Learning Forum

[Dubey92]

I am trying to melt a solid block using a moving laser. I am using VOF with Solidification/Melting. Earlier I was going with Implicit VOF but the convergence was slow and then after some time the solution started to diverge. Then I read somewhere that Explicit VOF is better for transient simulations. My domain model is as shown in the picture.

[Dubey92]

The Explicit VOF is converging better compared to Implicit VOF but shows a "floating point exception" after a sometime(after some percentage of completion).

[Rob]

Check the manuals. Explicit gives you more accuracy but time steps may need to be smaller, implicit allows larger time steps and steady state. You'll probably want explicit, so may need to review your time step size.

[Dubey92]

Thanks a lot for your suggestion. My total time itself is small( = 1.6e-3s). I started with the above parameters and solution was converging but abruptly shows "floating point exception" after 21 % of completion.
I cannot understand why it shows "floating point exception" after one-fifth of the solution is done and the solution is converging also.

[Rob]

Something has changed, possibly very quickly, and that's thrown the solver. Run the model up to a bit before it's struggling, save case & data, then look at the flow & temperature field.
The % complete is just a function of how many time steps the solver has done relative to the total number set. It's not an overly useful marker.

[Dubey92]

I increased the Global Courant Number and till now it has gone beyond that point.

[Rob]

Hmm, I'd tend to decrease the time step but let's see what happens. Keep an eye on the solid/liquid mass.

[Dubey92]

The solution is converging. But there are two things happening in between.
Reversed flow at N faces on pressure outlet. Sometimes this comes and then goes away. I read about it and learnt that it should not decrease the accuracy of the solution and sometimes flow reversal can take place in the domain. Is this true?
There was a warning coming for sometime. "Warning: Invalid Cp(0.000000 J/Kg.K) for steel at 20.00 K". But I have initialized the solution with whole domain at 300 K and the temperature can never reach so low. This warning was there for sometime but then the temperature in the warning kept on increasing and then warning was gone. Can you please tell how such a low temperature could be reached?
I have defined Cp using a UDF for temperature above room temperature(293 K). I am more surprised how temperature can reach such a low value as 20 K.

[Rob]

If the solver is struggling you may see a temperature over/undershoot. Given the fluxes and thermal gradients check you have enough mesh and temporal resolution.

[Dubey92]

Thanks for your suggestion. I have a doubt. The simulation shows "floating point exception" in the middle of the simulation. Then I make some change, usually in the Courant Number. Then I start the simulation from the same point and it gets completed. Does it affect the accuracy of the results?

[Amine Ben Hadj Ali]

Yes it does affect the quality of the results. Always good to report plot maximum of velocity, temperature so that to understand what is going on.

[Dubey92]

Thanks for your input.

[Rob]

And have a look at the stability controls in 2021Rx they may help with this case.

[Dubey92]

Is it in the theory guide or user guide? Or are you talking about the under relaxation factors under controls? Sorry to ask if it's a novice question.

[Dubey92]

I am using 2020R2. I don't see any stability controls under controls in setup.

[Rob]

You won't, you want 2021Rx

[Amine Ben Hadj Ali]

Stability controls are there but as TUI commands. Fluent User'S Guide does explain how to activate and what they do!

[Dubey92]

I will see if I can use that.

[Amine Ben Hadj Ali]

You should be able to use it as they are available.

[Dubey92]

Thanks