Solidification simulation under casting

[betleenkim]

Dear Bernd

How are you?
Long time no questions

These days i am trying to optimize a simulation with slow cooling rate.
But i have thickened boundaries as given below with slow cooling rate. Does this can be solved by reducing grid size?
How can i improve the quality of result?

Thickening of boundary.pdf

(96.37 KiB) Downloaded 285 times

Best regard

[Bernd]

Hi betleenkim,

your suggestion seems quite plausible: a too low grid resolution would result in such a broadening of interfaces which you showed us!

I assume that this type of simulation before was running successfully at higher cooling rates and higher grid resolution, and that you probably reduced resolution too much when switching to low cooling rates, right?

But in general, other types of problems can also lead to broadened interfaces...

[betleenkim]

Dear Bernd

Thank you for your reply.
As you mentioned, i do not have such a kind of problem with higher cooling rate.

Anyway, what is the other types of problem that causes such a problem. The simulation time will increase as i reduce grid cell size.
I firstly try to solve this problem with other causes.

Best regard

[Bernd]

As you mentioned, i do not have such a kind of problem with higher cooling rate.

What did you exactly change between the runs with higher and lower cooling rate?

[betleenkim]

Dear Bernd

I changed just cooling rate, 0.4k/s and 20k/s.
Then i do not have broadened boundary with higher cooling rate.

Best regard

[Bernd]

Hi betleenkim,

then it is definitively NOT a problem of grid spacing, because in that case we would expect problems with the high cooling rate!

You should check for other possible issues:

- check the driving force of those interfaces before and when they start to broaden (.driv output). Normally it should look quite smooth, the value of the driving force (at non-curved regions of the interface) should be in the same order of magnitude as the effect of curvature (difference of driving force between curved and non-curved regions). If the (absolute value of the) driving force is constantly too high, mobility is too low. If the driving force is fluctuating, mobility is probably too high.

- do you get any error messages?

- is the broadened interface consisting of dual interface or triple junctions (.intf output, 1 is dual, 3 or higher is triple junction)?

- check for other obvious differences between fast and slow cooling

Bernd

[betleenkim]

Dear Bernd

Thank you for your reply

I attached the result file form dri, phase and intf.
As you can see, there is not much fluctuation or high value of driving force. And the driving force is similar in broadened interface.
In addition, the problem occurs mainly at the higher junctions i think.

Broadening.pdf

(107.07 KiB) Downloaded 310 times

Secondly, i do not have any error messages both of the simulation with high and low cooling rate.

Finally, the obvious difference is the phase fraction between two simulations. Since the system goes to equilibrium condition at low cooling rate, the system has higher amount of ferrite phase during cooling.

Best regard

[Bernd]

Hi betleenkim,

From the outputs you show I get the impression that you probably do not have any broadened interfaces: The high order triple junctions (.intf output above 60!) mean that you obviously have a huge number of ferrite grains which are nucleated along the grain boundaries. In the .phas output, an interface between ferrite grains also appears as blue (-1), so that I guess the "broadened" region is just full of many small overlapping ferrite grains!
You can check this if you look at the .frac1 or .frac2 output: Probably, the interface between the two phases (ferrite and austenite) is not smeared at all!

If this is the case, I suppose the reason is that the seed types for ferrite nucleation which were working reasonably for fast cooling, are creating too many seeds in case of slow cooling. You can change that by either just "rescaling" the time parameters in the nucleation input (like checking interval, shield time) to the longer time-scale of slow cooling, or by making the nucleation scenario more "physical" (e.g. by using a higher value for the critical seed undercooling, so that with the lower supersaturation at low cooling rates nucleation is reduced automatically).

Bernd

[betleenkim]

Dear Bernd

Thank you for your reply
I will apply your suggestions for my simulation and see the result.

Best regard

[betleenkim]

Dear Bernd

How have you been?
Long time no questions about my simulation;;

So far i simulated very long cooling time and i finally get the result.
After cooling, i reheated my simulation domain at a heating rate of 20K/s.
However, although i did the simulation which is closed to the real situation in terms of cooling and heating rate, i could not get corresponding result with my experiment. So i attached summary file for my simulation and real situation.

In addition, it seems that the database (TCFE7) from thermo_Clalc cannot describe well for the partitioning of Si!!
In steel field, Si is known to be ferrite stabilizer and enriched in the ferrite area when dual phase such as ferrite and austenite is coexist. However, Si behaves totally opposite manner in both of simulation and thermo_calc.
In my opinion, Si should behave similar manner to Al. If then, more ferrite will survive at the elevated temperature due to stabilization effect from Si and Al.

Mismatch with simulation.pdf

(465.5 KiB) Downloaded 305 times

Best regard