Discussion Forum

Hi,

if the heat source follows a sine wave the long term average heat flow is zero.

Cheers
Edgar

--
Edgar J. Kaiser
emPhys Physical Technology
www.emphys.com

Yes, it's true, but I am interested in the amplitude (not the average) of the sine wave thermal signal inside the solid block, and this depends on the boundary conditions. For this reason, I was asking if the "cooling convection" is a good option when we have sine wave thermals signals o maybe I should define a "heat transfer in fluids" where I can specify the properties (static and dynamic) of the surrounding air.
Best!

Well, that's a matter of taste. I don't see that one or the other boundary condition should be incompatible with any kind of time dependent heat source.
Of course you have to take care that the time stepping resolves the time dependent source and you should keep in mind that 'thermal signals' don't reach far.

--
Edgar J. Kaiser
emPhys Physical Technology
www.emphys.com

Thanks for your ideas!
Anyway, how we can explain that the amplitude of the thermal signals is independent of the htc value of the convection cooling? Any idea? Such an amplitude depends on the area subjected to convection but not on the htc, and I am not able to explain it.

Hi

A few questions first:
you say you have an AC (sinus of frequency f0) thermal source (lets say to the left) with an amplitude "A", and an average value "Arms", I'm I correct ? Then this heat flux goes through a slab of bulk material (a wall) of density "rho", heat conduction "k" and heat capacity "Cp", hence a heat diffusivity of "alpha=k/rho/Cp [m^2/s]" (is this still the case so ?). And then the right boundary is exchanging with air by a convective exchange of the type "h[W/m^2/K]*(Text_ambient - T_wall)", and you solve all this with a time step solver, with sufficient steps per period (time stepping "Intermittent" ?).

If this is so, then the question is does the Twall change with time ? If yes you should see some difference, if no the dissipation is constant in time. And the fact that "T_wall" changes with time depends on the bulk wall length and the heat diffusivity and the AC frequency. roughly the AC thermal diffusing "wave" penetrates some 3*sqrt(alpha/f0), thereafter the temperature remains constant (it depends (with some delay) on any changes in "Arms" but not on "A".

You should check these intermediate values and look how the flux changes, i.e. change also "Arms" not just "A"

Now I might be wrong with my understanding of your model, but Heat diffusivity is sometimes counterintuitive, at least if you work often with true "waves" as in ACDC or optics.

4h later: I have just updated a few typos and the sign to follow COMSOL correctly sign convention
q_conv=h*(T_ext-T) (corrected in formuly above
--
Good luck
Ivar

Dear Ivar,
Many thanks for your comments!!
First, my model is exactly how you describe in your first paragraph, but the average value of power dissipated is zero.
About your questions, Twall does change with time, but there is no effects of htc on the amplitude of the AC thermal signals.
I will do some simulations including an average value of power.
Maybe the problem is that the average value of Twall is similar to Tambient, and then there is not a good convection in the "negative semicicle" of the AC thermal signal?
Thanks again!

Hi

when then I would suggest that you look carefully on the fluxes on the outgoing wall.
I suspect if your average is "0" that the Twall is hardly changing and that the amount of convecting power loss is rather small

--
Good luck
Ivar

Dear Ivar,
After doing many simulations (employing an average -or "dc"- value of dissipated power), I believe I have a problem with the "study" configuration.
Now I have:
-Parametric sweep (to change the freq. -f0- of the AC thermal signal)
-Step 1: Stationay
-Step 2: Time dependent
With this configuration, the effects of the dc dissipated power are not consistent with some preliminar calculations. However, I can see really see the effects of increasing f0 on the amplitude of the thermal signals.
On the other hand, if the "Time dependent" is disable (although it is disable, the transient analysis is also carried out, why?), the effects of the dc dissipated power are very similar to those expected, but there is no effects of increasing f0 on the amplitude of the AC thermal signals, which makes no sense.
I believe none of the two previous configurations is appropiate, but I do not know how to improve it. Any advice?
Best!
Ferran

Hi

well first of all 4.2 is rather old, I cannot remember if there where some issues in the physics coupling, there might be some patches proposed too.

Then another common error is the time stepping settings of your solver. COMSOL by default prepares the solver for diffusion problems, and not for AC "oscillatory" problems, so you should check that you use Solver - ... - Time Stepping - "Intermediate" with sufficiently small steps to resolve your AC component with at least 5 time steps per period.
This solver AC and time stepping detection issue has been improved in V5, but still I always change these time stepping settings for all my AC type models based on Diffusion equations.

Normally, if you disable a physics it should not solve.
I normally add in and solve with different Study nodes, one for stationary, another for time dependent (only) and often a third Study for coupled stationary (to set the "initial conditions correctly) directly followed by temporal study to get the trends

--
Good luck
Ivar