Conjugate Heat Transfer Simulation of a Desktop Computer
HTML-код
- Опубликовано: 21 авг 2024
- 5 [mm] conformal mesh using the polyhedral + thin meshers, with 8.8 million cells
Some notes / things I would fix if I were to do this again. I'd like to think that if I fixed these I would have a truly professional model.
-The hotspots at the VREGs and chipset were from having NO idea what the power draw would be
-Contact resistances weren't included in most scenarios, future refinement would include calculation of BGA resistances and 'solid' air gap resistance
-Fan curve on the CPU cooler was just a modified version of the 140mm case fan P-Q curve since the supplier chose not to release the data.
-Bottom vent / fan filter wasn't modeled. Although due to the negative pressure, I should have modeled it and added case feet.
-Power supply wasn't modeled as I didn't want to open it up for warranty / scary capacitor reasons and since I'd no clue what the heat generation / material properties / fan curve would be.
-Capacitors / cables / connectors could be included in the future for a very small increase in accuracy.
-Fan hubs were just a block of plastic with a heat generation ( 0.5 [W] ) used to model fan losses. The heat generation could be reduced but the modeling issue is unavoidable. Can't exactly take that part of the fan apart. Really no effect on accuracy.
-Fan efficiency (for fan swirl) was set to 0.50 because I have no clue what the real fan swirl would be.
-In hindsight I forgot to setup a rotation of the fan hubs (whoops!) probably a really, really small effect on accuracy.
-Fan filter P-V curve calculated from a paper on calculating the pressure drop over a simple woven bug screen. Future refinement would involve generating my own P-V curve in CFD
-Sphere boundary was a stagnation outlet, but I had issues with the natural pressure gradient due to gravity causing flow through it. I was able to fix this using a pressure outlet and a boussinesq model + constant density gas (as opposed to ideal gas which is what I used), but it made the simulation blow up after a hundred or so iterations. I'm sure Siemens knows how to fix this issue but I haven't asked them yet.
- Heatpipe fluid effective thermal conductivity was a total guess. Future improvement would involve doing stress tests on my computer and matching my model to it
- I have an updated version with a more accurate modelling of the porous media (using porous baffles and porous media), this isn't the better model.
- Didn't include anything in the rear partition of the case, like the SSDs or cables.