Issues with FA-mode Solidification Simulation and slow-down

[Ku shihyeon]

Hi, I am currently conducting simulations on the development of solidification microstructures for stainless steel alloys,
specifically focusing on F, FA, and A solidification modes.

I am presently working on the FA (Ferritic-Austenitic) mode, but I have encountered several problems
In the attached stainless.dri file, I have included the Austenite (FCC), Ferrite (BCC), and Sigma phases.
I have also configured the nucleation seeds to account for all stages, from F-A solidification and nucleation to sigma precipitation during the post-solidification thermal history.

Specifically, the FA mode solidification is modeled through the following stages;
Stage 1: Growth of primary ferritic dendrites.
Stage 2: Nucleation and growth of austenite at the interface between the liquid and the ferritic dendrite edge
Stage 3: Consumption of the ferritic dendrites by austenite, proceeding from the Austenite / Ferrite interface toward the dendritic core (F to A solid-state transformation), resulting in a vermicular (skeletal) ferrite morphology within an austenite matrix.
Stage 4: Simulation of sigma precipitation by applying a high-temperature thermal history to the final solidified microstructure.

For now, the F-A solidification and solute redistribution are highly unstable(image attached), and simulation is extremely slow.

stainless.png

(image from stainless.dri)

To address this, I have simplified the code to first implement a basic primary ferrite single-phase solidification (stainless4.dri).
I anticipate that stabilizing the complex F-A solidification microstructure can be achieved by optimizing parameters such as interfacial energy, anisotropy, and anisotropic stiffness for the Liquid/F, Liquid/A, and F/A interfaces.
I would greatly appreciate any advice or precautions regarding the implementation of these solidification processes and microstructures.

Furthermore, I am currently struggling even with the simplified single-phase ferrite solidification.
The primary issue is the abnormal simulation time.
Despite adjusting the re-linearization interval, diffusion recalculation interval, and PF solver time step,
the calculation remains extremely slow.

After the initial step, the simulation effectively stalls.
The log outputs the following:

Intermediate output for t = 1.00000E-36 s
CPU-time: 2 s
Current phase-field solver time step = 6.48E-08 s
Average conc. of comp. CR = 20.0383059, Variation = +0.0000000 wt%
Average conc. of comp. NI = 14.8802462, Variation = -0.0000000 wt%
Average conc. of comp. MO = 2.0113571, Variation = -0.0000000 wt%
Temperature at the bottom = 1710.00000 K
Temperature gradient = 0.00000 K/cm

Approximately 10 to 20 minutes after this output, the line "Updating of diffusion data from database..." appears,
but no further progress is made for several dozen minutes.
Could you identify what might be causing this problem?
I have attached the latest simplified file (stainless4.dri) along with the TCM and GESS files used for the calculation.

Thank you for your assistance and looking forward to your advice.

Best regards, Ku

[Bernd]

Dear Ku,

The reason for the extremely slow execution is simply that you use a much too fine meshing for a solidification problem. The grid resolution should be around 100 times coarser, which also corresponds to the expected length scale of solidification.

A second problem is that you probably want to use ferrite (BCC_A2) as primary phase and not BCC_B2.

Finally, if you do not have a significant temperature gradient in your simulation, you should not use "diagonal_dilute z" for diffusion data, because it does not make sense. You should rather use "diagonal_dilute g" instead.

Bernd

[Ku shihyeon]

Thank you for your reply.

I mistakenly thought that calculations at the level of 2000² cells had been possible in a previous project. I had not even considered resolution as a factor to take into account.

As for the BCC_A2 phase, it seems that it is not included in the TCNI12 Ni-based database I am using. At present, we are not able to adopt a new Fe-based database.

Based on equilibrium and non-equilibrium solidification calculations performed with the Ni-based database, the composition Cr-20 Ni-20 Mo-2 Fe-bal. resulted in the phase corresponding to ferrite being identified as BCC_B2#2. For now, I believe following this result is the best approach.

I also confirmed that the meaningless z option was being used. It appears that I had imported files from a previous project without modification.

Best regards, Ku.

[Bernd]

Dear Ku,

You are right. Using the TCNI12 database (instead of a dedicated steel database), BCC_A2 is replaced by BCC_B2#2 (disordered).

Bernd