Learning Forum

[flo.schuster]

Hello Everyone,

I am trying to simulate the melting behavior of a homogeneous material inside a furnace (which is also simulated). This simulation is transient and in Ansys Fluent

At one or multiple times during the simulation the residuals of continuity and energy drastically increase (see picture) but the simulation does not diverge. This usually happens, when the melt fraction would achieve 100% at any point for the first time. Up until then, the temperature and melt distribution looks reasonable but during the upset, temperature is redistributed homogeneously in the melt regime (see before and after pictures of melt fraction and temperature).

The mesh is developed to ensure high enough resolutions within the thermal boundary layers and is 2D but rotational symmetric. Some of the important solver settings are depicted below. Also, material properties are partially defined as smooth, named expressions as this was identified as crucial for better convergence. The melt flow is expected to be laminar and the respective model is chosen. Energy and solidification/melting is turned on (Mushy zone 10^8) and in the S2S model, all surfaces are included in the calculation for now.

Somehow, the pressure also reaches drastically high or low values ( +/-10^34) and changes during the mentioned upsets. However, the vacuum is modeled as regular air and patched to 1 Pa, which does not show any effects.

Reducing the timestep from 0.5 s to 0.1 s did also not help. Lowering the energy relaxation factor helped in some cases but led to unreasonable high melting times.

What other possible solutions might i try or where might thes upsets originate from?

Residual upset

Temperature distribution before upset

Temperature distribution after upset

Liquid fraction before upset (red portion is air, melt regime bottom middle that started melting)

Melt fraction after upset (none)

[flo.schuster]

 

Here a reference for the residuals upset

[Rob]

S2S isn't a good choice if you expect thermal radiation to be absorbed by either the fluid or "solid".  Is the melting material fixed density? How quickly is it likely to flow once melted? 

[flo.schuster]

Hi, 

thanks for the reply.

No, the melt-regime should not absorb radiation anyways, as only a combined thermal conductivity could be determined experimentally, also including radiation absorption. Additionally, it seems suited perfectly to model radiative heat transfer in vacuum.

The melting material is also sought to be described by either a piecewise linear description or a user-defined, named expression to fit experimental data. But this seems to be the problem after conducting some more tests. A constant value, simple sloped function or Boussinesq approach would work but do not reflect reality closely enough. The subsequent fluid movement in the melt regime is only driven by buoyancy forces so an accurate description is vital.

Based on geometry and viscosity, a flow speed between 1e-1 and 1e1 m/s is expected.

What approaches can be used to accurately describe the melting material's density? From room temperature up to 1300K the density should be defined as constant at roughly 1400 kg/m^3 and then at least gradually increase to roughly 2600 kg/m^3 at 1700K. So far I tried using a user-defined expression, namely the logistic function for a smoother transition than a piecewise linear description. 

Are there approaches that ensure higher solver robustness?

Thanks in advance for the help :)

[Rob]

It really needs to be fixed up until you reach the mushy point as the model works by fixing the "solid" in place.  What vacuum? 

[flo.schuster]

Ok thank you. Then I will try to implement a constant temperature with a piecewise linear density definition up to the beginning of the melting range. 

The geometry contains a vacuum regime that is modeled as air with a viscosity of 1000 (proved to increase robustness) and patched to 1 Pa pressure. Is that a valid approach?

[Rob]

Viscosity at 1000 cP won't help with stability.  I'm also very limited in what I can suggest as this is a public forum: try it and see what happens.