2D Glass Melt Simulation Setup
 

Hey all, I am currently trying to do a 2D study of a glass rod softening inside of a cylindrical heating element. The goal is to accurately simulate how the glass rod will soften and form a molten glass droplet.

I am running into a multitude of problems due to my lack of experience with FLUENT, and CFD software in general. I will try to be as comprehensive as possible with my current setup, and I will really appreciate any and all input on it.

Refer to this imgur album for relevant pictures: http://imgur.com/a/M6NKI

Here is a quick visual of the setup. A 15mm diameter glass rod is inserted 76.2mm (3in) into a 152.4mm (6in) long cylindrical heating element from the top. The cylindrical heating element generates 1240W of heat. Air makes up the environment within the element. Assume that it is an enclosed environment and no air can escape through the top or bottom.

Below is the geometry I have modeled. I assume it is okay to model half of the setup, and define the centerline as symmetry to simplify computation. Two surface bodies were created: the large surface representing the air environment and heating element at the wall, and the small rectangle representing half the glass rod.

Note the named sections I have created: heating element at the right wall, symmetry at the centerline, glass-air boundary (right side and bottom of glass rod), etc. Do these seem appropriate?

I am no expert in meshing, so I simply applied an automatic method to the entire geometry with min sizing of 0.0001, max face size of 0.0005, and max size of 0.0005. Is this the appropriate way to mesh for what I am trying to accomplish? Do I need to apply separate meshes to each surface body? Do I need to refine the mesh at the boundary between the two zones?

Due to my lack of knowledge, I am unsure of a few things with the initial setup. If I am assuming a constant density of the glass, do I need to make the solver pressure or density based? Absolute or relative velocity formation? Planar or axisymmetric? I'm assuming that a transient setup is appropriate, and I have enabled gravity (9.81m/s2) in the y-direction.

I have enabled the energy equation and S2S radiation. I know I need to consider radiation since it is the primary method of heat transfer from the heating element to the glass, but to be honest I know very little about the different radiation models within FLUENT. S2S seemed like the most simplified and applicable model to use, but I do not fully understand how to set up S2S, let alone any of the other models.

Material properties of the soda lime glass rod are important here, obviously. I am using a constant density of 0.000252 kg/m3. This is mainly due to my inability to find a polynomial relation for density vs. temperature for this material. If I assume constant density, does that make this problem incompressible from a setup standpoint? I have a polynomial relation for specific heat. Constant value for thermal conductivity as well. And finally, a piecewise polynomial for viscosity (two ranges, 0-823K with an arbitrary very high viscosity value of 10^12, and 823-1723K with the polynomial relation).

Cell zone conditions are straightforward, with air applied to the large surface and glass applied to the small surface.

For boundary conditions, I have simple wall conditions applied at the top and bottom of the environment. Symmetry at the centerline. At the right wall I have applied the heat flux of 1240 watts. Under the thermal tab, since I am considering radiation do I need to choose the mixed option? Or simply the heat flux option? In regards to the S2S setup, under the radiation tab I have defined the right wall and the glass-air interface surface as critical zones. Again, I'm figuring this out as I go along.

Should I be using hybrid or standard initialization?

I think I have covered all the bases. When I go to run the simulation, I get the following error:

Error: Divergence detected in AMG solver: x-momentum

I have gotten this error before in other simulations and have solved it by either making the mesh finer or changing the under-relaxation factor of momentum. Neither of those have worked, and I am assuming that it is because I have made a multitude of errors in setting up this simulation.

Sorry for the lengthy post, I wanted to be as thorough as possible to fully define my problem and give you as much information as possible. Again, any and all input is greatly appreciated, thanks!

Another thing worth noting is that I have chosen not to enable the melting/solidification model. This is because glass is an interesting material in that it does not actually melt, but rather softens. After doing research, I've concluded that the melting model should not be used because the glass is not going through a phase change, and therefore there is no energy exchange. The glass is simply getting less viscous as it heats up, which I hope will be modeled correctly with the piece wise polynomial. Does this logic make sense?

Hi marmz,
your problem is very curious and interesting.
Unfortunately I can't answer all your questions, but let me fix some points:

1) Mesh: I suggest using conformal mesh, in order to activate it you have to put glass and air in the same part (In DesignModeler if you are using it in workbench), creating a Multibody part. Otherwise, with a non-conformal mesh, to have coupled-wall (wall across which there is heat flux) you have to define an interface between this two solids.
About the cell-dimension you have to try different cases in order to find the best.

2) Set-up: You have to activate Axysimmetric
For the Solver I would start with Pressure-based (Segregated, maybe Coupled can be enabled afterwards).

3) Radiation Model: I suggest reading Fluent Manual, where it's well explained which model is more suitable for your problem (it depends on the type of surfaces and medium involved). Probably DO is the best in your case.

4) Droplet: if you want to simulate droplet fall you have to activate Discrete Phase Model (DPM), I'm not expert in this kind of simulation.

Sorry for my bad english

Regards

Thank you for the information! I will give this all a try. Can anyone else contribute anything?

After speaking with an expert on optics, I realize that at the temperatures the furnace will be producing, the wavelength of the radiative heat transfer will be in the infrared wavelength spectrum. This means that very little of this radiation will penetrate through the surface of the glass rod (penetration into the glass would require visible light wavelengths or higher). So in short, the radiation will only heat the surface of the glass, and then there will be heat transfer via conduction through the rest of the body. This makes me think that the Surface to Surface model was indeed the right choice to use since there will be no penetrative radiation heat transfer.

Had any success in modeling the melting of the glass rod?
 

Hello,
I am a newbie to this forum but have some experience with commercial CFD from years back. I'd like to model some simple glass fusing geometries and was wondering if you had any luck with solving your problem. I have access to ANSYS CFD/FLUENT.
Thanks,
Mario