Hi

Yes indeed amplitudes in frequency analysis of COMSOL should be seen as relative items, unitless

In anycase, for modal analysis, amplitudes depends on the quality factor Q that is directly coupled to the loss or damping values you use, therefore either you know the losses and you get correct absolute values (relative they will anyhow be OK), or you measure your damping factor via the resonance heights versus half widths values and you run a parametric sweep (or an optimisation) of your damping factor on your model to identify how to match both.

What you can find out though, with an eigenfrequency calculation, is the relative participation factors of the different modes, and how much relative energy there are in the different ones.
This is not directly derived by COMSOL, as you have it in Nastran, Ansys, COSMOS and most others FEM tools, because COMSOL uses a different eigenfrequency normalisation. But you can, if your model is not to heavy, export the matrices into matlab and normalise them ther.

Take a look at :

http://www.comsol.com/community/forums/general/thread/2650/

for the litterature you have good descriptions in "Actrive Control of Structure", A. Preumont and K Seto, Wiley, 2008, chapter 1.

By the way Prof. A. Preumont (Springer, and Kluever) has some very interesting books with good examples to run in COMSOL on Piezo structures and control thereof

Have a nice reading and
Good luck
Ivar

Read more: Solution of an eigenfrequency problem - COMSOL

http://www.comsol.asia/community/forums/general/thread/3811/#ixzz11A52aKog