[Ringelsy]

Dear all,

I am simulating water particles of 3 µm diameter entering a test chamber together with an air stream with a pre-defined, initial relative humidity of 40 % and temperature of 25 °C.
Particles are evaporating in the chamber, decreasing temperature locally (due to latent heat of evaporation) and increasing relative humidity.
I am using the Liquid Evaporation Model with two phases, a gaseous phase (moist air, containing air and gaseous water) and a liquid phase (water particles);
I have tried different test setups (different geometries/meshs, different particle mass flow rates, air velocities, amount of particles, particle iteration frequencies...)

Under certain conditions, the model is working. However, in many cases, there are local errors and temperature takes unrealistic values (e.g. in some cases more than 3000 K or less than 50 K). These errors always occur in the region where particles are located.

In a simple, stationary test case, I have found that the model seems to work if I chose a physical time step of 5*10^-4 s or lower. This makes sense to me, because particles evaporate quite fast, so a very low time step must be chosen.

However, chosing such a small time step, will need a lot of iterations, since I need to simulate at minimum the time the air stream needs to get into a steady state. In my test case I used for example a velocity at the inlet of 0.05 m/s; the total length of the box is about 50 cm, meaning I need to simulate a total time of at least 10 s. At a physical time step of 5*10^-4 s, this will need at least 20000 iterations in this case. In this test case, this might still be possible, but I want to apply the model also on a larger scale, where the simulation time will be too long.

I was thinking to calculate with variable time steps and add particles only to certain iteration frequencies and chose only at the moment of particle addition such a small time step. However, I do not know if this will solve my problem.


Has anybody an idea how to calculate two situations with different time scales at the same time? Is there maybe the possbility to use different time scales for the evaporation model than for the Navier-Stokes equations?

Or has anybody experience with the liquid evaporation model and found a different approach to fix this?

Thank you very much for your help.

[ghorrocks]

Widely varying time scales is a common problem in CFD. So when you have two or more different physics happening you can expect that one will be fast and one will be slow, and the end result is if you want to resolve both you need long simulations.

If you are doing steady state simulations the easiest way to speed things up is to advance the equations at different speeds. In your case you should be able to advance the fluid and heat equations at their rates and the multiphase equations at their rates.

Another approach involve separation of time scales. This is where you model the fast time scale by itself, with the long time scale physics assumed constant; and repeat the model at various points on the long time scale physics curve. This gives you a model of the fast time scale physics outputs at various points on the long time scale evolution, and then you can model the long time scale with the outputs only, rather than modelling it in detail.

This is a simple example of separation of time scales which is a whole research area in itself. There are much more sophisticated ways of doing it than this if you care to dig into this topic.

[Ringelsy]

Thank you very much for your reply. I think this will help me.



I will try to set different time scales for different models first and see if it is working.



If this is not working, I will try the second approach.