be a steady state approach (and this I've done in single phase), the switch to transient MRF in gas-liquid simulations in literature is what is puzzling me. What causes the transients?

In addition, as I try to run initial transient single phase simulations within two phase setup (Volume fraction equations switched off), I get oscillating solution contrary to my expectation of a solution approaching steady state. Could this have any relation with the macro-instabilities you've mentioned?

I don't think MIs will sustain if MRF is used, even if transient (but when using MRF for mixing, I always disable all fields except the scalar itself in transient mode). Wit sliding mesh, h we saw oscillations in the range 0.02-0.06 f/N = with N the stirring speed in 1/s, typically.

With gas-liquid flows, dynamic pressure differences can cause oscillations. There are some classic papers by Beckeret al. for a pseudo-2D bubble column with a central single bubble plume; the plume will inherently become unstable and sway from left to right. This case has become a classic for the validation of transient CFD codes due to the well defined fluctuation frequency. However, it may be a more tricky in stirred tanks where bubble plumes interact with forced impeller flow - which is why stirred aerated flows typically reasonably converge in a steady state approximation (much better than bubble columns) .

Is it possible that MIs could be sustained even with MRF?

My train of thought is that a RANS + k-e turbulence model "filters out" the turbulent fluctuations and the MRF technique "filters out" the periodic fluctuations due to blade passage. However, Roussinova et al (2003) [Low frequency MIs in a stirred tank...] suggest that the mechanism for MIs is pressure reflections from jet impingement, and the solution path for MRF (from zero velocity field) includes simulating the jet impingement .

It might be, never tried. I'll start one of my previous simulations with transient rather than frozen MRF to see what happens.

Good work and I wish to see the final result for this matter

The simulation is still running, but so far it seems you are correct. 30s of flow time (~ 1 mixing time at the given settings) have been completed so far, with about 6 periods in the velocity signal. At 5 RPS, that gives a frequency of f around 0.04N; roughly in line with earlier sliding mesh stuff I conducted. Of course there are some initial transients due to the change from steady to transient MRF, I'll leave it running for a second mixing time to see if it becomes properly periodic and then do FFT on the signals to properly get the frequencies.

Hi Godfrey, Sorry for not getting back to you earlier - was a bit swamped with other stuff. The simulation finished and I checked the frequencies with FFT at 3 points. With transient MRF, consistent oscillations were visible in a 2 impeller tank at a frequency f/N = 0.045-0.05 (N being the stirrer speed in 1/s). This is in agreement with sliding mesh results that gave f/N = 0.04 - 0.045, which we reported in a manuscript that is currently under review. It's inbetween aforementioned experimental frequencies.

This is well received. thank you.

Such a nice discussion. Actually, I recently submitted a manuscript in a journal and it was rejected concerning on using MRF in a transient study. The way of my investigation is the achievement of steady state flow by MRF in a gas-liquid stirred tank and then solve equation of oxygen concentration by fixing the flow field. My validation process is based on the time evolution of oxygen concentration in the reactor which has been compared with experimental measurements. For sure due to this strategy, there is discrepancy between simulation and experimental in initial minute, but we observed there is a good match after the initial mixing referring dissolved oxygen concentration. Any experience and idea would be appreciated regarding this issue. Tnx