Bainite modeling

Exchange about the physics background, diffuse interface theory, etc..
Post Reply
Razzak
Posts: 4
Joined: Tue Oct 29, 2013 12:27 am
anti_bot: 333

Bainite modeling

Post by Razzak » Wed Feb 26, 2014 5:37 am

Hi,
I have been trying to model austenite to bainite transformation using faceted model but facing some issues. Instead of having faceted ferrite (bainte) forming at the GB, I am observing GB instability instead. Please review my driving file (attachment) and let me know a possible solution.

Regards,
Razzak
gamma_bainite.dri
Driving file
(20.07 KiB) Downloaded 283 times
austenite_grain_korn_00012000.zip
starting grain structure
(20.07 KiB) Downloaded 278 times

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

Re: Bainite modeling

Post by Bernd » Thu Feb 27, 2014 1:53 am

Hi Razzak,

Welcome to the MICRESS forum!

I think, the instabilities in your case are most probably due to some extreme or unreasonable input parameters:

1.) In the facet model, you use a value of kappa=1, which is extreme! Please start from 0.5 and increase it stepwise if you feel that the edges are not sharp enough!

2.) In the nucleation model, you use a starting radius of 0.5µm, which is higher than the grid spacing, resulting in a "blocky" seed with wrong composition. Although when using linearized phase diagrams this may "heel out", it would create enormous problems in case of TQ-coupling. Thus, I would say it is at least "bad habbit", and you should set the seed radius to 0 (i.e. starting from a small grain with an initial fraction of 2xphMin in a single cell).

3.) You are using a very small value (1.E-8) for the interfacial energy of the 1/1 phase interaction which is 3 orders of magnitude smaller compared to the values of the other interfaces. This is very probably causing numerical issues even if you are not using extra triple point terms ("multi_obstacle")!

4.) I have the impression that the diffusivities are in general 3 orders of magnitude too small (perhaps a problem of the units, you have to use cm2/s!). The consequence is instabilities of the interface because the diffusion length is smaller than the interface thickness!

5.) For component 1 in phase 2, the activation energy of the diffusion coefficient is 1.E7, which leads to a diffusion coefficient which is huge and not representable in double precision any more! Apparently, MICRESS replaces it by 0...

6.) The interface thickness of 3.0 cells is too small! In no "fd_correction" is used, 3.5 is the smallest reasonable value, with "fd_correction" it is 2.5. So, you should either use fd_correction (what is recommended), or increase the value to 4.0.


I guess, that after resolving these issues, the simulation will behave completely different!

Good luck!

Bernd

Razzak
Posts: 4
Joined: Tue Oct 29, 2013 12:27 am
anti_bot: 333

Re: Bainite modeling

Post by Razzak » Thu Feb 27, 2014 5:59 am

Hi Bernd,
Thanks for your reply.
I have modified the driving file following your suggestions but still the problem persists. Instead of forming faceted ferrite at from the GB into the grain, I can only observe GB instability.
I am uploading the files, please have a look at them.

Thanks in advance,
Razzak
gamma_bainite.dri
Driving file
(15.62 KiB) Downloaded 275 times
austenite_grain_korn_00078000.zip
Initial grain structure
(19.68 KiB) Downloaded 269 times

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

Re: Bainite modeling

Post by Bernd » Fri Feb 28, 2014 7:14 pm

Dear Razzak,

your simulation case is not as simple as it may seem at first glimpse! What makes the situation complicated is the fact that diffusion of component 2 in phase 1 is very slow (essentially, it is practically zero). If phase 2 is growing, it hence must grow at the composition of 1.5wt%, which according to the phase diagram requires an undercooling of about 72 K!

In such situations (i.e. if there is an element which has a diffusion length which is smaller than the interface thickness), phase-field models show "artificial solute trapping", that means there is a partial overrunning of the corresponding concentration pile-up. Furthermore, if the slow element has a strong impact on the driving force, a "spreading of the interface" cannot be avoided: Partial transformation is possible at the outer side of the interface at relatively low driving forces, but complete transformation at the inner side requires 72 K undercooling!

One way to handle that would be to lower the interface mobility so much that the kinetik undercooling of the complete interface is higher than 72 K, so that the interface can stay together. But the much better and more physical solution is to use the "nple" model which MICRESS offers (see here) and which "projects" the high redistribution peak of component 2 onto the whole interface. Then, the whole interface has to overcome the high undercooling, and even nucleation needs at least 72K undercooling, but there is no spreading any more! (The contrary case would be para-equilibrium, if the front does not feel the pile-up of the slow component at all - "para" would be the corresponding option in MICRESS).

The special models "nple" or "para" can be switched on using the option "redistribution_control" in the phase interaction data and by specifying the corresponding model in the phase diagram data input of interface 1/2 for each component (i.e. "normal" for component 1 and "nple" for component two.

Best wishes

Bernd

Post Reply