Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.

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.

Stationary and Time-dependent study for Joule Heating

Please login with a confirmed email address before reporting spam

Hello,

I am working on a rather simple simulation of Joule Heating on a Pt resistor deposited on AlN substrate. Although I can perform the stationary study and have some of the results I want, I cannot perform a time-dependent study. The execution fails with error message "Failed to find consistent initial values. Last time step is not converged.".

Any ideas what might be wrong?

Cheers!

-O.

3 Replies Last Post 22.05.2012, 05:50 GMT-4
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 22.05.2012, 01:17 GMT-4
Hi

there can be several things, such as BC too far from resulting conditions, or too steep changes (no ramp up ...). This can also be linked to the mesh not respecting the sampling criteria linking the heat diffusivity, the time stepping n and the mesh size. Check the forum on HT and time response and heat diffusivity (unfortunately COMSOl does not propose alpha as a standard mesn to check the mesh for HT, while it uses similar coefficents in other diffusion models.

Obviously their HT team havnt discovered yet that they are solving the same PDE equations as the chemistry team ;)

--
Good luck
Ivar
Hi there can be several things, such as BC too far from resulting conditions, or too steep changes (no ramp up ...). This can also be linked to the mesh not respecting the sampling criteria linking the heat diffusivity, the time stepping n and the mesh size. Check the forum on HT and time response and heat diffusivity (unfortunately COMSOl does not propose alpha as a standard mesn to check the mesh for HT, while it uses similar coefficents in other diffusion models. Obviously their HT team havnt discovered yet that they are solving the same PDE equations as the chemistry team ;) -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 22.05.2012, 04:51 GMT-4
Hello Ivar,

Thank you for your immediate response! I read your propositions but I'm not sure how I can detect the specific problem and how I can correct it, since I am a COMSOL newbie. Would it be to much if I asked you to take a look at the model?

Thank you in advance,

Odysseas
Hello Ivar, Thank you for your immediate response! I read your propositions but I'm not sure how I can detect the specific problem and how I can correct it, since I am a COMSOL newbie. Would it be to much if I asked you to take a look at the model? Thank you in advance, Odysseas


Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 22.05.2012, 05:50 GMT-4
Hi

a few comments:
Note with the Joule heating strain reference temperature 0[K] is rather cold, I assume you ment "Tref0"
As your AlN is not conducting You can exclude it from the Joule heating source
I'm not sure the h=10W/m/K usual for air exchange is valid at 80[K] as air tend to flow or even freeze at those temperatures, so obviously you have something else He or H2 I assume

You do get a reasonable result in stationary, if you double the mesh you get the same shape and it goes from 8.5[K] to 9.4 temperature increase. your sharp corners make some singulatities thast could explain the change in value, you could also fillet slightly these corners, to allow a better current and heat flow (w.r.t. FEM resolution) since your results is mainly singularity driven

The temporal case is probably not runnign as you mesh size, material alpha (heat diffusivity) and time stepping are not lined up, check the forum. also you should ramp up the voltage to avoid the turn on dirac function you are creating
Change your V0 to V0*step1(t[1/s]) and define step1() in the Definitions. Do not then forget to add a Parameter t= 1[s] so that your steady state simulation will run OK, and allow the time dependent solver to overwrite the parameter t value nicely. then you need to increase the mesh, and perhaps play with the time stepping settings of the solver, but it should work

Advice: turn on the plot while solving for aall solver steps for the time dependent case to see haw it goes

To simplify your model, you could consider the Pt layer as a bulk thin film and not a 3D volume, as you do not resolve anything with 1 mesh element in thicknes, then you would need >= 3 and for your time scales you will not see any gradients therein, in my opinion

--
Good luck
Ivar
Hi a few comments: Note with the Joule heating strain reference temperature 0[K] is rather cold, I assume you ment "Tref0" As your AlN is not conducting You can exclude it from the Joule heating source I'm not sure the h=10W/m/K usual for air exchange is valid at 80[K] as air tend to flow or even freeze at those temperatures, so obviously you have something else He or H2 I assume You do get a reasonable result in stationary, if you double the mesh you get the same shape and it goes from 8.5[K] to 9.4 temperature increase. your sharp corners make some singulatities thast could explain the change in value, you could also fillet slightly these corners, to allow a better current and heat flow (w.r.t. FEM resolution) since your results is mainly singularity driven The temporal case is probably not runnign as you mesh size, material alpha (heat diffusivity) and time stepping are not lined up, check the forum. also you should ramp up the voltage to avoid the turn on dirac function you are creating Change your V0 to V0*step1(t[1/s]) and define step1() in the Definitions. Do not then forget to add a Parameter t= 1[s] so that your steady state simulation will run OK, and allow the time dependent solver to overwrite the parameter t value nicely. then you need to increase the mesh, and perhaps play with the time stepping settings of the solver, but it should work Advice: turn on the plot while solving for aall solver steps for the time dependent case to see haw it goes To simplify your model, you could consider the Pt layer as a bulk thin film and not a 3D volume, as you do not resolve anything with 1 mesh element in thicknes, then you would need >= 3 and for your time scales you will not see any gradients therein, in my opinion -- Good luck Ivar

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.