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.

Time Dependent simulations with Piezoelectric devices

Thibaut Meurisse
Hi all,

I have a problem I am sure the solution is obvious...but if it was already resolved somewhere on the website, maybe I just did not understand...

I simulate a piezoelectric disc just put on a roller, everything else being free boundaries, and I apply a sine voltage with frequency f0 and amplitude V0. I then launch the simulation from time 0, with time steps 1/20/f0 (to have 20 points every wavelength) and up to 20/f0 (20 wavelengths). So range(0,1/20/f0,20/f0).

First, I obtained strange results with time steps that were growing between each step (they were not constant). I probed the voltage on the disc, and it did not respect the frequency f0 (it was almost noise) and amplitude V0 (up to 3 times V0).
I then modified the Time Dependent solver, using for the probes "Times stored in output". Now the time steps are good, but the result is still anything. Voltage still goes up to 3 times V0, and frequency is not respected. In fact it is almost respected after some time, call it t0 (like half of the time steps). But if I try to begin at t0, I obtain the same result than before, just starting at t0.

Did I do something wrong ?

13 Replies Last Post 25 Oct 2016 03:40 GMT-04:00
Posted: 1 year ago 21 Oct 2016 07:27 GMT-04:00
Hi Thibaut,

I'm also working on piezoelectric devices simulation. Can you tell me how to apple a sine voltage with frequency f0 and voltage V0 ?

Thanks for help!
Hi Thibaut, I'm also working on piezoelectric devices simulation. Can you tell me how to apple a sine voltage with frequency f0 and voltage V0 ? Thanks for help!

Thibaut Meurisse
Posted: 1 year ago 21 Oct 2016 08:20 GMT-04:00
Hi Allen,

If you are working in Time Dependent study, to do that you should :
- Create a Function : click on "f(x) Functions", then Waveform, give a name to the function and a label (usually I give the same name to both, Exc), in the Parameters section select the type of waveform you want (for example : Sine), then the angular frequency (2*pi*f0), the phase if needed and the amplitude (V0). Now in your global definition, your function appears (in my case, it is written Exc(Exc) ). You may click on "Plot" to see the waveform you created.
- Electrostatic (es) : add a ground boundary, then a Electric Potential boundary (opposite to that you put the ground boundary). In the Electric Potential, go to the Electric Potential section and add after V0 the name of your function. For me, Exc(t). "(t)" has to be added to tell Comsol that the excitation depends on the time.
- Time Dependent : give the time vector you want to apply while computing the model.

Note that in my case, it does not work. I think I forget something, somewhere. And this is why I put this post here. Tell me if it works for you !
Hope it helps !
Hi Allen, If you are working in Time Dependent study, to do that you should : - Create a Function : click on "f(x) Functions", then Waveform, give a name to the function and a label (usually I give the same name to both, Exc), in the Parameters section select the type of waveform you want (for example : Sine), then the angular frequency (2*pi*f0), the phase if needed and the amplitude (V0). Now in your global definition, your function appears (in my case, it is written Exc(Exc) ). You may click on "Plot" to see the waveform you created. - Electrostatic (es) : add a ground boundary, then a Electric Potential boundary (opposite to that you put the ground boundary). In the Electric Potential, go to the Electric Potential section and add after V0 the name of your function. For me, Exc(t). "(t)" has to be added to tell Comsol that the excitation depends on the time. - Time Dependent : give the time vector you want to apply while computing the model. Note that in my case, it does not work. I think I forget something, somewhere. And this is why I put this post here. Tell me if it works for you ! Hope it helps !

Posted: 1 year ago 23 Oct 2016 21:59 GMT-04:00
Hi Thibaut,

I 'm working on a project including piezoelectric transducer simulation.I have tried the way you suggested when applying a sine excitation with the frequency of 4.6MHz and the duration of 1 second. But what I got when I clicked the 'Plot' button was totally unexpected.

So how can I get a correct sine wave with the frequency of 4.6MHz in COMSOL?

Thanks for your advice!
Best regards!
Hi Thibaut, I 'm working on a project including piezoelectric transducer simulation.I have tried the way you suggested when applying a sine excitation with the frequency of 4.6MHz and the duration of 1 second. But what I got when I clicked the 'Plot' button was totally unexpected. So how can I get a correct sine wave with the frequency of 4.6MHz in COMSOL? Thanks for your advice! Best regards!


Thibaut Meurisse
Posted: 1 year ago 24 Oct 2016 03:43 GMT-04:00
Hi Nick,

I tried what you did, this is the first time I use the Analytic function (usually I use Waveform).
I think the problem comes from the used sampling frequency. If you try with different frequencies, it works at 1Hz, 10Hz. At 100Hz it starts to do anything, and at 1kHz it goes back to 1Hz.

Unfortunately, I do not know how to impose the sampling frequency. What you can do is setting the Upper limit of Argument t (in Plot Parameters) to a value small enough so that the result is correct.

Or you may use "Waveform" instead of "Analytic" when selecting the kind of Function (which is what I describe in my previous message).

Hope it helps.
Hi Nick, I tried what you did, this is the first time I use the Analytic function (usually I use Waveform). I think the problem comes from the used sampling frequency. If you try with different frequencies, it works at 1Hz, 10Hz. At 100Hz it starts to do anything, and at 1kHz it goes back to 1Hz. Unfortunately, I do not know how to impose the sampling frequency. What you can do is setting the Upper limit of Argument t (in Plot Parameters) to a value small enough so that the result is correct. Or you may use "Waveform" instead of "Analytic" when selecting the kind of Function (which is what I describe in my previous message). Hope it helps.

Posted: 1 year ago 24 Oct 2016 04:35 GMT-04:00
Hi

I am not quite sure of what I am saying here since I don't know how you are simulating your structure.
But could it be that one roller is not enough in constraining your geometry?

maybe it could help if you add a "spring foundation" with a very low spring constant in the direction that your geometry can role .

Dara
Hi I am not quite sure of what I am saying here since I don't know how you are simulating your structure. But could it be that one roller is not enough in constraining your geometry? maybe it could help if you add a "spring foundation" with a very low spring constant in the direction that your geometry can role . Dara

Thibaut Meurisse
Posted: 1 year ago 24 Oct 2016 04:57 GMT-04:00
Hi Dara,

Thank you for your suggestion. Also, even if I am not sure how to use the "spring foundation", with different values I obtain the same kind of results as with a roller, or with nothing in fact.

The first problem I would like to resolve is related to the voltage applied to the piezo. I apply a sine voltage (see "Function_V" attached), and when probing the voltage during the simulation, what I obtain is different (see "Probe_V" attached). Note that in the probe file, there are many points, even if it looks like there is only a few (when they draw straight lines, they are drawn with many points). If I compute it over a longer time, it is even worth (see "Probe_V2"). Here, figure have been obtained with a "spring foundation", but it stays the same with a roller or nothing. I just wonder if solving this problem could solve all the others.
Hi Dara, Thank you for your suggestion. Also, even if I am not sure how to use the "spring foundation", with different values I obtain the same kind of results as with a roller, or with nothing in fact. The first problem I would like to resolve is related to the voltage applied to the piezo. I apply a sine voltage (see "Function_V" attached), and when probing the voltage during the simulation, what I obtain is different (see "Probe_V" attached). Note that in the probe file, there are many points, even if it looks like there is only a few (when they draw straight lines, they are drawn with many points). If I compute it over a longer time, it is even worth (see "Probe_V2"). Here, figure have been obtained with a "spring foundation", but it stays the same with a roller or nothing. I just wonder if solving this problem could solve all the others.


Posted: 1 year ago 24 Oct 2016 05:15 GMT-04:00
Hi,

I think maybe you have to reduce your time-step increments because it seems they are too big and that creates the jumps in your Probe_V results you showed.

If that is the case you can in addition change the time stepping in the solver to "strict" or "intermediate"

hope this helps

Dara




Hi, I think maybe you have to reduce your time-step increments because it seems they are too big and that creates the jumps in your Probe_V results you showed. If that is the case you can in addition change the time stepping in the solver to "strict" or "intermediate" hope this helps Dara


Thibaut Meurisse
Posted: 1 year ago 24 Oct 2016 05:35 GMT-04:00
Hi Dara,

Thank you for your help, now Comsol applies the good voltage !

But I am afraid that it does not resolve all my problems, as the deformation of the piezo is not perfectly related to the voltage (see attached, using a probe to determine the displacement of a part of the disc). I will try to find if this is normal or not, and add something here (or ask some more questions).

Thibaut
Hi Dara, Thank you for your help, now Comsol applies the good voltage ! But I am afraid that it does not resolve all my problems, as the deformation of the piezo is not perfectly related to the voltage (see attached, using a probe to determine the displacement of a part of the disc). I will try to find if this is normal or not, and add something here (or ask some more questions). Thibaut


Posted: 1 year ago 24 Oct 2016 05:54 GMT-04:00
Hi

happy that I could help

the results you get could really depend on the physics and how you set up the problem.

if ever, you could post the model here ( or an equivalent problem in case of secrecy :-) ), so people with more knowledge about piezos could help

Dara

Hi happy that I could help the results you get could really depend on the physics and how you set up the problem. if ever, you could post the model here ( or an equivalent problem in case of secrecy :-) ), so people with more knowledge about piezos could help Dara

Thibaut Meurisse
Posted: 1 year ago 24 Oct 2016 06:59 GMT-04:00
Hi Dara,

Attached is a simplified version of the problem (in 2D axisymmetric, which weight is lower), with just the basics. I obtain the same problem in 3D representation.

And of course, if anyone else could have any advice, it would be welcomed !

Thibaut
Hi Dara, Attached is a simplified version of the problem (in 2D axisymmetric, which weight is lower), with just the basics. I obtain the same problem in 3D representation. And of course, if anyone else could have any advice, it would be welcomed ! Thibaut


Thibaut Meurisse
Posted: 1 year ago 24 Oct 2016 10:31 GMT-04:00
Also, I wonder if that result could be due to oversampling ?
Also, I wonder if that result could be due to oversampling ?

Posted: 1 year ago 24 Oct 2016 16:09 GMT-04:00
Hi

First of all the structural analysis is not well defined without attaching the part with something to keep a "gauge" Z position. As you seem to only look for a radial contribution, it sort of works anyhow.

The waveform excitation is required for the voltage to get the sinus(2*pi*f0*t) shape.

I recommend always to do an eigenfrequency analysis to check the modes and you will see a radial mode at 193kHz, by plotting the mode participation factors and normalize the square of the MPF with the total mass you see which modes are the most "massive" to sort them by importance and direction. However in 2D axi you are already filtering many modes, be aware.

Then your time stepping seems for me quite reasonable, you get interactions with the higher modes because you apply a sudden acceleration and you excite many modes, as there is no damping in your model.
if you do a Fourier transform of your signal you will probably see the 193 kHz coming out, you might want an even shorter time stepping for that.
I would suggest to run an frequency domain sweep rather, start at your frequency f0 and a terminal voltage of 1 V and you will get a reasonable amplitude.

I always suggest to use intermediate time stepping mode when applying periodic load signal, as by default COMSOL is setting up the time stepping for non oscillating "diffusion" models, and might easily skip any periodic modulation. Anyhow periodic modulation implies normally a frequency sweep approach

Do not forget that COMSOL solves the true and full PDE of the physics such that all acoustic waves are also in there and when you hit a part with such a sinus step you excite many modes.

--
Good luck
Ivar
Hi First of all the structural analysis is not well defined without attaching the part with something to keep a "gauge" Z position. As you seem to only look for a radial contribution, it sort of works anyhow. The waveform excitation is required for the voltage to get the sinus(2*pi*f0*t) shape. I recommend always to do an eigenfrequency analysis to check the modes and you will see a radial mode at 193kHz, by plotting the mode participation factors and normalize the square of the MPF with the total mass you see which modes are the most "massive" to sort them by importance and direction. However in 2D axi you are already filtering many modes, be aware. Then your time stepping seems for me quite reasonable, you get interactions with the higher modes because you apply a sudden acceleration and you excite many modes, as there is no damping in your model. if you do a Fourier transform of your signal you will probably see the 193 kHz coming out, you might want an even shorter time stepping for that. I would suggest to run an frequency domain sweep rather, start at your frequency f0 and a terminal voltage of 1 V and you will get a reasonable amplitude. I always suggest to use intermediate time stepping mode when applying periodic load signal, as by default COMSOL is setting up the time stepping for non oscillating "diffusion" models, and might easily skip any periodic modulation. Anyhow periodic modulation implies normally a frequency sweep approach Do not forget that COMSOL solves the true and full PDE of the physics such that all acoustic waves are also in there and when you hit a part with such a sinus step you excite many modes. -- Good luck Ivar

Thibaut Meurisse
Posted: 1 year ago 25 Oct 2016 03:40 GMT-04:00
Hi Ivar,

Thank you very much for your explanations and recommendations, they explain almost everything.

Thibaut
Hi Ivar, Thank you very much for your explanations and recommendations, they explain almost everything. Thibaut

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.