Discussion Forum

Hi

you havent (or I havent understood) told us how you heat ?, by induction, radiation ? Are you sure your system is set up with a source therm that enters and exchanges into alldomains ?

--
Good luck
Ivar

Hi

I use a heat source. The domain is a dice of silicon where a small part is chosen as a heat source. The heat source effectively exchanges into alldomain. The same configuration works in with the stationary study.

I definitely don't know why the couple heat source/frequency domain study doesn't work.

Cédric

Hi

normally on uses the average power as output from the frequency scan to load the thermal part, but it all depends on how you couple your physics.

One you change the solver type, you also change adapt some of the variables COMSOL is defining internally

--
Good luck
Ivar

Hi

I don't understand what you mean with "but it all depends on how you couple your physics". In my case I have only one physics (which is heat transfer) and the heat transfer in solid concerns the all domains.

I looked the equations used in the physics and they are the same for the stationary and the frequency domain study. I don't know if some other parameters may have changed.

Cédric

Hi

it's probably that I have not fully understood your model description.
For me frequency domain solvers are typically used for ACDC, vibrations etc, generally not for HT so I understood you were looking at the thermal coupling of some frequency induced phenomena and a HT related thereto

If you are looking onto the T change during one periode, you need to work i transient mode. Frequency analysis works on the amplitude of the signal so its assumed that you have a regular sinus (posible with some phase delay)
as result too. but if your HT time constant is large compared to the frequnecy, you have only an average effect and you may couple average power, and not instantenous power. All depends on your model

--
Good luck
Ivar

Hi

My objective is found the transfer function (amplitude and phase) between the heat source and the increment of temperature. I do the hypothesis of steady-state, I am not looking onto the change during one period, this is why I need frequency domain study instead of stationary study.

Cédric

Hi

then I believe your issue is with the coupling of energy to your HT case, check the equations, and the values. One can often just run for initial conditions to fill the matrices, and then take a quick look at the different values in the postprocessing page

--
Good luck
Ivar

Hi

The equations used are the equations for frequency domain.

Although I have seen something strange in the solver. In the study part I have only one study: frequency domain. When I launch the simulation, the solver configuration has only on option named stationary solver 1 with options to parametrize the solver but anything about frequency domain. Is Comsol doing the simulation or just the initial condition?

The postprocessing page shows me the solution for the only frequency I do the simulation. The all dice is at the ambient temperature.

Hi Cédric,

I would like to find out if you finally were able so overcome this issue. I'm trying myself a very simple similar problem and all I get is a null temperature distribution for each frequency step I'm trying to simulate.

Cheers,
Diego

Hi

Not all physics are defined to use all solvers.

Normally one would not use a harmonic solver on diffusion models I would believe.

When mixing such cases one need to run in segregated way, and do some homework (hand calculations) to estimate the different time frames of the different physics.

Often one can separate the phenomena via their response time, and run one "master physics" with one solver i.e. transient, and have the others solving in stationary only. Or other combinations

--
Good luck
Ivar

Dear Ivar,

It would be nice if it were possible to use the frequency domain or frequency-domain,perturbation study steps in heat transfer problems. Like previous authors in this discussion mentioned, adding a frequency-domain study step to a problem does not seem to work, since if I look at the equation for the heat source (Show equation assuming: Frequency Domain) the term containing the heat capacitance is absent (rho*Cp*dT/dt). If this term is absent in the heat equation, time or frequency dependent effects can clearly not be captured by a frequency analysis.

If I add a time-dependent study step, the equation does show the heat capacitance term rho*Cp*dT/dt. Therefore it seems that to get a thermal frequency response analysis the best way is to perform a time-dependent analysis and convert it by a Fourier transform to the frequency domain. Or is there another way to directly get the thermal frequency response?

Best regards,

Peter Steeneken

Hello

first of all, to get clear answers of what is or not, and especially what is to come for COMSOL in the future, then you must address that directly to "support" or to your closest rep, here on the Forum we are all users (me including ;)

Today what I can say is that when COMSOL does not propose a solver for a given physics alone, that means that it related equations are not implemented (yet). Sometimes you can get around, either by changing the physics yourself, or by considering the effects at different time scales, and use i.e a time dependent solver for a slow process (accepting a time stepping solver), and link the result to a stationary analysis for a "fast process" (assuming the latter does not have the time solver setting implemented), and then loop around

Then I have heard of people using modal development of HT cases, although for me it sounds somewhat strange, Diffusion PDEs do not "oscillate" by their physical laws for t>0. Certainly you might drive it by a oscillating BC.
But I have never had time to look deeper into this issue, so why can't there be something interesting there too ? probably with some limiting hypothesis that MUST be respected.

But if you have an oscillating BC you can run a temporal analysis, but one always need to wait long enough for the transients to settle

--
Good luck
Ivar

Hello, Ivar Kjelberg
I have a question about freauency domain solver.
what dose frequencies mean?
If frequencies is 1000Hz, 1000Hz is frequency of current source?
I want to apply sine wave current source to the cross sectional area of coil in axial symmetric 2-D simulation
But, I can't search method.
I don't know how to apply waveform to current source.

Hi

take a look at the equations tab in v4 (not the equation sub node) for your physics. You will see that the frequency domain solver does a harmonic development of the excitation signals (right hand side of the equation). You define the amplitude, possibly the phase and you sweep the frequency domain (equivalent to a sinus sweep spectrum analyser exciting your models)

Tkae a new look at the doc, too

--
Good luck
Ivar