Modal identification using Frecuency domain descomposition
Modal identification using Frecuency domain descomposition
(OP)
Hello, I´m spanish, my english is very bad...
I try to make a program that use the method "frequency domain descomposition" (FDD) (Brincker et al. (2000)).
I try to calculate the modal parameters of a cantilever beam. This method take n-signal and make the spectral density matrix. Then compute de singular value descomposition for each frequency and take for a new approximation, for the frequency response of my mechanical system, the maximum singular value. The eigenvector at the natural frequency are the shape mode, and the damping ratio is estimate from the inverse fft of the svd around the peak of the natural frequency.
I have a problem to calculate the PSD of each signal. I´m usin the command pwelch of MATLAB, that use the welch's method to estimate de PSD. The parameters of this command are the length of the window to average, the overlap of these windows and the number of samples that I anlyze. I see that the spectrum's amplitude and shape change with the value of the length window.
Someone could help me.
Thanks!
I try to make a program that use the method "frequency domain descomposition" (FDD) (Brincker et al. (2000)).
I try to calculate the modal parameters of a cantilever beam. This method take n-signal and make the spectral density matrix. Then compute de singular value descomposition for each frequency and take for a new approximation, for the frequency response of my mechanical system, the maximum singular value. The eigenvector at the natural frequency are the shape mode, and the damping ratio is estimate from the inverse fft of the svd around the peak of the natural frequency.
I have a problem to calculate the PSD of each signal. I´m usin the command pwelch of MATLAB, that use the welch's method to estimate de PSD. The parameters of this command are the length of the window to average, the overlap of these windows and the number of samples that I anlyze. I see that the spectrum's amplitude and shape change with the value of the length window.
Someone could help me.
Thanks!





RE: Modal identification using Frecuency domain descomposition
RE: Modal identification using Frecuency domain descomposition
Your problem is a signal processing one, and is not really a result of the FDD technique.
Making the window longer will change your frequency resolution. The better your frequency resolution, the more unreliable the measurement is.
Overlap, in excess of 50%, is a nice way to get smoother curves, but is not really increasing the information available to the system.
Increasing the total length of the data analysed is the ONLY way to get significantly more information into the model over a given frequency range.
Cheers
Greg Locock
Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.
RE: Modal identification using Frecuency domain descomposition
htt
It certainly fitted a few test cases I gave it at any rate. And it's free and written in Matlab. I converted it to Scilab tho' as thats free as well, unlike matlab
RE: Modal identification using Frecuency domain descomposition
Cheers
Greg Locock
Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.