Discussion Forum

Very interesting problem,

You have use a different model than heat transfert with phase change acting on the Cp value. Do you have compared your phase change model with a simple model using comsol phase change feature?

You have to tell Comsol 2 conditions, gradT sign and T>Tmelt. It seems difficult to converge... Can you share your model or a more simple one?

Regards

Julien

Thanks for your answer.

For my issue the model of comsol isn´t sufficient because melting and solidification is there modeled as "the same just with other sign". But my material has reasonable differences between melting and solidifcation, meaning the curves for the enthalpy over Temperature look very different for melting and solidification. I attached a very simplyfied Picture for better understanding.

In my enthalpy based method the heat equations is the following:

(dH/dt)*rho=(d/dx)(k(dT/dx)

With Values out of DSC for H

For a normal melting i get with my in equation model self written enthalpy based method nearly the same results as the comsol model. And a whole cycle is no Problem too. My only provlem is to teach comsol to use the in my Picture marked red path. This should simulate solidification when the material isnt´t already completely solid.

If i would say use the the solidification curve if gradT is negative in this case, i would put Kind of energy in the System whithout physical reason. Thats why i want to follow the red curve. But i cant imagine how to do that with comsol.

Regards
Thomas

Very good Thomas,

Is your model is 1D?
Do you use some PDE/ODE for the enthalpy formulation? How is the convergence compared to the precooked phasechange feature of Comsol?

I was thinking that when the phase change is not complete, you have to follow the same curve (no switch on the other curve). So, I don't understand why your condition based on the sign of gradT doesn't work.. It will be easier to help (if I can) if I can see your file.
Your material have maybe a broad temperature width, not a pure material?

Maybe you can have a look on how is implemented hysteresis loop H-V curves for magnetic materials in Comsol.

Regards

Julien

Well my model is in 3D you can also replace the d/dx by an Nabla Operator just didnt´t mangaged to get the Symbol in the Forum so i made a 1D out of it ;-)

The Material is a PCM material, a Compound of some paraffine, binder and a material to improve thermal conductivity. I allready researched something about modeling such incomplete melt-solidification cycles and there are validated models in literature which tell you not to follow the same curve back but the red path shown in my attaced Picture from above.

I allready thought about magnetical Hysteresis, too. But even there i found no Comsol model until now which could help me with my Problem. (But i´m also not sure if Magnetization and Crystallization behave the same way.).

I think my model wouldnt be very helpfull. Ist just using a PDE for the heat equation. Material properties k and rho and the enthalpy curves which allready inclued Cp (which i´m not allowed to show). The if condition is also not spectacular. And with the Change from melting curve to solidification curve i have no Progress or idea how to handle it until now.

Hello Thomas,
Regarding : as you can see, you can use LaTeX notations in the Discussion Forum. See how at this link:
www.comsol.com/community/forums/general/thread/33025
Best,
Jeff

Thank you very much Jeff,

Looks better, next time i will use this: :)

Regards
Thomas

The problem is that you have 2 different temperatures for the same H(T) value,
Just an idea; is it possible to write the function T(H) and swith at the same temperature between melting of solidification curves? ...

Regards

Julien

I think every time i do a sudden switch from one to another curve is either not physically correct or makes numerical Problems. Another thing is that temperature is my dependent variable so enthalpy will not Change with time without a temperature Change.

I´m student of materials engineering, so i´m no expert in numerics, so correct me if I´m wrong with any of my statements.

Regards
Thomas

Can i say comsol something like:

Condition 1: if gradT<0 -> H(T) = H_melt(T)

Condition 2: if gradT>0 -> H(T) = H_sol(T)

Condition 3: if gradT=0 -> H(T) = const. as long as H_melt(T) < H(T) < H_sol(T) and dont care about
condition 1 and 2 in this intervall

Regards
Thomas

But temperature changes between melting curve and solidification curve (H constant), do you think that graT=0?

No in this Intervall gradT is for sure not 0,
thats why i want to ignore condition 1 and 2 for this Intervall, meaning if you notice gradT=0 Keep H(T) constant until H_melt>H>H_sol. If this is the case take care about condition 1 and 2 again

Hi Thomas,

It's possible to make that you want in storing the last temperature value with the nojac operator in a pde on the whole domain with simple if cycles but needs small time steps.
It works with thermal conductivity, for enthalpy you have to derivate I don't know if it's work.

Best regards

Julien

After several cycles of heating cooling and 2 non complete heating cycles, the thermal conductivity of the domain looks like the attached picture

Best regards

Julien