error in multiphase modiling

mary-engine · April 12, 2012, 00:05

Hello, everyone, please help me. I'm modeling two-phase liquid-air flow in complicated atomizer. I have to consider the simplest calculation as possible. So my parameters for modeling are follows:

1. Homogeneous multiphase model ( I'm trying to do with and without free surface)
2. Inlet of liquid - 37.5 g/s, inlet of air - 17.5 g/s, opening - 1 atm
3. Turbulence model - SST
4. Timestep is conservative automatic ( I had set even smallest timesteps but problems are almost the same)

Here is my problems:

1. Almost in all cases local mach number begin to growing ( I saw in cfd-post there are a local pressure drop appears, but mesh in this place is fine enough)
2. Almost in all cases supersonic flow appear in place where it should be subsonic, after interface connection.

ghorrocks · April 15, 2012, 18:17

This means your simulation is diverging. I bet you will need to do this transient to get it to converge.

mary-engine · April 16, 2012, 05:52

Thanks, ghorrocks, I've got a small timescale and free surface model and it's not diverging. It's slowly reaching a staedy state field. Can you tell me is there clear physical meaning of timescale and why transient can help? Maybe someone can name some book from which it can become clear. I thought timesteps in transient are almost the same as timescale and in a cases with bad initial guess timescale should be very small to reach steady state and not to diverge. Are there some advices on how to choose and change timescale during the solution except the one in cfx help?

ghorrocks · April 17, 2012, 01:56

In a steady state simulation several terms are omitted as they do not affect the final converged solution. This means the final converged solution is correct, but the evolution through time to get there is not correct. For instance you can advance different equations are different time steps - this is not physically possible so is just an acceleration scheme for steady state simulations.

The reason free surface flows are very difficult to converge in steady state is that the surface usually has small waves on it which reflect and bounce around with little dissipation. They are resolved in a transient simulation but mean a steady state converged result is tricky to obtain.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Difference of multicomponent and multiphase homogenous flows	Luk_Fiz	CFX	11	April 4, 2013 05:29
Use data from multiphase into DO radiation	yashmash	FLUENT	0	October 7, 2011 03:57
multiphase model, water in air?	bugra	Main CFD Forum	10	January 2, 2010 13:09
multiphase multicomponent physics	ckleanth	CFX	3	June 4, 2009 20:15
Multiphase flow modeling	Paul	CFX	2	August 11, 2003 09:41

April 15, 2012, 18:17		#2
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,703 Rep Power: 143	This means your simulation is diverging. I bet you will need to do this transient to get it to converge.

April 16, 2012, 05:52		#3
mary-engine New Member Marina Zayceva Join Date: Apr 2012 Posts: 2 Rep Power: 0	Thanks, ghorrocks, I've got a small timescale and free surface model and it's not diverging. It's slowly reaching a staedy state field. Can you tell me is there clear physical meaning of timescale and why transient can help? Maybe someone can name some book from which it can become clear. I thought timesteps in transient are almost the same as timescale and in a cases with bad initial guess timescale should be very small to reach steady state and not to diverge. Are there some advices on how to choose and change timescale during the solution except the one in cfx help?

April 17, 2012, 01:56		#4
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,703 Rep Power: 143	In a steady state simulation several terms are omitted as they do not affect the final converged solution. This means the final converged solution is correct, but the evolution through time to get there is not correct. For instance you can advance different equations are different time steps - this is not physically possible so is just an acceleration scheme for steady state simulations. The reason free surface flows are very difficult to converge in steady state is that the surface usually has small waves on it which reflect and bounce around with little dissipation. They are resolved in a transient simulation but mean a steady state converged result is tricky to obtain.