Cooling problem

[Juha]

Hi,

If that "body" is supposed to be part of air flow domain, you should activate the Navier-Stokes
equation for that? Now i think only heat equation belongs to Equations set 1? Also then you
should not have (velocity) BCs on the bondaries of that body. Maybe i misunderstood ?
-Juha

[tibich72]

Hi Juha,

well, that body (the heatsink), should not be part of the air domain, but maybe I don't understand how to set up the model. Basically, I have a body that models the air, and for that I solve the heat equation as well as Navier-Stokes; all the other bodies (including the heatsink) are not fluid, so only the heat equation is solved for them. So yes, for body #2, only the heat equation is applied.

Are you suggesting I should (a) I eliminate the boundary conditions if I apply only the heat equation to solids or (b) apply Navier-Stokes to the solids too?

On a (possibly) separate note, I've added the wall law to the boundaries between solids and air, and there was no change in the pattern of airflow.

Thanks,
Tibi

[Juha]

Hi,
no i'm suggesting neither a) nor b). I'm afraid i just don't quite understand the original problem,
of "flow going around the heatsink not through it" then. Sorry to be somewhat slow on this...
Juha

[tibich72]

Right, sorry, I should've started with the geometry. Here's the system without the air box:

3d_view.jpg

Model without airbox

(206.7 KiB) Not downloaded yet

So, I have a component (at the bottom, with the rounded edges) and on top of it is a heatsink (the one with a lot of rectangular fins -- it's not the most performant heatsink out there, just something I could create fast). The slice images that I was uploading are roughly half-way through the fins, parallel with the bottom of the heatsink (the Z axis in the figure is orthogonal to the plane of the slice). The heatsink is made of aluminum and is solid. However, there are gaps between the fins and I'd expect the air to flow through them -- but, as the simulation indicates, there's no airflow in the area of the heatsink (between the fins), only around the entire heatsink, which probably accounts for overheating.

Hope it's clearer now -- let me know if you need a different view or file.

Thanks,
Tibi

[Juha]

OK got it, thanks. I don't have immediate explanation other than that the thing obviously offers some effective
resistance to the flow and it (of course) decides to go where it can flow more easily. I don't think it's not
flowing through it at all, just (a lot) more slowly than around it. Whether or not the proportion going around
versus through the heatsink is anywhere near physical i cannot tell you. The only thing i can suggest is to
further refine your mesh and see whether it makes a difference.
-Juha

[Juha]

Hi,

had a look at your "demo.sif" file again, and it seems to me that you are not iterating
the coupling between the n-s, k-e & h-e equations,e.g. you have written there

Code: Select all

"Steady State Max Iterations=1"

Maybe this was just for testing?
Juha

[tibich72]

That's part of the inheritance, I don't know why it's that number of iterations. Is there a better value that I should put there?

For the boundary conditions between air and solids, I've set the wall law to true and set the boundary thickness, but did not set the velocity to zero. It did show air going between the fins and the temperatures more in line with what's expected, but it still did show a bit of velocity inside the fins. Does it make sense to set the wall law to true without setting the "normal-tangential velocity" to true and "velocity 1" to zero?

Thanks,
Tibi

[tibich72]

I've searched in the documentation for Steady State Iterations and, in the examples, it's always set to 1. Probably the sif file was created from examples in the documentation, so that's why it's one.

I've also looked at Solver 4 (the KEpsilon solver), and it has the line "Exec Solver = Never" -- it seems that this is never executed. So, the issue seems to be in the interaction between the Heat and Navier-Stokes equations.

I've increased the # of steady-stated iterations to 5, but I got the same result.
Tibi

[Juha]

Yes,
well the max iteration number is mainly there to give control to stop the simulation after a while.
Reaching "Steady State Convergence Tolerance"'s by different solvers should usually stop the
simulation before that. You definitely want to iterate the coupling, might it be between flow and
heat or the turbulence variables.

If you are not solving the turblence equations it's kind of artificial to have the "VIscosity Model" or the
"Conductivity Model" set to "K-epsilon". Same goes for the wall law as this is part of the
turbulence model. You are now just effectively using larger viscosity value computed from the
initial values of K & Epsilon... You could just as well set the viscosity to some arbitrary number?
-Juha

[Juha]

>Does it make sense to set the wall law to true without setting the "normal-tangential velocity"
>to true and "velocity 1" to zero?
The normal velocity should be set to zero by some means, there are more than
one way to do that though.
-Juha