Menu

Discussion Forum

New Discussion
Filtrer

Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

Heat Flow - Stationary vs Time Dependent study

Posted 17 avr. 2015, 13:15 UTC−4 Heat Transfer & Phase Change, Computational Fluid Dynamics (CFD), Modeling Tools & Definitions, Parameters, Variables, & Functions, Studies & Solvers, Structural Mechanics Version 5.0 6 Replies

Jonathan Wilkins
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Good Evening,

The mph files for both are added below.

My model is shown in attachment (3).

It consists of a copper strip on a piece of GaAs which rests on a heat sink made of copper. A 20 V Voltage terminal will be on the copper strip and the heat flow will be studied through the GaAs and down into the heat sink. The stationary solution is shown in attachment (4), along with the model in attachment (5).

The next step is to look at a time-dependent study, with a 20 V Voltage Pulse, which repeats on a repetition frequency of 10kHz. The variables and parameters are set up, but the solution does not reach a thermal equilibrium in the time specified, and the result is completely wrong (attachment (2)). I am not sure how to rectify this problem, I have tried different ranges of time yet the solution still does not seem to be working.

My time-dependent model is attachment (1).

Thank you for your time.
6 Replies Last Post 21 avr. 2015, 18:16 UTC−4
Luke Gritter Certified Consultant
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago 17 avr. 2015, 18:07 UTC−4
Jonathan,

In pulsed problems like this, you need to make sure that the solver catches all the pulses. You can do this by specifying at least one output time during and between each pulse and setting the time-stepping to "Strict" on the Time-Dependent Solver sub-node. Alternatively, you could set the maximum time step to a value less than the pulse width.

--
Luke Gritter
AltaSim Technologies
Jonathan, In pulsed problems like this, you need to make sure that the solver catches all the pulses. You can do this by specifying at least one output time during and between each pulse and setting the time-stepping to "Strict" on the Time-Dependent Solver sub-node. Alternatively, you could set the maximum time step to a value less than the pulse width. -- Luke Gritter AltaSim Technologies
Jonathan Wilkins
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago 17 avr. 2015, 18:35 UTC−4
Dear Luke,

I specified the Time-Dependent node as attached, yet obtained the same result.

My pulse duration is 2us.

Thanks.
Dear Luke, I specified the Time-Dependent node as attached, yet obtained the same result. My pulse duration is 2us. Thanks.

Attachments:

Daniel Connelly
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago 19 avr. 2015, 19:10 UTC−4
I'd be tempted to run with sinusoidal voltage rather than step functions. Since heat generation is proportional to the square of the voltage you need to match <V^2> between multiple waveform options, so for example if the amplitude is V0 for the square wave, <V^2> = V0^2 / 2, so then if I have an amplitude on the sine of A, with the voltage varying from 0 to A, then the average of V^2 in this instance is 3 A^2 / 4, so A = (4/3) V0 to match the two, if I did the math right (I may well not have). An alternate option is to make V^2 vary sinusoidally from 0 to some maximum value: that might actually be better.

The issue would be that interpolating voltage across the step is going to give incorrect values so depending on how the simulator chooses transient step size you could get substantial errors during the transitions. However if you use a function with a continuous first derivative that would be more realistic. In any case there's an RC time constant associated with the system so you can't really change voltage instantaneously anyway.


I'd be tempted to run with sinusoidal voltage rather than step functions. Since heat generation is proportional to the square of the voltage you need to match between multiple waveform options, so for example if the amplitude is V0 for the square wave, = V0^2 / 2, so then if I have an amplitude on the sine of A, with the voltage varying from 0 to A, then the average of V^2 in this instance is 3 A^2 / 4, so A = (4/3) V0 to match the two, if I did the math right (I may well not have). An alternate option is to make V^2 vary sinusoidally from 0 to some maximum value: that might actually be better. The issue would be that interpolating voltage across the step is going to give incorrect values so depending on how the simulator chooses transient step size you could get substantial errors during the transitions. However if you use a function with a continuous first derivative that would be more realistic. In any case there's an RC time constant associated with the system so you can't really change voltage instantaneously anyway.
Daniel Connelly
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago 19 avr. 2015, 22:48 UTC−4
Actually, now that I think about it more, I like the idea of making V^2 sinusoidal about Vmax^2/2. Then the peak value of V will be the same in both cases, as the average of V^2 in both cases will be Vmax^2 / 2.
Actually, now that I think about it more, I like the idea of making V^2 sinusoidal about Vmax^2/2. Then the peak value of V will be the same in both cases, as the average of V^2 in both cases will be Vmax^2 / 2.
Walter Frei COMSOL Employee

Please login with a confirmed email address before reporting spam

Posted: 10 years ago 20 avr. 2015, 10:18 UTC−4
Dear Jonathan,

For the kind of periodic loading that you are describing in this problem, you will want to use the Events interface (rather than strict timestepping or altering the maximum timestep)

This will ensure that the solver captures the on/off events appropriately. For more details about how and why to do this, please see:
www.comsol.com/blogs/modeling-a-periodic-heat-load/
Dear Jonathan, For the kind of periodic loading that you are describing in this problem, you will want to use the Events interface (rather than strict timestepping or altering the maximum timestep) This will ensure that the solver captures the on/off events appropriately. For more details about how and why to do this, please see: http://www.comsol.com/blogs/modeling-a-periodic-heat-load/
Luke Gritter Certified Consultant
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 10 years ago 21 avr. 2015, 18:16 UTC−4

Dear Luke,

I specified the Time-Dependent node as attached, yet obtained the same result.

My pulse duration is 2us.

Thanks.



Jonathan,

I tried running the model you attached with the 0.1 um maximum step (as shown in your attached image), and the results looked reasonable to me. The model tracks the pulse shape well and generates heat accordingly. The amount of heating is very small due to the low duty cycle, short run time, and high heat transfer coefficient.

As Walter mentioned, using an explicit event will give you a much shorter runtime than limiting the maximum time step for this model.

--
Luke Gritter
AltaSim Technologies
[QUOTE] Dear Luke, I specified the Time-Dependent node as attached, yet obtained the same result. My pulse duration is 2us. Thanks. [/QUOTE] Jonathan, I tried running the model you attached with the 0.1 um maximum step (as shown in your attached image), and the results looked reasonable to me. The model tracks the pulse shape well and generates heat accordingly. The amount of heating is very small due to the low duty cycle, short run time, and high heat transfer coefficient. As Walter mentioned, using an explicit event will give you a much shorter runtime than limiting the maximum time step for this model. -- Luke Gritter AltaSim Technologies

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.

Suggested Content