multiphase flow, quick divergence of contuinity eq
 

Hi all,

I am modeling a two phase flow, air and molten glass. temperature ~ 1600K. the viscosity ratio of the two phases are high ~ 10^5 and the density ration is also high ~ 10^3. The domain consists of one horizontal velocity inlet of pure molten glass, one vertical outlet (pointing downward) of molten glass. There is another upward vertical vent where the molten glass is in contact with air.

I am using VOF model so that the interface of air and molten glass can be tracked. The flow domain is mainly molten glass. I've tried all the possible ways suggested in the User's Guide, like using "node based" gradient option, CICSAM scheme for the Volume Fraction Discretization, enabling "Specified operating density", reducing under-relaxation factor for density, using very small time steps, etc. However, the continuity equation diverges after several iterations.

I am desperate now and any suggestions about making it converge would be helpful. Thanks a million.

BTW, I used unstructured mesh. There are a few highly skewed meshes (>0.97) on the volume connecting the horizontal domain and the upward-facing vent. The geometry are originally constructed in Solidworks and that is the best mesh I could get. I wonder if this affects the divergence greatly.

Re: multiphase flow, quick divergence of contuinit
 

hello do you patch the initial condition?

Re: multiphase flow, quick divergence of contuinit
 

yes, I did. I patched a volume fraction of 1 (for air) to the area that is originally occupied by air.

Re: multiphase flow, quick divergence of contuinit
 

r u using presto scheme or not. if not first thing is to switch it to presto. do it pronto.

Re: multiphase flow, quick divergence of contuinit
 

just tried using "Presto!" for pressure, diverged in 7 iterations (Error: > (greater-than): invalid argument [2]: wrong type [not a number] Error Object: nan), though it does seem like it is the "velocity" fields that are diverging.

Re: multiphase flow, quick divergence of contuinit
 

Try putting some boundary layer mesh at the interface between the two fluids. this will make it easier for the interface tracking algorithm. Having highly skewed mesh will in fact affect convergence, but it shouldn't be that significant.

For the specifying of operating density, please specify the lighter of the two fluids, and also please specify a reference point that is explicitly inside the lighter phased fluid.

To be honest, I faced exactly the same problem with my simulation. All the momentum and volume fraction equations were converging, but the continuity equation diverged, and very rapidly I have to say. The funny thing is, (since I have access to both FLUENT and CFX) I tried the VOF simulation in CFX, and it converges very nicely indeed (I have now ditched FLUENT for CFX for all my VOF simulation). So I'm suspecting that the segregated solver inside FLUENT may not be up to the task for VOF, as CFX is inherently a coupled solver (if there are any multiphase flow developers from FLUENT reading this, please verify this please!). Although I haven't tried this myself, try using one of the density based coupled solver and see if that works. Do let me know how it goes.

Hope that helps.

Re: multiphase flow, quick divergence of contuinit
 

Thanks very much for you detailed response.

1. I am not sure how to do this. The interface is not known a prior, at least I do not know. It should be somewhere in the upward pointing pipe, but I do not know how much the molten glass will fill the pipe until I do have a sucessful simulation. I will try to ask someone else to see if they have information regarding this. 2. Yes I used the lighter fluid density (air) for the operating density. I have pressure boundary conditions at the vent, so according to the User's guide, the reference point for pressure should be useless. 3. If this is the case, I'll have to stay desperate since I do not have CFX.

Thanks again!

Re: multiphase flow, quick divergence of contuinit
 

could you try a simple test for me on cfx, it is useful for us.

along with air try a fluid with very high viscosity say : 8E4 and density say 1200 and let me know if CFX can handle it.

Fluent can only handle this with presto scheme, and it is really tough to do. I wish to see how cfx does when viscosity ratio is very high. Thanks.

@violet :Were you able to figure it out later? I'm stuck with a similar case right now.