A floating point exception has occurred: floating point exception [Overflow].
 

Hi! I'm working with a multiphase flow of liquid and vapour heptane. I've activated the Segregated Fluid Temperature Model and I've selected the User-Defined EOS. I've chosen a range of temperature between 200K and 500K and I have values of density for both the liquid and the vapour, depending on temperature, so I've selected Table(T) as density input in the model!

After few iterations I get this error: "A floating point exception has occurred: floating point exception [Overflow]." and the calculation has diverged.

I've tried both a polyhedral and a trimmed mesh, not very refined, because right now I don't have much computing resources.

Someone could help? Any advices? Thanks!

Are you seeing pressure or velocity spikes anywhere in the domain? Can you determine what is diverging? I would suspect that if the model is setup right that your EOS is causing the issue, but its hard to say without any additional info. Sometimes this can happen due to boundary conditions or even your mesh, especially if it is too coarse in areas with high flow gradients.

Thank you for your reply! Yes, velocity spikes and so other quantities I'm monitoring! I actually think that my EOS model isn't set correctly! I didn't modify anything else in my model besides density definition.. maybe should I initialize pressure or other quantities in a different way?
I'm working with constant pressure @inlet and constant pressure @outlets, turbulent flow and so far it was isothermal! I'd like to introduce quantities temperature dependence and I'm literally walking in the dark! :(

I've switched to 1st convection order on Velocity and Pressure and I finally could refine a little bit the mesh.. It seems to work now! But I have to return to 2nd convection order eventually!

I don't think it had to do with your convection scheme. It was probably a mesh issue that you worked out when you refined it. A few suggestions moving forward.

1. Layer in complexities. Start with a simple model then add in your physics once you are comfortable with what you are doing. For example, try running with constant density or a predefined EOS before using a custom setup.

2. If you can run a coupled analysis there is an option for expert initialization that works well for me. It essentially solves simplified NS equations on a series of finer grids to give you a somewhat realistic starting condition.

3. Ramp up your courant number over the first 50 iterations or so to reduce the risk of divergence.