Hi all,
I have encountered a possible artefact. I am trying to simulate austenite formation from ferrite with a simple 1D, isothermal model (input files attached). Initially both the parent ferrite (phase 1) and austenite have a nominal C concentration of 0.039 wt%. When I start a simulation at 1100K, at the first time step the initial austenite slab split into 2 (see the figures attached) and gave a drop in austenite fraction and a strange concentration profile. Obviously the simulation will still reach the equilibrium fraction of austenite, but the morphology is not correct.
Interestingly enough, this does not happen in each of my simulations but only a few of them. For example, the same simulation at 1150K (input file also attached) gives a nice growth of the austenite slab. I have both input files attached here.
I wonder if you have any suggestions to get rid of this artefact.
Thanks in advance,
Billy
Artefact of Phase Content and Concentration Profile in a 1D Simulation
-
- Posts: 25
- Joined: Fri Feb 01, 2019 8:53 pm
- anti_bot: 333
- Location: University of British Columbia
Artefact of Phase Content and Concentration Profile in a 1D Simulation
- Attachments
-
- Eq_1150.in
- (15.91 KiB) Downloaded 285 times
-
- Eq_1100.in
- (15.92 KiB) Downloaded 258 times
-
- Posts: 25
- Joined: Fri Feb 01, 2019 8:53 pm
- anti_bot: 333
- Location: University of British Columbia
Re: Artefact of Phase Content and Concentration Profile in a 1D Simulation
Here in another simulation of mine where I also included Mn (component 2), a similar artifact occurred. This time however, the interface appears to be very thick and I cannot see the austenite slab in the middle.
- Attachments
-
- Eq_EqPhad_conc1_2.png (23.99 KiB) Viewed 2854 times
-
- Eq_EqPhad_conc1_1.png (22.06 KiB) Viewed 2854 times
-
- Eq_EqPhad.in
- (16.86 KiB) Downloaded 260 times
Re: Artefact of Phase Content and Concentration Profile in a 1D Simulation
Hi Billy,
this artefact is caused by a strong composition gradient which is shorter than the interface thickness and thus causes strongly opposite chemical driving forces on both sides of the interface. You can see them in DP_MICRESS if you activate the driving force output (.driv). Without stabilisation measures, this leads to a spreading of the interface, i.e. a broadening which is accompanied by a destroyed interface profile. You can visualize the interface profile using the .frac1 or .frac2 output and the virtual EDX feature of DP_MICRESS (for following up the initial interface degeneration, you need to make the output interval smaller, e.g. all 0.1 s).
Assisted by the fact that phase 2 is shrinking at the beginning (due to the non-equilibrium compositions at start), the two degenerated interfaces from both sides of the initial phase 2 layer overlap. Finally, central instance of phase 1 condenses, leaving the wrong microstructure which you showed.
Having analyzed the problem, there are several solutions:
a) Decrease the interface thickness by increasing grid resolution and/or decreasing the number of cells in the interface
b) Prevent the strong gradient of the driving force across the interface by dG averaging (using e.g. the standard value of "avg. 0.5") in the 1/2 phase interaction data
c) Use interface stabilisation (second optional parameter in same line with interface energy, up to ~10x the value of the interface energy). Just increasing the interface energy would have the same effect and would not hurt in 1D-simulations...
d) Avoid the initially strong composition gradients by starting with equilibrium compositions
Bernd
this artefact is caused by a strong composition gradient which is shorter than the interface thickness and thus causes strongly opposite chemical driving forces on both sides of the interface. You can see them in DP_MICRESS if you activate the driving force output (.driv). Without stabilisation measures, this leads to a spreading of the interface, i.e. a broadening which is accompanied by a destroyed interface profile. You can visualize the interface profile using the .frac1 or .frac2 output and the virtual EDX feature of DP_MICRESS (for following up the initial interface degeneration, you need to make the output interval smaller, e.g. all 0.1 s).
Assisted by the fact that phase 2 is shrinking at the beginning (due to the non-equilibrium compositions at start), the two degenerated interfaces from both sides of the initial phase 2 layer overlap. Finally, central instance of phase 1 condenses, leaving the wrong microstructure which you showed.
Having analyzed the problem, there are several solutions:
a) Decrease the interface thickness by increasing grid resolution and/or decreasing the number of cells in the interface
b) Prevent the strong gradient of the driving force across the interface by dG averaging (using e.g. the standard value of "avg. 0.5") in the 1/2 phase interaction data
c) Use interface stabilisation (second optional parameter in same line with interface energy, up to ~10x the value of the interface energy). Just increasing the interface energy would have the same effect and would not hurt in 1D-simulations...
d) Avoid the initially strong composition gradients by starting with equilibrium compositions
Bernd
-
- Posts: 25
- Joined: Fri Feb 01, 2019 8:53 pm
- anti_bot: 333
- Location: University of British Columbia
Re: Artefact of Phase Content and Concentration Profile in a 1D Simulation
Hi Bernd,
Thanks for your suggestions! After some tweaking I found that changing the composition in phase 2 is the most effective way to avoid this artefact. Apparently I forgot some basic principles and used unreasonable numbers in my previous simulations.
Billy
Thanks for your suggestions! After some tweaking I found that changing the composition in phase 2 is the most effective way to avoid this artefact. Apparently I forgot some basic principles and used unreasonable numbers in my previous simulations.
Billy