Modeling piezoelectric transmitters with acoustic and piezo extensions

  • Last Post 25 June 2018
RaphaelR posted this 22 June 2018



I have been modeling piezoelectric acoustic transducers to determine the frequency response spectrum in water.


The specific transducer types I am modeling are monopole ring and dipole bender bar transducers. They are to be operated in a pulse mode: a single sine differential voltage pulse is applied. The frequencies I am modeling are between 500 Hz and 20 kHz.


I have been using a harmonic analysis in workbench 19.1 with both the acoustic and piezoelectric extensions. Materials are modeled as isotropic as a simplification.


My models (for both ring and bender transducers) all suffer from the same issues:


(1) The model shows a much higher response than the experiment at resonance

(2) The model shows a lower response than the experiment at both higher and lower frequencies than resonance

(3) The location of the peaks in the model vs. experiment is substantially different

(4) Typically the difference in response for the model between two designs is much greater than reality


I have attached the response of two different ring transducers for the model vs. experiment to illustrate the problem.


What's wrong and what could be used to improve the fidelity of the model (and be able to use it to predict the behavior of the actual part)?


Much higher mesh densities?

Adding damping to the transducer? If so what is the best method?

Modeling via fluent rather the acoustic extension?

Doing a transient analysis rather than a harmonic analysis to more closely match the experiment?

Or am I expecting too much from this type of model?


Any help or insights would be much appreciated as I have very little experience with Ansys and acoustic/piezoelectric modeling and don't know where to start.





Attached Files

Order By: Standard | Newest | Votes
peteroznewman posted this 24 June 2018

I did a "back of the envelope" calculation to scope out the maximum element size.

Your maximum frequency, f, is 20 kHz. The speed of sound, c, in water is about 1500 m/s.

Let's calculate the wavelength of a 20 kHz sound in water:  lambda = c/f  = 1500/20000 = 0.075 m or 75 mm.

At least 6 quadratic elements are needed along one wavelength to adequately represent the wave. That means the maximum element size in the model would be 12.5 mm for 20 kHz.  I don't know how big your domain is, but that maximum element size seems reasonable.

If the response is too high at resonance, then you must add damping. That can be applied in the Solution branch, but that applies to all materials so it may be more appropriate to apply damping to each material in Engineering Data since the contribution to damping from the water could be very different from the contribution to damping from the piezo material.

Best regards,


RaphaelR posted this 25 June 2018

Hi Peter,


Thank you for looking into this. My max element size is around 18 mm, this is probably leading to some error, but we should see more error at higher frequencies than lower frequencies which is not the case. I will be reducing the element size, but I agree that the biggest issue is likely inadequate damping.

What is strange however is that the response is too low at both low and high frequencies. Wouldn't adding damping resolve the issue at resonance but result in an even bigger error at low and high frequencies?

Do you think most of the damping is coming from the piezo material? What would be the best damping model that would allow me to approach experimental data?

Thanks again,


peteroznewman posted this 25 June 2018

Hi Raphael,

What does "pulse mode" mean?  I read that it might not be a continuous sine wave that would match what a harmonic analysis is assuming. Therefore, if you apply the real pulse train to the piezo material in a Transient Dynamics analysis, that might give you a closer agreement to experimental data.

How is the experimental data measured?

I expect the water will provide much more damping than the piezo material itself.

If you can share your Project Archive, that would help me understand your model.

Best regards,