Page 1 of 1

Ferrite / Cemenite to Austenite

Posted: Wed Nov 21, 2018 9:59 am
by R.Hess
Dear Bernd,
after our personal talk about my simulation, i was able to make some progress, but i am still far away from my target.
The problem is still, that tne austenite is not growing fast enough into the ferrite. Austenite formation is a diffusion controlled process, therefore i thought, that if the diffusion of carbon (in this case) is fast enough, the transformation should be also. In reality, we can observe austenitisation even at temperature gradients >1.000.000 K/s. Therefore, in my case (with gradients of 10.000 & 100.000 K/s) i should also observe a fully austenitisation.
The problem in my 10.000 K/s case is, that some austenite strings are growing into the ferrite, but the rest seems to be frozen.
At dT = 100.000 K/s even those strings are missing.
In both cases i tried to use the mob_corr function.

I will upload both input files and both result images. Maybe you can check them and tell me what i have to change, because i don´t have any other approach, how i can fix the problems.

Thank you!

Re: Ferrite / Cemenite to Austenite

Posted: Wed Nov 21, 2018 12:21 pm
by Bernd
Dear Rafael,

I found 5 issues which are problematic in your input files:

1.) In the time input data section of the driving file, you limit the numerical phase-field time step width by setting a minimum of 1.E-4 s. This is a huge number in comparison with your time scale and certainly has just been forgotten to adjust. Forcing MICRESS to use bigger time step than those predicted by the automatic time stepping will strongly reduce the interface mobility (please check with .mueS output) in order to maintain stability. You need to reduce this value considerably until this mobility reduction applies only to a small percentage of the interface cells.

2.) In your simulation with a cooling rate of 100000K/s you did not use the mob_corr and paraequilibrium option for the austenite/ferrite phase interaction. Probably, this was only a test...

3.) You update diffusion data in an interval which corresponds to a temperature change of 100K. This is not sufficient and may lead to wrong diffusion data.

4.) You also update thermodynamic data in a much too big time interval.

5.) I don't understand why you use a linearized phase diagram description for the Cementite phase. It is very difficult to do this consistently, given that the data for austenite/ferrite is strongly changing with temperature. It would be much better and easier to use the database.

Bernd

Re: Ferrite / Cemenite to Austenite

Posted: Mon Nov 26, 2018 11:58 am
by R.Hess
Dear Bernd,
thank you for your reply!

I´m going to implement your suggestions in the file and try to get some results.

Re: Ferrite / Cemenite to Austenite

Posted: Mon Dec 03, 2018 12:30 pm
by R.Hess
Hello Bernd,
i have changed some of your mentioned points. The first problem is, in case of using thermocalc for all phase interactions, that the simulation time rises extremly. Furthermore, the result is worse. I cannot observe nucleation and the interfaces become thinner. what is my problem in this case?

Another problem arises without using thermocalc . By changing the phasefield solver time steps and the timesteps of diffusion data the results look much better. i can observe a adequate austenitisation. the problem now is, that the interfaces of the origin structure dont move. i will upload a result as a picture.

- why doesnt transform the cemenite? I think it is a diffusion problem. how can i solve it ? by furhter adjustments of the updating time steps?

- what may be the problem, that the interfaces of the initial strucutre are such stable?

thank you in advance

Re: Ferrite / Cemenite to Austenite

Posted: Mon Dec 03, 2018 4:59 pm
by Bernd
Dear Rafael,

As a first problem you explained that using TC-coupling for all phase interactions makes the simulation very slow. To avoid that, MICRESS has different global options for updating thermodynamic data which allow the user to reduce the computational effort for TC-coupling to a minimum. Furthermore, as the updating interval in most cases can be decreased at the same time, increased numerical stability is obtained. Although global updating may be slightly less accurate compared to a local one, it is still much much better compared to a linearized or mixed description which which cannot be consistent if used with different temperatures. As there is no temperature gradient (I mean a spatial one) in your simulation, I would advise you to use "globalG" for all 3 phase interactions.

The reason why the initial ferrite grain boundaries stay visible and are broadened is not clear. I guess that there are remnants of another phase behind (ferrite or cementite). You can visualize them by looking at the .frac* outputs which show the exact fraction distribution for each phase. Without this knowledge it is impossible for me to know the reason.

The broadened interfaces around the cementite phase show that there are also numerical instabilities in this region. The slow dissolution of cementite could be due the the diffusion kinetics, but it could also be an artifact of your mixed thermodynamic description. You should try to use full thermodynamic coupling...

Bernd