Diverging results on heat sink simulation

Hi folks,

I’m trying to simulate a heat sink to see what size and parameters would be good starting point for the actual build.

I’ve spent a bit of time browsing through demo projects and have tried to replicate the techniques I’ve seen, but so far no luck. I keep running into issues where a “velocity field started diverging” or a “temperature field started diverging”. On one run the device temperature began to climb gradually (up to a very realistic point) then shortly after started going down and errored out. I’ve tried messing with most of the mesh settings which ultimately ends in a meshing failure, often “ran out of memory”. I’ve got result control on the device temp and on the air outlet. I’ve also tried this with a few different variations of the CAD model with exactly the same errors.

The model consists of the heat sink, a mock up of the device generating the heat (an LDMOS transistor), and a shroud to direct the airflow. Based on my testing with these LDMOS devices I’ve determined that a skived copper heat sink is my best chance to maintain a safe junction temperature without going to water cooling or heat pipes.

I estimate that the device will generate up to around 750w of heat, and that the case temperate should be no higher than 120degC.

I’m very new to this type of modelling so any tips are appreciated. Thanks!

Hi!

A few pointers/questions about the setup:

  • Can you double check the velocity inlet BC? We are currently looking at around 1000 m/s:




  • I’m a little bit concerned about the mesh non orthogonality (max 86 at the moment). Some changes in the mesh may be necessary, to try and improve that.
    This geometry is a little difficult to work with, but you might be able to get a better non orthogonality by using 1 layer instead of 3 and also applying further refinement on the fins. See this article for more info.

  • Lastly, if the runs are still unstable, it might be worth it to decrease the temperature relaxation a little bit. Perhaps 0.5 or 0.6 would help a little bit.

1 Like

Thank you for pointing out the velocity… I mixed up my units there. Should have been in the ballpark of 0.08m^3/s. I think the velocity has a lot to do with why this simulation is struggling. I created a smaller model with similar geometry in the hopes of trouble shooting this model and it seems the velocity has a big impact on the stability of the simulation.

Thanks for linking the article as well, very informative. I’ve been paying attention to the log as I try different geometries and it seems the non-orthogonality is almost always 75 or more, even with 1 layer and refinement. Any tips for reducing this further?