MICRESS Forum

Dear Bernd

The simulation is now going well.

I have several technical questions about 1D-temp field.

1. I have read your manual. However, i cannot fully understand the mechanism of pseudo_3d.
In your manual, we have to turn on pseudo_3d when solutal diffusion fields do not touch the calculation domain boundaries.
What is the meaning of the sentence " solutal diffusion fields do not touch the calculation domain boundaries"?
In addition, how can i define the critical matrix fraction?

2. In my simulation, i put two grains (BCC) at the two bottom corners to simulate delta-ferrite dendrite growth. And as temperature decreases, there is nucleation of austenite at the interface between delta-ferrite and liquid. In that case, should i turn on pseudo_3d for both of phase FCC and BCC?

3. In 1D-temp mode, i need to define the features of 1D-temperature field such as entalphy, Cp and so on.
In your AlCu example, you used information from the file named "AlCu_Temp1d_latHeatData".
How do you get such information?

4. How can i control cooling rate? Should i adjust "heat transfer coefficient?"

4. When i turn on just latent heat mode (no 1d-temp), i have to define heat flow.
I am wondering that what this heat flow is. Heat flow from the bottom to the out of domain? or heat flow inside the simulation domain?

Sorry, i have many questions. And not organized also. But still i am looking forward to your reply.

Best regard

Hi betleenkim,

These are really many questions, let me try to give the answers:

1.) The good news about pseudo_3D is that it is possible to make corrections for the systematic error which arise from 2D vs. 3D simulations. The bad side is that it is not easy to do this in a correct way...
Chapter 4.11 of the MICRESS manual describes in detail how this correction is working. As is described there, a good correction is possible as long as the particle or dendrite under consideration is growing freely. In this case, one can just "expand" the 2D domain in a corresponding way into the third dimension, and get a corrected value for the 3D phase fraction (and thus latent heat amount). If there is no free growth (e.g. Scheil-like conditions), then the phase fraction must not be corrected, as the fraction is dominated by thermodynamics and therefore already correct!
The condition for when one case is converting to the other is when the diffusion fields of the growing particle are touching the domain boundary (or the diffusion fields of other particles): Then, thermodynamic limits for the phase fraction start to get important.
Therefore, a good estimate for the critical fraction of the matrix phase (e.g. melt) for terminating correction should be when the diffusion fields start touching. This you can estimate from the simulation itself.
Even if the matrix fraction, where correction should be stopped, is not so easy to estimate, it is still better to estimate than not to correct for 2D effects at all!

2.) In case of a peritectic reaction, only the BCC phase is growing freely, and only at the dendrite tips! If you consider directional solidification using coupling to a 1d temperature field (1d_temp), then the pseudo_3D correction is performed for each z dimension range which is corresponding to one grid cell of the 1d temperature field. Therefore, you should choose the critical fraction of liquid such that the correction applies to the dendrite tips, but not to the lower regions where the diffusion fields of the two dendrites are already touching. FCC should not be corrected at all because there is no free growth!

3.) You get this information directly from the .dTLat output! Therefore, the question is just how to start with the first iteration. You can either start from a simulation without 1d_temp but with lat_heat and add the heat conductivity values by hand, or start from data of a similar alloy/case.
A trick for the first iteration is to already use 1d_temp, but make the 1d temperature field exactly as long as the height of the 2D/3D simulation domain. Then, it doesn't matter if the first dataset is completely wrong, because it is only applied outside the 2D/3D domain

4.) However you like! It depends on the process itself which boundary condition is most realistic. The local solidification conditions depend on both, the distance of the 2D/3D domain from the boundary condition (which is important for the temperature gradient) and the boundary condition itself. I typically use a constant heat transfer coefficient, because I think it is most realistic for typical casting conditions where cooling is not explicitly controlled.

4. ) In that case, temperature diffusion is not solved, and heat extraction is considered as uniform. This is realistic for small particles like in Differential Thermal Analysis (DTA). Heat extraction (per volume) as well as latent heat lead to a uniform rising or dropping of temperature (for this as well as for the homoenthalpic approach see B. Böttger, J. Eiken, M.Apel, Phase-field simulation of microstructure formation in technical castings – A self-consistent homoenthalpic approach to the micro–macro problem, J. Comput. Phys. 228 (2009), 6784-6795.)

Best wishes

Bernd

Dear Bernd

Thank you for your answers

But i have uncertanity about the answer 3.

I firstly simulate my project with lat_heat mode turning off 1D_temp field.
Like you said, i got the .dTLat file containing thermodynamic parameters.
Based on the value in the file, this time i turned on 1D_temp field. And put the entalphy and Cp value as constant based on the file.
However, the phase 1 is always disappeared at the beginning of the simulation. I attatched my driving file.

Secondly, in your answer number 3 "the question is just how to start with the first iteration. You can either start from a simulation without 1d_temp but with lat_heat and add the heat conductivity values by hand, or start from data of a similar alloy/case" Is this sentence means that the 1d_temp mode is equivalent to the lat_heat mode? what does the first iteration mean? I do not think that there is no only first iteration in the lat_heat mode. Once i turn on the lat_heat mode, the simulation is going straighforward. And i cannot turn on the 1d_temp mode during simulation.

Best regard

Hi betleenkim,

Please have a look on the example "AlCu_Temp1d_dri" which comes with the MICRESS redistribution! In this case, the latent heat data are read from the file "AlCu_Temp1d_latHeatData" which is actually a copy of a suitable .dTLat file. This is the way you should do it! With a constant enthalpy value only, there is no latent heat

With iteration I did not mean those inside one simulation, but iterative repeating of simulations. The first simulation is based on the estimated latent heat data or those taken from a similar simulation. The second one is then using the new .dTLat output from the first iteration, and so on. You repeat several times until the results do not change any more. This is the idea of the "Iterative Homoenthalpic Approach".

Best wishes

Bernd

Dear Bernd

Thank you so much for your answer. Now i can run the simulation successively
So i can simulate solidification with different position from surface.

Hear is my another question.
I try to reheat the result from 1D temp field simulation. Is it valid that i still apply 1D temp field for the solid state phase transformation?

Best regard

Dear Bernd

Sorry for another question;;

In 1D temp filed assumption, we have to put initial values for starting a siumation.
I basically got the values from .dTLat file that can be got successive simlation in lat_heat mode.
In the file, we have to put Avg. Lambda(?) [W/cm/K] value to get initial temperature gradient. Am i right?
But i cannot distinguish between the value and the heat transfer coefficient that i have to input in the driving file when i assume "constant" temperature in B-direction. What is important for the heat transfer coefficient even though i already input the Avg. Lambda.

Best regard

Hi betleenkim,

If you finished a solidification simulation which uses the 1d_temp option, the normal way to continue with e.g. a heat treatment simulation would be to use the restart option, and change the boundary condition of the 1d temperature field accordingly. At present it is not possible to continue without 1d_temp option, the reason is the format compatibility of the restart file (.rest). But this will be possible in future, I am working on it right now!
If it is important for you to have e.g. a constant heating rate in the second process step, you could use a trick (I never tried that but it should work): When reading the .rest file, change the temperature boundary condition on one side of the 1d temperature field to a constant heating rate (by reading a temperature condition from file) and shift the position of the microdomain to this boundary!

In your second question, I think you are mixing up different things:
- The average lambda (heat conductivity [W/cm/K]) which you have to specify in the 1d_temp simulation is needed for calculating the temperature field during simulation. (After the first iteration of the Homoenthalpic Approach, these values are automatically updated and written to the .dTLat, so that you can use them in the next iteration.)
- The initial temperature gradient you input via the initial temperature at the bottom and at the top (or by reading the initial temperature field from a file).
- The heat transfer coefficient [W/cm2/K] which you can input together with the temperature boundary condition at the bottom or top of the 1d-temperature field describes heat transfer between an outer medium (e.g. a quenching bath with the given temperature) and the boundary (outer grid cell) of the 1d-temperature field. It has nothing to do with heat conductivity lambda inside the casting (1d-temperature field).

Bernd

Dear Bernd

Thank you very much for your answer.

Now i am trying to do heating simulation with 1d_temp field.
However, there is linearization problem during simulation.

At the beginning there is no problem at all. However, as time reached around 7s, MICRESS cannot proceed simulation showing "try harder" message. As i asked previously, "try harder" message can be partly overcome by resetting mobility data. But i think that the mobility is not problem since the simulation goes well when i simulate cooling section.

I attached the driving file for the heating simulation. Could you check it for me? If i make mistake constructing driving file?
For the information i give you the content of .dTLat file for starting simulation as below,

# Time of Temperature fraction Delta T. Delta T. Cp effect. Avg. lambda Avg. H dH/dT enthalpy enthalpy enthalpy
# sim. [s] [K] solid Lat.Int.[K] Lat. Ext.[K] [J/(cm^3*K)] [W/cm/K] [J/cm^3)] [J/(cm^3K)] phase 0 phase 1 phase 2
0.000 749.2038 1.0000 -286.38 -1000.5 4.2049 1.9786 1423.565 6.4079 0.000000 897.0423 1809.916
0.01 749.1857 1.0000 -286.38 -1000.5 4.2049 1.9786 1423.487 6.4374 0.000000 896.9561 1809.848




Best regard

Hi betleenkim,

it is important to check whether temperature is changing in the way you want. If the heating rate inside the simulation domain is in a similar range as the cooling rate before, simulation should work properly. But if the heating rate is much higher, the lengthscales could change and anything can happen! You can check temperature e.g. in the .TabL output.

I wonder whether the .dTLat output you show is what you are using as input (3wt%_lat_heat_dTLat) or whether it is the first output from the heating simulation. This file which you show could only be used as input in the temperature range between 749.1857 and 749.2038! The temperature range in this file should include all temperatures encountered in the 1d-temperature field, otherwise strange things may happen due to unrealistic extrapolation!
By the way, the interval between .dTLat outputs is the same as for the .TabL and can be specified with the tab_log keyword in the "Selection of outputs" section of the driving file.

Apart from that I cannot see anything strange in the drivint file, and I would need more information to understand what is going wrong. I assume that you successfully did the solidification simulation using 1d_temp option, and simulation ended at about 750K.

The first thing I would check is whether the 1d-temperature field is correctly read. At the beginning of heating it should be exactly like at the end of the solidification simulation (check _1DTemp.temp file). Then, with changing the boundary condition of the 1d-temperature field to 1550K, temperature should rise correspondingly, and (shortly afterwards) also temperature in the microstructure domain.

If all that is ok, then you should check what is happening at the time when you see the errors. Is there nucleation, are there extreme driving forces? You can trace the grid cells where the problem occurs with the .numR output: negative values indicate errors during updating thermodynamic data.

Please keep me informed about your success

Bernd

Dear Bernd

Thank very much for your answer.

First of all, I changed heat transfer coefficient that is considered to effect on the heating rate. And i will see the result.

Secondly, for the "3wt%_lat_heat_dTLat", i used this file for initiating the simulation. And the output .dTLat file contains every other information with respect to temperatures.
In my driving file, i put the bottom and top temperatures as 750K then i think "3wt%_lat_heat_dTLat" file contains every information about the 1D temp field. Am i wrong?

Thirdly, i put the "temp" image at the end of solidification and the at the beginning of heating simulation. As you can see, the temperature is slightly different. So i further reduced the heat transfer coefficient to avoid dramatic change in the temperature field at the beginning of heating.

Finally, i am sorry that i cannot find .numR file;; How can i get this file? I think the among the solutions you suggested, this is most relevant reason for error since i found the error message at specific time and temperature. If then, what is the solution that i can avoid the error?

I will send the information files for my simulation via an e-mail, since the files size is rather large so i cannot upload it.

Best regard