Weld solidification

[superabc]

Hi Bernd,

Thanks!
But there seems to still have some problems.

When I decreased the initial temperature at the bottom to 1735, 1730 or 1725 from 1740K and kept the temperature at which the initial equilibrium will be calculated unchanged, the dG value was still positive but it still run.
When I decreased the temperature at which the initial equilibrium will be calculated to 1735 or 1730 from 1740K, and kept the initial temperature at the bottom unchanged, the dG value was minus, but it still showed phase 1 disappeared at ... time.
When I decreased both two temperatures to 1735,1630 or 1625 from 1740K, the dG value was minus and it seemed to be correct.

I am confused these two temperatures meaning. I understand that temperature at which the initial equilibrium will be calculated is the real liquidus temperature from experiment. The initial temperature at the bottom is just the setting temperature. And these two temperature values are no relationship with coupled database. I understand that If cooling rate is not so high, initial temperature can be same as real liquidus temperature. If cooling rate is very high, it is better to set relatively higher initial temperature at the bottom than real liquidus temperature in order to keep stable at the beginning.

Superabc

[Bernd]

Hi superabc,

The initial temperature at the bottom is the important value which sets the temperature at the beginning and thus determines whether the phase will grow or not.

The temperature at which the initial equilibrium will be calculated is just a numerical parameter for initialisation of the phase interaction, like it is also typical in Thermo-Calc. The output of this initialisation is given in the .log file, thus the given temperature determines which driving force is written there, but apart from that it has (in most cases) no further influence.

Typically, both temperatures should be set to the same value (the temperature at which the simulation starts). But there are some cases where the (numerical) initialisation fails, and a temperature different from the initial temperature must be given to get a numerically correct initialisation. This is the reason why there are two temperatures to be specified!

Bernd

[superabc]

Hi Bernd,

Thank you for your detail explanation
I have already run simulation for about 6 days and stop running. When it run to about third or fourth day, university power turned off and I restarted. Right now, the questions are as follows:

1) FCC should appear at about 1700K or lower. However, it did not reach that step.

2) By checking .TabF, when temperature decreased to about 1663K, in the next running step the temperature and fraction of phase 1 keep stable. In my opinion, i guess that that is due to the limit of Z direction length when setting grid size. So, I think i should increase this size if i want to try again. is it right?

3) This time the purpose of my simulation is just to compare the results with some other paper's using same parameters. They also run using Micress. In the paper it showed the initial temperature at the bottom was 1740K, and the anisotropy between secondary phase was 0.2. However, in case of initial temperature 1740K, I have already asked you this question when the temperature was1740K, the phase 1 disappeared due to be not stable. In case of 0.2 anisotropy between secondary phase, you explained that right now there was no model corresponding to the anisotropy between same phase. Thus, i do not how they finished the simulation.

I attached some documents. Please!

All the best

[Bernd]

Hi superabc,

Let me try to answer your questions one by one:

1.) The reason why fcc phase is not formed is that the time interval for checking nucleation of fcc is chosen much too big, so that it did not happen yet (1.0 seconds!). As the cooling rate under welding conditions is very high, you should use much smaller intervals. Please also check other time intervals in the input file (e.g. updating of diffusion data or thermodynamic data) whether they have been reasonably adapted to the high cooling rate!

2.) In your simulation setup you use the moving frame functionality which makes the simulation domain follow the movement of the dendrite tip. This automatically has the consequence that, from the moment when the dendrite comes close to the top of the domain, the bottom temperature will not decrease anymore. For high-alloyed materials, and without using extremely high temperature gradients, it is typically not possible and reasonable to make the simulation domain that large that complete solidification is reached at the bottom.
Thus, it is up to you to use moving frame and to focus on the high temperatures (selection behaviour, tip temperature, fcc precipitation etc.) or not to use moving frame and to look what happens at lower temperatures (secondary phases, secondary arm spacing, segregation, etc.). A sort of compromise would be to use moving frame first until a stationary growth is reached, and then to make a restart from this point with the moving frame switched off. Please note that (at least for the present MICRESS version 6.1) the 1d_far_field option should not be used for the concentration field in this case, because it hurts afterwards when the dendrite tip hits the top of the domain!

3.) It could be, e.g., that they used a different database (with a slightly higher liquidus temperature), or that they started with big initial grains which only partly melt (but not disappear) until the temperature is lowered sufficiently. Which are the papers you are referring to?

Furthermore, I found two details in your input file which I want to comment on: First, the moving frame distance should typically be identical to the 1d distance for the 1d_far_field option. Otherwise you do not use the full extent of the domain. Secondly, I personally would use a smaller interface thickness of 3 cells. This reduces the numerical artifacts of the interface (solute trapping), or allows you to slightly increase the grid spacing and thus performance!

Bernd

[superabc]

Hi Bernd,

Thanks again for your advice.

Due to big size of paper (more than 3 M), I sorry i can not upload this paper. This paper (in Japanese) comes from conference proceedings of Japanese welding society.
Here I attach PPT which includes paper's title, abstract, mainly parameters and picture results.
In addition, our own database is TCFE 7 and MOB2.

I hope this time the result is good!

[Bernd]

Hi superabc,

Thanks really a lot for the information! It is always interesting to see scientific publications which contain MICRESS results

If you could give me the exact reference, we could include it in the MICRESS references on this website!

Bernd

[superabc]

Hi Bernd，

Right now, I only have two papers about Micress application. If you want, please give me your e-mail and i can send it to you. The size of each paper is more than 3 M, thus I sorry i can not upload. Another problem is these two papers are in Japanese. Maybe it is difficult to understand for most foreigners except Japanese and Chinese. It is some regret.

And in case of my own research, I still have some questions:

1) In .log file, the slop m(Ni)/Liquid is -0.18 but the slop m(Ni)/FCC_A1 is 0.32. I do not know why one is minor and another is plus.
2) Ni element content of our materials are 20.13% (input parameter). However, in .log file, Co(Ni)/Liquid is 20.133252 and Co(Ni)/FCC_A1 is 20.9277, which are both more than initial concentration. Theoretically, initial concentration should be between equilibrium concentration of solid and liquid. Here, the results are incorrect.
3) Also in .log file, i do not understand the meaning of dcdT(element)/Liquid or dcdT(element)/solid. Can you explain?
4) Yes, as you said the driving. mcr is important output, we can check whether it is stable or not by judging driving force fluctuation. But some times i think it is difficult to judge the boundary of fluctuation and not fluctuation. Can you help me judge our driving force is fluctuation or not? I will upload the driving force picture. And can you show me the fluctuation picture or not fluctuation picture?
5) In my research, dendrite tip radius is an important result. I want to know how to measure dendrite tip radius precisely. As a result of high cooling rate, the radius is too small. Right now, the measurement method which we used is as follows: Firstly, ImageJ software is employed to measure. Then, dendrite tip will be magnified as large as possible. In case of phase field graph, red region means liquid, blue region means interface and white region means solid. Afterwards, a circle can be made and then its size and position will be adjust until arc of this circle is tangent to the edge of white cell or just touches the corner. Finally, diameter of this circle could be measured by using line tool. However, I guess the error is too much. When magnifying dendrite tip as large as possible, the mosaic will appear. The size of circle will depend on the the shape of solid part mosaic. e.g. if solid part mosaic in tip is triangle the circle is small but if it in tip is plane like a quadrangle, maybe the circle is larger.

In order to understand what i said, i will attach some documents.

Best regards!

[Bernd]

Hi superabc,

I have sent you a PM with my e-mail address. It would be nice if you could send me the papers

Your questions:
1.) The slopes which you mention are calculated using multibinary extrapolation. This type of extrapolation of quasi-equilibrium data can lead to pseudo-demixing of single components, which is predominantly occurring in high-alloyed systems where interactions between elements are strong. The opposite signs which you mention are the reason why you are using the new ternary extrapolation scheme which makes the multibinary slopes useless. Otherwise they could make trouble!

2.) The slight deviation from the nominal composition is due to the exact way how we calculate the initial equilibrium: A fraction of 1-phMin (phase minimum) of liquid + phMin of fcc is taken as phase mixture. The mixture composition is calculated by weighted averaging of the initial phase compositions, which are either input by the user, taken from initial equilibria of these phases in other phase pairs, or taken from the default composition ("Start Composition for iteration of quasi-equilibrium" in the .log output) which are defined in the database. In your case, the composition of fcc is taken from this default composition which is 51.06662 wt%. This is the reason why the values are slightly above nominal composition.
But please keep in mind that this initial equilibrium does not define your initial condition but is only a numerical initialization to be prepared for the following "real" quasi-equilibria! Thus, this deviation has absolutely no influence on the simulation results.
Anyway, you could get much closer if you would reduce phMin or if you would input the initial compositions of all phases!

3.) dcdT is the extrapolation coefficient of the equilibrium compositions with temperature. You can find its definition and use in eq.(57) in Eiken et. al, Phys. Rev. E 73, 066122 (2006).

4.) I cannot see any sign of fluctuations! They would look like freckles or extremely high or low values changing from time step to time step.
But what I see is that the mobility is probably too low as there is no change of sign of the driving force when the curvature changes from positive to negative! This means that the kinetic undercooling is not smaller than curvature undercooling, and that perhaps curvature is not taken into account correctly!

5.) I also do not have a standard method for evaluating the tip radius, but it is obvious that you should use another type of output (not .phas) which contains the complete information of the diffuse interface, like the phase fraction output *.frac1.
What you could do then furthermore is, for example, to make an ASCII output of this result file and to perform a line-wise integration (e.g. in Excel). By plotting this integral of the fractions of a given line (=temperature) over the z-position you should get a much more precise shape of the dendrite tip. This shape you could then use for fitting circles or whatever.

Bernd

[superabc]

Hello Bernd,

I have already run last simulation for comparing with that paper in order to check our dirving file. In that paper, they used database FE-DATA ver.6, while my database is TCFE7. Last time you told me that different database will resulting in a slightly liquidus temperature difference. And right now i have some questions to ask you:

1) After simulation, i measured the dendrite tip radius. My value is about 0.36μm, while their value is 1.05μm. I am not sure whether only different database can lead to so different value. And we set the same mainly parameters to run, such as same dimension, same cooling rate, same temperature gradient, same interface energy, mobility and anisotropy and so on from that paper except initial temperature. Are there some other parameters can cause this difference?

2) Another problem is that two secondary dendrite arms along z direction grow fast, looks like primary dendrite. In fact, we only set three initial nucleation grains. Right now it looks like 5 primary dendrites. But, that paper' result shows only three primary dendrites. Can you explain this cause? And give me some suggestions how to set other parameter to eliminate these two long secondary dendrite arms.

3) From in. file, in the "phase diagram-input data", it generated automatically FCC_A1#2 in the "The database contains 4 phases:". And I only set three phases liquid, bcc_a2 and fcc_a1. I do not know why.

4) It is about calculated dendrite tip radius using KGT model. I tried to use different elements to calculate tip radius. However, the difference is too much. For example, when using Ni the radius is 0.81 μm, when using Mn the radius is 2.28 μm and when using Si the radius is 4.96 μm. Could you help me explain?

Here I attached the in. file, log. file and image at 0.31 solidification time.

Superabc

[Bernd]

Hi superabc

1.) Getting the correct (2D-) tip radius is not an easy task, because the tip radius depends on the correct tip undercooling, and this you will get only after calibration of the interface mobility or if the grid resolution is extremely high - in your case, the tip radius and the grid resolution are of similar size!
Apart from that, it is possible that they did not get it correctly, also...
An important parameter which you did not mention is the diffusivity of the elements in the melt. Using the standard mobility databases for steels, you will get only a value of 1E-5 cm2/s, which is not based on any measurements and which may be wrong...

2.) At least in one of the two papers, these "secondary" dendrites are also appearing, but are overgrown after a while. So, it seems that the situations are comparable, any slight change of parameters can lead to overgrowth or not!
The question whether such secondary dendrites form or not is strongly depending on the initial growth at the very beginning. You could e.g. suppress them by slowing down lateral growth along the bottom boundary condition by specifying "w" and a corresponding wetting angle in the boundary conditions for phase-field.

3.) This is done automatically by Thermo-Calc when you create the .GES5 file, if M(N,C) carbonitrides can be formed from the specified elements. Thermo-Calc treats FCC_A1#1 and FCC_A1#2 as one phase with two composition sets (existence regions in the composition space), while for MICRESS they would be two phases.

4.) You cannot calculate the Tip radius for each element, as they all together are relevant for the shape of the dendrite tip! As KGT model works only for binary systems, you should calculate a pseudo-binary phase diagram. I am not sure whether there is a standard procedure, I lately did it like that:

Pseudo-binär.png (25.26 KiB) Viewed 4781 times

Bernd