Prev: MEX calloc fails
Next: index of npermutek
From: Simon on 2 May 2010 19:07
Hi,
I am attempting to fit a mimo grey box model to some structure vibration frequency response measurements (frd) using idgrey and pem (System Identification tbx), so far with mixed results.
I get a warning that initial state estimation is being disabled for some reason, I am not sure why, and I think this may be hindering the pem algorithm.

Warning: The value of the "InitialState" property cannot be 'Estimate', 'Model' or 'Backcast' for frequency response estimation data (IDFRD). A value of 'Zero' will be used during estimation.
> In warning at 26
In idgrey.pem at 289
In identify_TF_mimo at 185

****my code****where frd_mimo is an 4x4 FRD object
Nmodes=6;
w=[214, 297, 350, 380, 510, 550]*2*pi;
z=[0.056,0.059,0.038,0.03,0.012,0.012];
b=[-411.8543 330.8876 277.4365 -340.4028...
170.0714 540.4001 -199.2972 -586.6016...
302.3111 -0.6186 180.2823 738.1254...
-826.9749 60.3639 -278.6268 106.7860...
783.0839 62.2010 -582.9300 -190.2057...
165.0831 46.0552 548.1654 -406.6991];
c=[-71.8100 518.9095 61.3916 -391.0497...
-111.3715 731.4216 50.9352 -429.0903...
13.8883 773.7163 63.3963 520.5923...
-644.2266 195.2176 -406.1334 432.71575...
889.6835 -11.7904 -463.9117 -9.4552...
-129.2022 -22.9474 744.4212 -135.6812];
x0=[.23 -.04 0.009 0.004 -0.04 -0.04] ;

m_substr = idgrey('substr2',[z,b,c,x0],'c',[Nmodes w],0);
m = pem(frd_mimo,m_substr,'DisturbanceModel','None','InitialState','model') ;

Any ideas would be kindly appreciated.
Thanks,
Simon
From: Simon on 3 May 2010 18:57
I fixed the problem by using idss object instead of igrey to create the structured model. Also, by fixing the values of freq and damping ratio I was able to better fit my data.

I have one question though, correct me if I am wrong, but does the idgrey have the ability to define a larger class of model parameterizations of which the idss can define just a subset of? i.e. which is more general?

-Simon

"Simon " <sem5t(a)virginia.edu> wrote in message <hrl0in$1os$1(a)fred.mathworks.com>...
> Hi,
> I am attempting to fit a mimo grey box model to some structure vibration frequency response measurements (frd) using idgrey and pem (System Identification tbx), so far with mixed results.
> I get a warning that initial state estimation is being disabled for some reason, I am not sure why, and I think this may be hindering the pem algorithm.
>
> Warning: The value of the "InitialState" property cannot be 'Estimate', 'Model' or 'Backcast' for frequency response estimation data (IDFRD). A value of 'Zero' will be used during estimation.
> > In warning at 26
> In idgrey.pem at 289
> In identify_TF_mimo at 185
>
> ****my code****where frd_mimo is an 4x4 FRD object
> Nmodes=6;
> w=[214, 297, 350, 380, 510, 550]*2*pi;
> z=[0.056,0.059,0.038,0.03,0.012,0.012];
> b=[-411.8543 330.8876 277.4365 -340.4028...
> 170.0714 540.4001 -199.2972 -586.6016...
> 302.3111 -0.6186 180.2823 738.1254...
> -826.9749 60.3639 -278.6268 106.7860...
> 783.0839 62.2010 -582.9300 -190.2057...
> 165.0831 46.0552 548.1654 -406.6991];
> c=[-71.8100 518.9095 61.3916 -391.0497...
> -111.3715 731.4216 50.9352 -429.0903...
> 13.8883 773.7163 63.3963 520.5923...
> -644.2266 195.2176 -406.1334 432.71575...
> 889.6835 -11.7904 -463.9117 -9.4552...
> -129.2022 -22.9474 744.4212 -135.6812];
> x0=[.23 -.04 0.009 0.004 -0.04 -0.04] ;
>
> m_substr = idgrey('substr2',[z,b,c,x0],'c',[Nmodes w],0);
> m = pem(frd_mimo,m_substr,'DisturbanceModel','None','InitialState','model') ;
>
> Any ideas would be kindly appreciated.
> Thanks,
> Simon
 | 
Pages: 1
Prev: MEX calloc fails
Next: index of npermutek