Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
Frequency-transient Electromagnetic Heat Source & Heat Transfer with Phase Change
Posted May 21, 2015, 6:24 a.m. EDT RF & Microwave Engineering, Heat Transfer & Phase Change, Materials, Mesh, Modeling Tools & Definitions, Parameters, Variables, & Functions, Studies & Solvers Version 5.0, Version 5.1 6 Replies
Please login with a confirmed email address before reporting spam
I have created a coupled multiphysics model which refuses to solve. The model consists of a 2D periodic domain containing an EM wave incident onto a surface which uses the 'Heat Transfer with Phase Change' physics type to simulate a phase change between two states with different EM properties. The EM wave heats the phase change material and hence there is two-way coupling between the EM and thermal simulations, so the situation is non-linear.
When I begin a frequency-transient study, COMSOL correctly solves up to the transition temperature but then iterates ever closer to that temperature without ever arriving at it. Additionally, if I begin the simulation at a temperature within the transition interval, COMSOL is unable to define the initial state and the simulation fails.
Is there any way I can make this work?
Many thanks
Robin
When I begin a frequency-transient study, COMSOL correctly solves up to the transition temperature but then iterates ever closer to that temperature without ever arriving at it. Additionally, if I begin the simulation at a temperature within the transition interval, COMSOL is unable to define the initial state and the simulation fails.
Is there any way I can make this work?
Many thanks
Robin
Attachments:
6 Replies Last Post Jun 2, 2015, 11:23 a.m. EDT
Please login with a confirmed email address before reporting spam
Posted:
10 years ago
Hallo,
sadly i dont have the required module to open your file. However, i had created many models to simulate other types of contact-less heat coupled with phase change. Therefore, i faced your problem multiple of times and i can give you hence some tips to reach convergence.
1. If your material properties are not dependent on temperature or the expected temperature increase is not dramatic (within a few 10s of K) that the material properties stay constant; then it is advantageous to decouple your system and create a solver with step1: frequency domain and step2: time dependent. Comsol knows automatically, that the answer of step1 will be used in step2. The trickiest part is to create a variable in definition for the places where the heating occur and give it a name such as Q1. The function is the source of the heating, example: Physics.Qtot (check equation view to know the correct name). Finally manually create a heat source on the domains that are only subjected to heating. Don't forget to decouple the physics in this case.
Remark: using this method, you might have an error in the frequency domain solver (something like iomega undefined). That occurs if you define the frequency in the solver as a constant, that trick to solve this problem is to define a parameter called "freq", and define the frequency value there and use it in the frequency solver.
2. Phase change is numerically challenging and sensitive physics. Therefore, I suggest that you either try to refine the mesh or increase the value of transition region to help your system to converge.
I am sure that your model will converge if you used one or both explained methods.
Good luck,
Zedd
sadly i dont have the required module to open your file. However, i had created many models to simulate other types of contact-less heat coupled with phase change. Therefore, i faced your problem multiple of times and i can give you hence some tips to reach convergence.
1. If your material properties are not dependent on temperature or the expected temperature increase is not dramatic (within a few 10s of K) that the material properties stay constant; then it is advantageous to decouple your system and create a solver with step1: frequency domain and step2: time dependent. Comsol knows automatically, that the answer of step1 will be used in step2. The trickiest part is to create a variable in definition for the places where the heating occur and give it a name such as Q1. The function is the source of the heating, example: Physics.Qtot (check equation view to know the correct name). Finally manually create a heat source on the domains that are only subjected to heating. Don't forget to decouple the physics in this case.
Remark: using this method, you might have an error in the frequency domain solver (something like iomega undefined). That occurs if you define the frequency in the solver as a constant, that trick to solve this problem is to define a parameter called "freq", and define the frequency value there and use it in the frequency solver.
2. Phase change is numerically challenging and sensitive physics. Therefore, I suggest that you either try to refine the mesh or increase the value of transition region to help your system to converge.
I am sure that your model will converge if you used one or both explained methods.
Good luck,
Zedd
Please login with a confirmed email address before reporting spam
Posted:
10 years ago
Hello Zaid and thanks for your input.
Looking at 1) - decouple the EM and thermal simulations - sadly this is not an option. As you say, this works if material properties stay constant, but unfortunately they're not. If I do simulate the system with a material independent heating function that mimics laser heating, the simulation completes, but this is not an accurate description of the real situation - the real heating function is dependent on phase state.
2) looks more promising. I reduced the mesh size in the area containing the phase change material to 1nm maximum (from 40nm). I then increased the temperature range over which the phase transition takes place (I assume this is what you mean by 'value of transition region') to 10K. The solution still doesn't finish.
It's important to note that my simulation does converge and doesn't report any errors. Instead it iterates endless convergent solutions closer and closer to the start of the phase transition, but never actually arriving at it. Also, if I set the initial temperature at a value within the temperature range of the phase transition, COMSOL fails to find a valid set of initial values.
Looking at 1) - decouple the EM and thermal simulations - sadly this is not an option. As you say, this works if material properties stay constant, but unfortunately they're not. If I do simulate the system with a material independent heating function that mimics laser heating, the simulation completes, but this is not an accurate description of the real situation - the real heating function is dependent on phase state.
2) looks more promising. I reduced the mesh size in the area containing the phase change material to 1nm maximum (from 40nm). I then increased the temperature range over which the phase transition takes place (I assume this is what you mean by 'value of transition region') to 10K. The solution still doesn't finish.
It's important to note that my simulation does converge and doesn't report any errors. Instead it iterates endless convergent solutions closer and closer to the start of the phase transition, but never actually arriving at it. Also, if I set the initial temperature at a value within the temperature range of the phase transition, COMSOL fails to find a valid set of initial values.
Please login with a confirmed email address before reporting spam
Posted:
10 years ago
Bumping - I still need some help with this.
Please login with a confirmed email address before reporting spam
Posted:
10 years ago
It seems that no community members have been able to help with this. What's my next step here? Contact COMSOL support directly?
Please login with a confirmed email address before reporting spam
Posted:
10 years ago
Hi Robin,
I don't see in your model the coupling between emw and heat transfert. You need to define a source term and choose in the list which source you want to heat the gallium
You need to use a finer mesh and increase mushy zone width. Try a segregated solver approach like step 1: emw and step 2: heat transfert
Best regards
Julien
I don't see in your model the coupling between emw and heat transfert. You need to define a source term and choose in the list which source you want to heat the gallium
You need to use a finer mesh and increase mushy zone width. Try a segregated solver approach like step 1: emw and step 2: heat transfert
Best regards
Julien
Please login with a confirmed email address before reporting spam
Posted:
10 years ago
Hi Julien,
Thanks for taking a look at the model!
The coupling between emw and heat transfer is in [Multiphysics->Electromagnetic Heat Source 1]. It definitely works.
I have tried using a finer mesh and increased mushy zone - file attached. No improvement.
I can't use a segregated solver because my model depends on the phase change affecting reflectivity.
Thanks for taking a look at the model!
The coupling between emw and heat transfer is in [Multiphysics->Electromagnetic Heat Source 1]. It definitely works.
I have tried using a finer mesh and increased mushy zone - file attached. No improvement.
I can't use a segregated solver because my model depends on the phase change affecting reflectivity.
Attachments: