I often encounter convergence problems when using DPM with interaction with the continuous phase. Part of this is probably due to the finite number of particles/trials. Lowering the relaxation factor for the DPM sources or the number of continuous phase iterations per DPM iteration, as well as increasing the number of particles/trials helps. But I wonder if anybody has an advice which of these is more efficient, i.e. will lead quicker to a good result?

Hi,

I don't have the solution, but I know one thing, you mustnot lowering your relaxation factor for the DPM source (or very close to 1) or you will not consider all the action of the dpm on the continuous phase. It's quiet difficult to explain but in contrary to other relaxation factor, you will consider only a ratio of the action, you will do iteration on continous phase with this terme source, calculation will converge or not, and you won't have seen all the action of the DPM. In contrary, to others relaxation factors , you can't see if the solution about the DPM is converging, that's why you can make this mstake without knowing, but effectively you are far from the rreality. Else the number of continuous phase iteration by DPM iteration is a good point but if you are in unsteady calculation, the best is to reduce your time step. Sorry if my english isn't very good (Im french), good luck. Good luck

Boris, Is this also the case for steady state calculations? I can see your point for the transient case. Do you mean that, for instance, momentum transfer is only partial from the discrete to the continuous phase? This would be quite worrying if this is also the case for the steady state case.

Yes, that is exactly what I mean, (this information was given to me by a support engineer of Fluent), I have some doubts about a steady state case, but I think it's the same. To be sure, a simple case would be to inject just one droplet in steady state(evaporation with consideration of energy) and to compare the lost of mass of this one with the total gain for the continuous phase(with a small relaxation factor for the dpm source), it should be equal but if only one ratio is considered, a difference in proportion to relaxation factor should appear.

Boris I find your feedback interesting. I have run several chem/spray cases using 1 step rxs and according to a fluent engineer, the URF for DPM should not go above 0.5 for these types of cases. When I start the interaction between the 2 phases, it's at 0.01. After 100 itrs I change it to 0.1 for another several 100's and if it lags to max 0.5. Using these factors, the case remains stable and I get convergence (I check this by fluxes and total heat xfr=sum dpm enthalpy source).

I agree with Noell. I only do steady state calcultions. Most of the time the only way to get a DPM odell running is to start with a very small DPM urf. I usually start with 0.05 and increase it gradually up to 0.5. If you monitor flow variables that are effected by the DPM you can judge when its time to increase the urf. You can also use figure 23.7.1 from fluents manual to get a feeling how much the urf will affect your solution convergence.

RoM

Ok, I thank you for those informations, I had some doubts about the way to consider those URF in a steady state case, I used it only in unsteady case and in those cases calculation methods are different

I checked it with a fluent engineer, he says there should be no issue with underrelaxation. My mass balances are also correct, I don't see 'missing mass', so I guess that using underrelaxation is allright, as one would expect.