Discussion Forum

Hi

you shuld check the Forum (search) there are many questions around this. The main reasn is that an eigenfrequency is NORMALISED to an arbitrary value, COMSOL has selelcted one norm, other tools have other norms. In 4.1 you can now also select different norms (see the eigenvalue solver sub-node) including selecting the mass partcipation value, i.e. each eignfrequency is scaled w.r.t. its mass participation w.r.t. the three depedent varibles u,v,w.

An eigenfrequency analysis has no normalised energy input so normally all modes are shown with large displaceemnts even if this does not make really any sens i.e. u=6m for a 10um large part ;)

Then you want a Q factor, but Q factor is the dissipation, and by default COMSOL porposed "0" dissipation. An eigenfrequency analysis will not give you anything else than 1) the dissipation factor you enter 2) + some artificial numerical dissipation that migh show up depending on the solver and model used, to avoid NAN's all over

To perform stressed eigenfrequencies you must fist build up the stress with a stationary solver under some BC load, then save the solution and run an eigenfrequency solving (somewhat similar to the default "buckling" analysis of COMSOL) as eigenfrequency analysis is a linear approach, you mut define the starting point onto which the solver will do its linearisation. In 3.5 you needed matlab to link these two steps, in V4 this is done in the solver sequence (note that the early v4.0 version has several weakness and not yet implemeneted limitations, you should upgrade to 4.1)

As I said there is even far more on the forum in the diffeent threads
--
Good luck
Ivar

Hi Ivar,

Thank you so much for your reply. However I still get confused with these issues:

1. "you shuld check the Forum (search) there are many questions around this. The main reasn is that an eigenfrequency is NORMALISED to an arbitrary value, COMSOL has selelcted one norm, other tools have other norms. In 4.1 you can now also select different norms (see the eigenvalue solver sub-node) including selecting the mass partcipation value, i.e. each eignfrequency is scaled w.r.t. its mass participation w.r.t. the three depedent varibles u,v,w."

Yes, I have read some other threads about this issue, but there were not clear enough How big is the normalization?? Lets say I have 7m of displacement. May I assume that the real displacement of my cantilever is times of 10e-7 or 10e-6..So that my real value of displacement is around 0.7 um or 7 um (I saw it on another thread/forum)..I thought it is more realistic..:)..But I am not quite sure...It still does not make sense for me to see 6m displacement for 10um geometry. Since I want to know what is the smallest dimension of cantilever that I can create for my purpose, but without sacrificing the detectable deflection/displacement. If I just make 'assumption', then my results are BIG QUESTIONS..:). Please see the attached file..
And I was thinking that COMSOL cannot say the real value of displacement by eigenfreq analysis, since I have read one of paper from COMSOL conference about modelling Cantilever. And there are only the pictures of deformations of the cantilever in the first and second mode of resonant freq...And of course the value of 5 different modes of frequencies..That research group was comparing the theoretical value of resonant frequency with FEM and measurement. However, I did not see the legends/values which tell the reader how big is the deflection/displacement.
So, if I would like to get the real value of displacement..Do u have any suggestion what should I do?Should I use other solver? However, I am much more interested when my cantilever is working on high mode resonant frequency..

2. "Then you want a Q factor, but Q factor is the dissipation, and by default COMSOL porposed "0" dissipation. An eigenfrequency analysis will not give you anything else than 1) the dissipation factor you enter 2) + some artificial numerical dissipation that migh show up depending on the solver and model used, to avoid NAN's all over"

SO, I am not able to see find the Q factor directly from comsol. How to enter the dissipation factor? I was using sequence solver(2 solvers in 'study' tree)..The first one is eigenfreq solver, then the second is stationary..However, the results are separated. I do still see the same values of eigenfrequencies compared to "only using eigenfreq solver"..
Those results make me thinking that for eigenfreq analysis, I am only allowed to use 1 solver...Is it possible in COMSOL to see the common 'bell-shape' resonance frequency curve with frequency values in X-axis and Amplitude value in Y-axis??If so, how could we do that?

3. I am also curious about 'is COMSOL able to simulate the effect of humidity and temperature for a cantilever?". But what I need to analyze is resonance frequency. The idea is similar with "busbar_box", but when I use stationary solver, I do not get any frequency result which can be analysed..I just saw the displacement, von misses Stress, and so on..Since it will be great if those things can be simulated in comsol..Could you suggest me on this?

Thanks so much...

Hi

I believe you should first make clear for yourself the physics involved here, which parameters are involved, and how things are related.

From the eigenfrequency analysis you get ONLY frequencies out (as absolute values). Eigenfrequencies are properties of the material and the geometry only. The amplitude of the oscillation is a dynamic issue related to energy conservation and momentum transfer, which is fully different. To get amplitudes out you must know the forces in and the previous time variations involved, or with some PSD estimate of a input energy distribution versus frequency

The "mass participation factor" normalisation of the eigenfrequencies, allows you to tell the ratio of the energy that the different modes would have, if excited by a flat-energy random excitation, but only relative to each others, and for the three different directions. This i generaly of interest for engineers, as we then can identify the modes having the highest potential impact.

Quality factors of modes are related to the material damping, hysteresis effects, air damping etc. Comsol can include these, if you add them to the parameters in the implemented physics equations. However the damping of material are the parameters that are generally the most difficult to find, not to say that bolted, or soldered items have very varying damping factors, that can normally only be estimated by measurements, and are rather case to case specific.

The effect of thermal expansion can be included with COMSOL, but you need to tell and analys where the heat sources are, where and how the energy leaves, to define correctly your BC.

Humidity is another story, but there you need to define how the humidity affects your canteliever, which physics, which other structural material properties are affected and how. I believe this requires "home brewed" equations, perfectly possible but requires a) a perfect understanding of COMSOL b) a good understanding of the physics involved,c) a global understanding of the system and the expected results

Last, find out with your local COMSOL rep. how to get hands on v4.1, there are too many things not ready in the early 4.0

--
Good luck
Ivar