setting of separated simulation

[liyc]

Dear Micress team,

I have a question in the simulation of solid phase transformation during cooling. In my simulation, austenite transforms to both ferrite and bainite when cooling down. The program doesn't work when I set both ferrite and bainite seeds together in one job. Therefore I try to use restart option to separate it into two parts. In part 2, I use .korn, .phas, and .conc1 files from previous part as the input data. But the program says 'unable to read from xx.conc1 file'. I guess perhaps I set in a wrong way. So I wonder how to set the 'grain input' and 'initial concentration' parts using the data from previous part. Or is there any method that I can simulate both ferrite and bainite transformation in one program?

Best regards,
Liyc

[Bernd]

Dear Liyc,

welcome to the MICRESS forum!

I do not see the reason why you cannot nucleate both phases in one simulation run - prehaps you should explain us a bit more how you plan to treat bainite, which is not a "real" thermodynamic phase, and which problems occur.
What you call a "restart" is essentially a new simulation run, reading in from outputs files of a prior simulation run. I would not do that, unless it is really necessary. The problem with reading the xx.conc1 file could be that you did not convert it to ASCII format, but let us first check out whether you really need to use this procedure!

Bernd

[liyc]

Dear Bernd,

for the explanation of bainite I copy the setting of bainite in my simulation as following:
# Options: recrystall no_recrystall
no_recrystall
# Is phase 4 anisotrop?
# Options: isotropic anisotropic faceted antifaceted
faceted
# Crystal symmetry of the phase?
# Options: none xyz_axis cubic hexagonal
none
# Number of type of facets in phase 4
1
# kin. anisotropy parameter Kappa?
# only one value for all facets/phases
# 0 < kappa <= 1
0.6000000
# Number of possible orientations of a facet 1
6
# 1 -th normal vector facet 1 ? 3*
-1.000000
3.000000
4.000000
# 2 -th normal vector facet 1 ? 3*
3.000000
4.000000
-1.000000
# 3 -th normal vector facet 1 ? 3*
-1.000000
4.000000
3.000000
# 4 -th normal vector facet 1 ? 3*
3.000000
-1.000000
4.000000
# 5 -th normal vector facet 1 ? 3*
4.000000
-1.000000
3.000000
# 6 -th normal vector facet 1 ? 3*
4.000000
3.000000
-1.000000
# Should grains of phase 4 be reduced to categories?
# Options: categorize no_categorize
no_categorize

I think this will show more clearly than just describing it.
In my simulation of cooling process, austenite should transform to ferrite first, then bainite begins to appear. When I set ferrite and banite seeds in one job, the program comes to error after the bainite nuclei come out. The .log file shows like this:

Seed number 165 set at time t = 3.78E+00 s
------------------------------------------
at an interface
Phase: 4
Seed type: 4
Temperature at the bottom = 898.50 K
Undercooling = 217.73 K
Grain number = 432

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Warning in spfAddPhInkr, t= 3.780079
Wrong fractions at nTupelp= 2017
sum = 0.000000
x,y,z = 5 1 199
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!


!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Warning in spfAddPhInkr, t= 3.780151
Wrong fractions at nTupelp= 1982
sum = 0.000000
x,y,z = 6 1 199
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

Then the program stops. Using this setting I can simulate bainite and ferrite formation respectively. Is there something wrong with this setting?
About the .conc1 file, I have translated the .conc1, .phas and .korn files all to ASCII format. But the simulation still does not running. Or I only need to use .conc1 file in ASCII format?
Looking forward to your reply. Thank you very much!

Liyc

[Bernd]

Dear Liyc,

the error messages give me the impression that either the numerical parameters (e.g. interface mobility) are far from reasonable, or there is a conceptual problem: How do you define the thermodynamic properties of bainite? Are you using linearized phase diagram descriptions for the interaction of bainite with ferrite and austenite? Or do you use coupling to thermodynamic databases for all or part of the interactions?
It is very "easy" to make errors when defining thermodynamic data for multicomponent systems. If you say, ferrite and bainite both work fine as long as you do not nucleate them simultaneously, it could be a problem with the interaction between these two phases...

If you don't mind, you could show us the whole input file. If you can't make it public, send it to me as a personal message!

Bernd

[Bernd]

Hi Liyc,

I got your input file as PM, and it looks quite complicated - which are the phases 1 to 4 which you defined? You said that bainite is phase number 4. The initial structure which is read in from a file consists of phases 2 and 3, probably ferrite and pearlite. Further I guess that during heating, pearlite is dissolved and ferrite transforms to austenite (phase 1). Then, on cooling, formation of ferrite and bainite are growing in competition, is that right?

Bainite is interacting only with phase 1 (austenite) and is formulated as a faceted phase. The anisotropy coefficient which you define for the interfacial energy (static anisotropy coefficient) is quite extreme (3.E-3), keeping in mind that 1 means "no anisotropy" and 0 "infinite anisotropy"(Facet Model in MICRESS)! Don't you think this could be the source of the problem?

The problem could be "amplified" somehow through the existence of phase 2, so that you do not see it otherwise...

Bernd

[liyc]

Dear Bernd,

the steel I simulate is DP steel. Phase 1 is austenite. phase 2 and 3 is ferrite and martensite, respectively. In heating, austenite is formed. During cooling, it should be that ferrite comes out first, and then bainite is transformed. I'll try to make some change in the static anisotropy coefficient part, and see if it affects the result. When question still exsists, I'll tell you later. Thank you for giving me this advice.

Regards,
Liyc

[liyc]

Dear Bernd,

these weeks I tried to simulate this cooling process, but the problem was still there. I changed the anisotropy coefficient of bainite to 0.1, and also tried to set the min. undercooling with the value read from .log file. Still after bainite began to nucleate, the error came: (phase 4 is bainite)

Seed number 103 set at time t = 3.75E+00 s
------------------------------------------
at an interface
Phase: 4
Seed type: 4
Temperature at the bottom = 903.00 K
Undercooling = 207.41 K
Grain number = 370

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Warning in spfAddPhInkr, t= 3.756375
Wrong fractions at nTupelp= 5034
sum = 0.000000
x,y,z = 115 1 126
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!


!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Warning in spfAddPhInkr, t= 3.758248
Wrong fractions at nTupelp= 4887
sum = 0.000000
x,y,z = 318 1 186
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

Is there other setting which I can make a change on?
Besides, after changing anisotropy coef. to 0.1, the bainite nuclei become oval shape, which is not expected. About this situation, is there a rule to set the kin. anisotropy parameter Kappa and orientations normal vector of bainite?
Thank you very much!

Regards,
Liyc

[Bernd]

Dear liyc,

It is hard to say what else could be the reason without trying to run it myself. I have two advices which could solve the problem:

- the interface mobility of the 1/4 interface could be too high.
- the nucleation undercooling at the moment of nucleation is already very high (>200K!). Maybe, you should check for nucleation earlier, i.e. increase the maximum nucleation temperature for seed type 4.

If this does not help, please send me the actual input files (driving file, input grain structure and mobility files) per PM, and I will try myself!

Bernd

[liyc]

Dear Bernd,

I think I cannot solve the problem...I tried small 1/4 mobility and higher max. nucleation temperature. Still no effective result. And higher nucleation T led to earlier coming out of the error. So I sent the whole job with all the files per PM. Hope you can find the solution of it. Thank you very much!!!

Regards,
Liyc

[Bernd]

Hi liyc,

I got your input files and could try to run the simulation myself. The problem is that the system is thermodynamically not consistent: phases 2 (ferrite) and 4 (bainite) are both growing from phase 1 (austenite), and they touch each other. You did not define any phase interaction between them, but in order to describe the diffusion between the phases, we need some phase diagram description!
In case of TC coupling, what we usually do in such cases is to use a "constant K" approximation: As the new interface without phase interaction originates from triple junctions, we know the composition of both phases (2 and 4 in your case) in each point and start extrapolating from this point assuming a constant partition coefficient. This works pretty well for many cases.

Essentially, we try to do the same in case of linearized phase diagrams, if there are missing interactions. But there are two reasons why it is much more difficult:
1.) Linearized phase diagrams are globally defined, therefore the constant K approximation cannot start at the correct compositions in all regions of the domain.
2.) In linearized phase diagrams, the user might choose to put values for c0 which are far from realistic compositions. This is essentially why it crashes so hard in your case: you have chosen 0 as reference composition in phase 2, and this value is used for the constant K approach in interface 2/4!


In principle, there are three ways out of the problem, from which No. 3 is by far the best:
1.) Try to shift (by shifting T0) the c0 in the phase diagrams of the 1/2 and 1/4 interface such that the resulting c0 of 2/4 lie inside a reasonable constant K ratio.
2.) If constant K is too wrong, try to go to the other extreme and define one of the phases as stoichiometric (would have to be phase 4 in your case...)
3.) Include phase interactions 2/4 and try to construct a phase diagram description which is consistent for all three phases at all temperatures.

To be honest, I don't think that 1) or 2) will do it, it makes only sense if the system is close to one of the two extremes...

Bernd