Dear Bernd
Thank you for your kind explanation. Will the grid spacing is the smaller the better (if I do not mind the long time simulation)? When I increased the grid spacing to 0.24μm, I found that there are lots of lamellae that I did not define grew up at the beginning. After I decreased the grid spacing, these lamellae disappeared and only the lamellae I defined grew up. Do you know why this happean?
thank you very much!
lihaoge
Cr-Ni eutectic alloy simulation
Re: Cr-Ni eutectic alloy simulation
- Attachments
-
- 0.47.png (16.5 KiB) Viewed 3500 times
-
- 0.3.png (23.33 KiB) Viewed 3500 times
Re: Cr-Ni eutectic alloy simulation
Hi lihaoge,
You are correct with your conclusion that higher resolution usually mean more correct results. However, this is not the case for nucleation, especially when empirical nucleation models are used. E.g. the "stabilisation" model (in the section of nucleation input) may behave completely different when changing resolution. It is a "single grid cell" model which has an inherent "critical radius" which is equivalent to the grid size. The higher resolution is (and the lower the grid spacing [gr]Delta[/gr]x), the smaller is this "critical radius", and the higher the inherent "critical undercooling" which has to be overcome.
In some occasions, this "critical undercooling" (or the equivalent curvature undercooling of a sphere with radius [gr]Delta[/gr]x) is written to the screen output (see here).
If this value is higher than the user-specified critical undercooling for nucleation, nucleation may be impossible, or nuclei stop growing before reaching the size of a full grid cell. This could be the case in your simulations where a lot of such remnants can be seen as blue dots. The solution then is to use "analytical_curvature" instead of "stabilisation" in the input for the given seed type and define (directly in the next line) a critical radius which is bigger than Δx). This model replaces the normal phase-field curvature which is maximal when the grain reaches full size, replaces it with an analytical value and thus implies a lower critical undercooling.
Bernd
You are correct with your conclusion that higher resolution usually mean more correct results. However, this is not the case for nucleation, especially when empirical nucleation models are used. E.g. the "stabilisation" model (in the section of nucleation input) may behave completely different when changing resolution. It is a "single grid cell" model which has an inherent "critical radius" which is equivalent to the grid size. The higher resolution is (and the lower the grid spacing [gr]Delta[/gr]x), the smaller is this "critical radius", and the higher the inherent "critical undercooling" which has to be overcome.
In some occasions, this "critical undercooling" (or the equivalent curvature undercooling of a sphere with radius [gr]Delta[/gr]x) is written to the screen output (see here).
If this value is higher than the user-specified critical undercooling for nucleation, nucleation may be impossible, or nuclei stop growing before reaching the size of a full grid cell. This could be the case in your simulations where a lot of such remnants can be seen as blue dots. The solution then is to use "analytical_curvature" instead of "stabilisation" in the input for the given seed type and define (directly in the next line) a critical radius which is bigger than Δx). This model replaces the normal phase-field curvature which is maximal when the grain reaches full size, replaces it with an analytical value and thus implies a lower critical undercooling.
Bernd
-
lihaoge1994
- Posts: 9
- Joined: Mon Apr 02, 2018 2:33 pm
- anti_bot: 333
Re: Cr-Ni eutectic alloy simulation
Dear Bernd
I do not know why I can't login with my username, maybe because something wrong with my computer. So please allow me to use a new username temporarily.
I found a problem that the liquid phase still exists when the temperature is cooled below the eutectic line. I simulated with different cooling rates as 10K/s, 20K/s and 50K/s, however, there are eventually about 10% liquid phases exists.
I send the results to you, is that because there is a problem with the initial phase fraction? Should I adjust initial composition？
Thank you very much!
lihaoge
I do not know why I can't login with my username, maybe because something wrong with my computer. So please allow me to use a new username temporarily.
I found a problem that the liquid phase still exists when the temperature is cooled below the eutectic line. I simulated with different cooling rates as 10K/s, 20K/s and 50K/s, however, there are eventually about 10% liquid phases exists.
I send the results to you, is that because there is a problem with the initial phase fraction? Should I adjust initial composition？
Thank you very much!
lihaoge
- Attachments
-
- 10k.txt
- (15.13 KiB) Downloaded 306 times
-
- results.pptx
- (585.99 KiB) Downloaded 245 times
Re: Cr-Ni eutectic alloy simulation
Dear lihaoge,
To be honest, I do not understand how you obtained the diagrams on slide 1 of your attached ppt file. I assume that you use the moving_frame option so that there will always be liquid in the simulation domain?
Bernd
To be honest, I do not understand how you obtained the diagrams on slide 1 of your attached ppt file. I assume that you use the moving_frame option so that there will always be liquid in the simulation domain?
Bernd
-
lihaoge1994
- Posts: 9
- Joined: Mon Apr 02, 2018 2:33 pm
- anti_bot: 333
Re: Cr-Ni eutectic alloy simulation
Dear Bernd
I obtained the diagrams on slide 1 by the TabF.txt. In reality, there should be no liquid after some time. Is that if I use the moving frame option in order to track the solidification front, there will always be liquid phase exist?
In addition to the comparison with the experiment, what other methods can detect the correctness of the simulation?
Thank you all the times!
lihaoge
I obtained the diagrams on slide 1 by the TabF.txt. In reality, there should be no liquid after some time. Is that if I use the moving frame option in order to track the solidification front, there will always be liquid phase exist?
In addition to the comparison with the experiment, what other methods can detect the correctness of the simulation?
Thank you all the times!
lihaoge
Re: Cr-Ni eutectic alloy simulation
Dear lihaoge,
When using moving frame, of course, the domain can never solidify completely.
Therefore, instead of evaluating the total phase fractions of the domain (which is furthermore not at the same temperature due to a specified temperature gradient) in .TabF, you should rather check the phase fraction over temperature, e.g. by integrating the phase fractions given in the .frac* output files for each z-coordinate. You can do that by saving the .frac* outputs as ASCII file using DP_Micress and import e.g. in Excel, or by writing an own script for that purpose.
Bernd
When using moving frame, of course, the domain can never solidify completely.
Therefore, instead of evaluating the total phase fractions of the domain (which is furthermore not at the same temperature due to a specified temperature gradient) in .TabF, you should rather check the phase fraction over temperature, e.g. by integrating the phase fractions given in the .frac* output files for each z-coordinate. You can do that by saving the .frac* outputs as ASCII file using DP_Micress and import e.g. in Excel, or by writing an own script for that purpose.
Bernd