Timestepping in two - phase Simulations using RSM
 

Hi all

I have a question about the correct set up for the timesteps in a two - phase air - water flow simulation in subchannels of a nuclear reactor.

The following models were used for the simulations:
Two fluid model (air and water, buoyant flow with standard components), monodispersed bubble distribution (here 3mm)

Since the set up contains no heating, the isothermal model was used.

Alle the available Drag and non - Drag forces were activated (in agreement with simulations carried out before, the Grace model for the Drag force, the Tomyiama Lift force model, Frank wall lubrication force model and Favre-averaged Drag force model for turbulent dispersion were activated - furthermore, Satos enhanced eddy viscosity model for the turbulence transfer was activated).
Furthermore, the BSL Reynolds Stress model was used (see below for justification; in fact, the recommendation of changing this model is not the key for success).

For the outlet, an opening condition was used, setting a static pressure condition of 1 atm (and consistend values for the other parameters).


The figure attached shows one plane of the smulated geometry and the results obtained for single - phase flow (which does not matter, because the mean tendencies in the velocity fields, which are shown in single phase, can also be seen in two - phase flow).

The vector plot shows the structure of the arising secondary motions. These motions, which can also be seen in two phase simulations, are only calculated correctly if a RSM model is used (and you can believe me, it has to be like that: approaches using two equation tubulent models fail).

The three png graphic files show typical convergence behavior for the carried simulations carried out. It can be seen, that the values for the volume fraction are quite bad, where the other parameters seem to behave quite well.

My basic question is about timestepping: it has shown during the simulation work, that the set - up of the time step has a sever influence of the quality to the results - but i was not able to figure out, which rules should be taken as reference (and the documentation did not help me out...). My general question: is there a rule, which timescales has to be resolved by setting up the timestep? Or is there a general remark which you could make?

Best regards
Torsten Betschart