Discussion Forum

As far as I know, you cannot accomplish this in the Comsol GUI itself. However, I have done this several times using Comsol Script/Matlab. I think this may be you're only solution. From the GUI, you can save your model as an M-file and then run it in Matlab assuming you have a license for Comsol Script.

If anyone does know how to do this without using Script, I'd be very eager to hear it.

Dear Alessandro (and Rick),

it is actually possible to run an eignefrequency combined with a parameteric analysis in the COMSOL GUI.

This can be done thanks to the parametric sweep feature (added in version 3.5a)

Just go to the solve menu > parametric sweep. There you can set one or several parameters that can vary independently.
The analysis will be then performed using the solver settings defined in the solver parameters window (or using the solver settings defined in the solver sequences of the solver manager window).

A nice example can be found in the structural mechanics module's model library:

Model library > Structural mechanics module > dynamics and vibration > rotating blade

Note that the parametric sweep features works will all solver type (stationary, eigenvalue or time-dependent).

Best regards,

Remi Magnard

Dear Remi
thank a lot for the help!
I finally managed to perform the analysis within Comsol GUI.
Dear Rick, if you are interested, I can upload my file in the Model Exchange section.

Cheers.

Alessandro

Can you please upload your code for me to look at? I would be very interested in learning how you were able to accomplish this. Thank you.

I have a slightly related question and I was hoping someone could provide some insight. I am attempting to create the band structure plots for a square lattice of circular cylinders in 2d. Essentially these plots consist of listing the degenerate eigan solutions for the first brillouin zone i.e., the reciprocal lattice. My understanding is that people create these plots by sampling a continuum of wave vectors (or altering the direction of the incident plane waves) and plotting these wave vectors against freq.

My question is this: I am using Comsol to solve Helmholtz's equation for the all of the eigan frequency solutions of the reciprocal unit cell of my lattice near the location of the principle stop band. It then returns a bunch of frequencies, some with only one degenerate and others with more. This is fantastic and all but how do I extract the wave vector direction from these solutions to assign an x-axis position to the band structure plot given that the eigan value approach uses no source terms?

see attached plot for example of what I want to create

I am using periodic boundary conditions see attached sim file

fsdasdf

fsdasdf

Sorry for the mess...
I would like just to reply to John by attaching a file but "Attach File" command does not work.
In any case my reply (without attachment) is the following:

Use two modules: moving mesh (ALE), solid-stress strain.
You first have to program a sequence made of a static run with the ale_module to change
cantilever geometry and an eigenfrequency step with calculates eigenfrequencies on the
deformed geoemtry. Activate the option "solve using solver sequence".
Then you go in "solve/parametric sweep" in order to set the change of the parameter which
modifies cantilever geomety.
If you want cantilever to expands along x axis through a "disp" parameter, for instance set in the X field of the subdomain settings of the ale_module: (X+50e-6)/100e-6*disp.
In this way, you will obtain the expansion of a cantilever beam whose initial lenght is 100e-6m and whose geometry model has been obtained as block centered in (0,0,0) and having l = 100um.
I hope this help.

Alessandro

I believe there is an example available in 2d in the COMSOL RF Module that allows you to calculate the eigenfrequencies. I would like to know how to do the same thing for a 3d structure. Also, it would be interesting to learn how to vary the size of holes and plot against the eigen frequencies. Of course, the k-vector would have to be set to one particular point (normal incidence, k=0).