Cell size for solidification/melting multiphase problem
 

Hi everyone,

I am currently running some simulations of a multiphase problem including a solid metal phase and an argon phase.
I've turned on the solidification and melting module and have an energy soruce term that turns on and off to allow solidification and melting of my materials.

Previously, while simulating an aluminium alloy with cell size of 3 μm^3, everything was ok as the solidification range was quite wide.

I am now trying to do the same simulation with a stainless steel alloy which has a really narrow solidification range (between 1658 and 1723 K), and, although it is overall ok, as you can see on the attached graph (I plotted the liquid fraction against the temperature for where the metal volume fraction is equal to 1), I am unsure of the accuracy of the mushy zone (the metal between the solidus and liquidus temperature, corresponding to the two green lines on the attached figure), as the range is really narrow regarding the cell size (the cells are 6 μm^3 cuboids)

I'd love to have your opinion, if the cell size need to be decreased to capture the evolution of the liquid fraction in the mushy zone or if it is still ok

Thanks a lot in advance !

if you are doubting it always better to refine mesh and compare results, ofc if you have time for that

Hi AlexanderZ,

Thanks for your reply, I might give it a try as I agree with you, just to be sure.

However, in welding/3d printing modelling literature, which I am modelling, the size of cells is often 3 microns, and it is already a fine mesh, leading to really long simulations (over a week). Regarding the size of the mushy zone on my figure, I don't really want to go down to 3 microns or even finer mesh hence I was wondering if there was maybe an other way.

Especially regarding the fact that it doesn't seem to be a problem for the temperature field to have a big gradient within one cell (50 K difference within one cell), so I was also wondering why I was having this issue with the fraction of solid which is calculated with the temperature.

The thing is that I export all the data in a tecplot file and and "post process" it on Paraview. When I do the post process from ASCII files in Matlab, there is a difference in the figure between the Cell centre values (more accurate I believe) compared to the Nodes values.