07-20-2016 02:05 AM - edited 07-20-2016 02:08 AM
I didn't propose to use a bandpass 🙂
And you can't expect to get reasonable identification results for a frequency region where you didn't provide signals with a at least some information 😉
If you only provide information in a restricted frequency band you don't have/provide prior system knowledge, you can't expect to get reasonable results outside the frequency band provided,
One common way is to restrict the system order to the lowest possible order that still can explain/fit your results. And if you know that your system response is about zero at higher frequencies, restrict your system models to types that behave like that.
07-21-2016 09:26 AM
Hi Henrik:
I carefully read your last message and change the frequency range of excitation to 0 to 500 Hz, and also adjust the filter's parameters. The result becomes better, but still not ideal. And I think both the frequemcy models of input signals and the nonparametric identification are right now, as the pictures below show:
figure.1: frequency response of the system input to the estimation vi
figure.2:magenitude response of the nonparametric model identificated by nonparametric identification vi
figure.3:phase response of the nonparametric model identificated by nonparametric identification vi
We can see after the preprocessing of the input signals, the nonparametric identification vi can precisely estimate the nonparametric model
But the problem is, when I put the same input signals to the parametric identification vi, it cannot figure out reasonable result no matter what I do, it is ridiculous that with such clear input signals of which the first three modals cannot be more obvious ,the parametric identification vi just cannot give the right transfer function, only to give these anwsers like below:
figure.4:magnitude response of the parametric model identificated by parametric identification vi
I have no idea now, and just turn to matlab which can easily give me the right result using the system identification tools.