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.

Artifacts in temperature profile across interfaces

Please login with a confirmed email address before reporting spam

I am a relatively new COMSOL user, and I am having trouble with the Heat Transfer module (Multiphysics 5.5) where there are artifacts in the temperature profile of a system with interfaces or boundaries when solving the time dependent solution.

To illustrate the problem, I made a simple mock file (attached) that has a 2D geometry of alternating metal and water square domains (picture attached), where the initial temperature of the water is 300 K and the metal has a linear Gaussian profile of elevated temperatures spanning multiple domains. At early times (0-0.1 ns, picture attached) there are temperature spikes at the interface, and at some time points (0.01 ns, picture attached) the temperature profile within some domains is somehow inverted from expected, and at later times the whole temperature profile is inverted from expected with lower temperature in the center than the edges (10 ns, picture attached). In this case I established an interface conductance that is relatively high (faster than the timescale of the simulation), but the problem persists. I've also seen similar problems even with a slab of say metal with a 2D Gaussian temp profile, including ripples and spikes in the temperature profile.

I would greatly appreciate input on how to remedy such unphysical artifacts!



2 Replies Last Post 11 juin 2020, 14:22 UTC−4
Henrik Sönnerlind COMSOL Employee

Please login with a confirmed email address before reporting spam

Posted: 4 years ago 11 juin 2020, 10:08 UTC−4
Updated: 4 years ago 11 juin 2020, 10:08 UTC−4

Hi,

If you look at the Y-axis legend for the 10 ns plot, you will see that there is only one single value. What you see is just numerical noise of the order of 1 mK. In practice, the value is constant.

By plotting T-300 or T-300.063 instead of T you can examine the values more in detail.

As for the ripple at small times, that will probably decrease if you tighten the tolerance by some orders of magnitude.

-------------------
Henrik Sönnerlind
COMSOL
Hi, If you look at the Y-axis legend for the 10 ns plot, you will see that there is only one single value. What you see is just numerical noise of the order of 1 mK. In practice, the value is constant. By plotting T-300 or T-300.063 instead of T you can examine the values more in detail. As for the ripple at small times, that will probably decrease if you tighten the tolerance by some orders of magnitude.

Please login with a confirmed email address before reporting spam

Posted: 4 years ago 11 juin 2020, 14:22 UTC−4

Thank you Henrik, that made a big difference but I wouldn't have thought to look at the study tolerance to fix it on my own! That also flipped the 10 ns plot to a physically meaningful orientation (though I understand the actual values should be taken with a grain of salt).

Thank you Henrik, that made a big difference but I wouldn't have thought to look at the study tolerance to fix it on my own! That also flipped the 10 ns plot to a physically meaningful orientation (though I understand the actual values should be taken with a grain of salt).

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.