negative concentration during cementite precipitation

Discussion of numerical issues which are not directly linked to specific applications
Post Reply
qinyu
Posts: 1
Joined: Thu May 21, 2015 5:16 pm
anti_bot: 333

negative concentration during cementite precipitation

Post by qinyu » Thu May 21, 2015 5:52 pm

Hello,

I encountered a problem about negative concentration when I was simulating Bainite transformation of steel (22Mn5). When considering cementite precipitation, the concentration of carbon in the matrix is positive at the beginning, but becomes negative as the cementite grows up. I tried different settings but still can not figure out the problem.

I used thermocalc database to define the phase diagram and stress coupling is also considered. The driving files are attached.

Could you please help me with this problem? Thanks a lot!

Yu Qin
IEHK RWTH Aachen
Attachments
Stress_aniso_dri.txt
using database in phase diagram definition
(31.08 KiB) Downloaded 98 times
Stress_aniso_dri.txt
using linearTQ in phase diagram definition
(30.46 KiB) Downloaded 94 times

Bernd
Posts: 1138
Joined: Mon Jun 23, 2008 9:29 pm

Re: negative concentration during cementite precipitation

Post by Bernd » Thu May 21, 2015 8:21 pm

Dear qinyu,

Welcome to the MICRESS forum.

Having a look on the attached input file reveals that there is a lot of complexity in your simulation setup:
- stress coupling
- use of mixed Thermo-Calc/linTQ
- use of the para-equilibrium model (paraTQ)
- parallel computing
- etc.

Building up a driving file of this complexity requires many weeks of work even for a skilled user. I am wondering whether you started from an already existing setup, or whether this is your first version. In the first case it would be important to know which changes have been made (since it worked correctly) which provoke the problem.
In the second case, I would strongly recommend to start with a simpler setup (small simulation domain, without stress and para-equilibrium) and increase complexity step by step once the simpler setup is working as expected.

In particular, I have the following comments/questions:
- An interface thickness of 3 cells (at the very end of the input file) in your case is too small for a proper functionality of the phase-field method. The smallest reasonable value is 3.5 (without fd_correction) and 2.5 (with fd_correction).
- The use of "fd_correction" (in the same line with "double_obstacle" in the "Flags and settings" section of the input file) is always recommended. If you do so, an interface thickness of 3 cells is ok.
- Using linearised phase diagram descriptions together with Thermo-Calc coupling is very complicated and may - if not done correctly - easily lead to negative compositions or any other problems. I am wondering why you do that instead of using Thermo-Calc coupling for all interfaces.
- The "paratq" option is activating the Thermo-Calc paraequilibrium model, and thus cannot be used together with linearised (lin or linTQ) descriptions. In the present MICRESS version, this is not explicitly forbidden, but will not work. In future MICRESS versions, this input will be not allowed anymore.
- All the redistribution models "nple", "para" and "paratq" cannot be used for stoichiometric phases, because a broad solubility range is required (phase interactions 1/3 and 2/3!). Even if there is a certain solubility of Mn modeled in CEMENTITE, this will cause trouble!
- please post the .in file, which is a cleaned-up input file and which can be read much more easily.
- In complicated cases it is often not possible to directly diagnose problems by only going through the input file. Then, in order to debug the setup, it is necessary to have not only the driving file but to post (or send by PM) all the input files (.ges5 file, other input files if used).
- It is always recommended to use the latest MICRESS version (6.2), especially when advanced models like para-equilibrium are used which did not always work perfectly in older versions.

Bernd

Post Reply