Hi, when we have a curve or a signal (as a curve of second order system), how can find a model of this curve, there are functions in Matlab do that??

My problem: I have a curve, and i know that this curve can be modeled as a second order equation: H(p)=G*w^2/(p2+2*w*zeta*p+w^2). the idea: how i can optimize the three parameters(g, w, zeta) to obtain a model of my curve,

thank you in advence.

[Patrik Ek]

How about p and p2, are they known? Also w is that a constant? This looks like a signal and in general this is a function of w. p can of course be included in zeta, but p2...

[khaled SAFI]

sorry, p^2 so: H(p)=G*w^2/(p^2+2*w*zeta*p). p is the time variable: 1, 2, ..., n and w is the frequency of this signal oulr curve.. so we have 3 parameters to estimate : G, w and zeta

[khaled SAFI]

H(p)=G*w^2/(p^2+2*w*zeta*p+w^2)

[Patrik Ek]

So is p the continous shift operator and w is a constant? the one called w0 in many cases?

[khaled SAFI]

yes, w, G and zeta are the constants that we should estimate them, to find finally a model confirm with my curve(vector data).

[khaled SAFI]

>I assume that you are working with automatic control or something similar. yes my data are generated as the mean square diplacement, and the idea, i want to find a model for this data, and in general, when we plot it, its form like the second order plot, that why i want to model it as a second order system that have this equation:

[khaled SAFI]

H(p)=G*w^2/(p^2+2*w*zeta*p+w^2) I used fit(xdata,ydata,rat02), but it doesnt make a good estamations for G, zeta ans W. >Do you have frequency dependent data or time dependent data. I have a time dependent data. as: xdata=0:0.1:10; and ydata as in the attached file

thank you very much :)