Pearlite to Austenite/Cementite

solid-solid phase transformations, influence of stresses and strains
Post Reply
Ali
Posts: 9
Joined: Thu Jun 24, 2021 10:56 pm
anti_bot: 333

Pearlite to Austenite/Cementite

Post by Ali » Wed Jul 28, 2021 11:11 am

Dear Bernd,

I am trying to simulate the microstructure evolution from pearlite to cementite/austenite structure.
Unfortunately, the austenite phases could not grow fast and after several changes of the interaction mobility between the phases, the ferrite phases still remain in the structure. Do you have any idea which parameters are better to change to get an austenite/cementite structure?

please see the attached images related to before and after evolution, as well drving file

Thank you in-advance,
Ali
Attachments
Per to Aus-Cem.jpg
Per to Aus-Cem.jpg (320.83 KiB) Viewed 2306 times
H2-8.dri
(29.79 KiB) Downloaded 140 times

Bernd
Posts: 1504
Joined: Mon Jun 23, 2008 9:29 pm

Re: Pearlite to Austenite/Cementite

Post by Bernd » Wed Jul 28, 2021 6:36 pm

Dear Ali,

It is difficult to find anything definitively wrong in your input file without trying to run the problem and without seeing the different types of information from the output. There e.g. could be problems with the dimension of physical parameters like the diffusion coefficients (which can be easily checked in .TabD), with an inconsistent phase diagram, or already within the initial microstructure.

What do you expect from the simulation? Should the initial pearlite (mixture of phases 2 and 3) disappear completely on heating with formation of austenite (phase 1), and then precipitation of cementite (phase 3) should occur on cooling? The fact that there seems to be no ripening of pearlite at least brings me to the conclusion that the interface mobility 2/3 could be much too low...

If you could also show the input file which you used for creating the initial microstructure (or send me the H1-8.rest file to my private mail account if not too big), I could try to run the simulations myself and find out more precisely what goes wrong.

Bernd

Bernd
Posts: 1504
Joined: Mon Jun 23, 2008 9:29 pm

Re: Pearlite to Austenite/Cementite

Post by Bernd » Thu Jul 29, 2021 7:56 pm

Dear Ali,

Thanks for sending me the missing restart-file by email. While running your simulation I found that the problem of insufficient fcc growth is essentially caused by interaction of two issues:

1.) The linear phase diagram description you use is inconsistent. It must be inconsistent, because correct equilibrium at triple junctions (i.e. 3-phase-equilibria) can generally be described by linear phase diagrams only for one temperature, while for the other temperatures information is contradictory: The driving force of one phase with respect to the other two phases is not identical.
The same problem exists also for nucleation: Nucleating phase 1 from phase 3 (3 is reference phase) gives completely different nucleation temperatures than nucleating phase 1 from phase 2 (2 is reference phase). In your case, nucleating from phase 3 has led to earlier nucleation (at lower temperature), but the seeds disappeared because of missing driving force of the 1/2-interface.
Given the fact that the C-composition in phase 3 (cementite) is not significant anyway (you apparently use a numerically high slope value in order to mimic stoichiometric behavior), you should never use phase 3 as reference phase for nucleation. Using a stoichiometric condition for cementite (which would be chosen automatically in case of TC-coupling) is very recommendable - then phase 3 as reference phase would be impossible anyway...

2.) As I already mentioned, the interface mobility 2/3 is very small. This fact not only prevents ripening of the pearlitic structure, it also slows down the fcc growth: Dissolution of cementite into the fast-diffusing bcc phase is hindered, so that fcc can grow only in direct contact with cementite.
Instead of manually selecting a mobility value, I recommend using "mob_corr" with a high physical value for interface 2/3 in the same way as you already do for 1/3. This leads to a much higher numerical mobility value and allows for dissolution of cementite without direct contact to fcc.
Please note that reducing the value of the "minimum time step" may be necessary in order not to cut the interface mobilities (see here for more information). This will lead to a considerably increased simulation time (you could do some performance optimisation by finding out how big the minimum time step value can be chosen without affecting the bcc/fcc transformation kinetics...).

Please find attached a modified input file, which takes these findings into account. It will run much longer but hopefully give you better results. To make it faster I changed the initial temperature to 1000K, as I noticed that the bcc-fcc transformation does not start before having passed this temperature.

Furthermore, you should think about using TC-coupling in order to prevent the inconsistencies of your phase-diagram description. Please also note that I did not address your first simulation step leading to the initial microstructure used for your simulation here, although it appears to be badly resolved and dominated by numerical grid anisotropy ;)

Bernd
Attachments
test2-8.dri
(29.79 KiB) Downloaded 154 times

Ali
Posts: 9
Joined: Thu Jun 24, 2021 10:56 pm
anti_bot: 333

Re: Pearlite to Austenite/Cementite

Post by Ali » Fri Jul 30, 2021 10:46 am

Dear Bernd,

thank you very much for your nice feedback.
I will try to correct them based on your comments. I will update here again.

Regarding the first simulation, I only considered isotropic growth, but I do not know how they grow anisotropically.
Do you know which parameters might be leading to this problem

Best regards,
Ali

Bernd
Posts: 1504
Joined: Mon Jun 23, 2008 9:29 pm

Re: Pearlite to Austenite/Cementite

Post by Bernd » Fri Jul 30, 2021 11:32 am

Dear Ali,

I think the biggest issue is the too coarse resolution which increases grid anisotropy (i.e. preferential growth of the lamellea in the grid directions). I think there should be some room for increasing the grid resolution (decreasing Δx) because currently most time is spent in list-related operations ("list time" and "PF time" in the .TabP output) which scale only linearly with grid resolution...

Bernd

Ali
Posts: 9
Joined: Thu Jun 24, 2021 10:56 pm
anti_bot: 333

Re: Pearlite to Austenite/Cementite

Post by Ali » Tue Aug 24, 2021 10:47 am

Dear Bernd

I have changed the grid size for the first simulation based on your suggestion from 0.1 to 0.02. I observed two different results. But I don't know which one is correct?
And the simulation with the grid size 0,02 takes very long time. after about a week it is still running. how can I decrease the run time?

for the first heat-treatment, I want to simulate the microstructure evlution from austenite to pearlite and cementite.

thank you very much in advance.

Best regards,
Ali
Attachments
Austenie to Pearlite and Cementite.jpg
Austenie to Pearlite and Cementite.jpg (352.59 KiB) Viewed 2185 times

Bernd
Posts: 1504
Joined: Mon Jun 23, 2008 9:29 pm

Re: Pearlite to Austenite/Cementite

Post by Bernd » Tue Aug 24, 2021 11:49 am

Dear Ali,

Generally, if you do the same simulation with different grid resolution, the finer resolved one should always be more trustworthy.

However, if possible you should always compare to experimental results, and it may happen that low resolution hides wrong model assumptions which become eminent only with higher resolution. I am wondering whether and why you should have regions with cementite particles and others with eutectic. In the simulation (and also in reality) this may be strongly dependent of the interface energy settings.

If you want to check whether performance can be improved, you should first have a look at the .TabP output to see which parts require the highest part of the simulation time. This approach has been described in more details here. If you are lucky, there is a clear bottleneck which can be addressed to increase performance.

Bernd

Ali
Posts: 9
Joined: Thu Jun 24, 2021 10:56 pm
anti_bot: 333

Re: Pearlite to Austenite/Cementite

Post by Ali » Tue Aug 24, 2021 3:23 pm

Dear Bernd,

thank you for reply.

My material is Fe-1%C and I want to simulate the red marked microstructure. Is this wrong?
I sent you already the dri file by e-mail. Should I decrease the Cementite nucleation distance?

Best regards,
Ali
Attachments
Iron-Carbon.jpg
Iron-Carbon.jpg (124.35 KiB) Viewed 2180 times

Bernd
Posts: 1504
Joined: Mon Jun 23, 2008 9:29 pm

Re: Pearlite to Austenite/Cementite

Post by Bernd » Wed Aug 25, 2021 12:07 am

Hi Ali,

This is what I meant: You do not expect to get rows of round cementite particles. This could be due to a too high nucleation undercooling, i.e. supersaturation of C. The rows of particles further suggest that you do not check for nucleation often enough, or that you have a too long shield time and/or distance. It is necessary to play a bit around with the nucleation parameters to match best to the real microstructure. Perhaps you should select a smaller part of the domain for that in order to get faster the results.

I did not get any driving file from you by mail yet. You can simply attach it to your answer if there is nothing confidential about it.

Bernd

Post Reply