Abstract
This paper presents a new modification of the multiple integration method [1, 2, 3] for continuous nonlinear SISO system identification from measured input - output data. The model structure is changed compared with [1]. This change enables more sophisticated systems to be identified. The resulting MATLAB program is available in [4]. As was stated in [1], there is no need to reach a steady state of the identified system. The algorithm also automatically filters the measured data with respect to low frequency drifts and offsets, and offers the user a potent tool for selecting the frequency range of validity of the obtained model.
Highlights
This paper presents a new modification of the multiple integration method [1, 2, 3] for continuous nonlinear SISO system identification from measured input – output data
The functions fi are supposed to be linear in parameters ai, j, i.e., to have the form mi å fi(u (t), y(t)) = ai, j × gi, j (u (t), y (t)) ; i = 0, 1, K, n, (5) j =1 where gi, j are known functions of the measured data and ai, j constants
Comparison with [1].) Let us suppose that u and y are the only measurable quantities in the system
Summary
Let us take a nonlinear continuous time-invariant SISO (Single Input – Single Output) system, described by the state equations (see Fig.). ; i=n and an output equation y(t) = g -y1[g u(u (t)) + x1(t), u (t)]. The functions fi are supposed to be linear in parameters ai, j , i.e., to have the form mi å fi(u (t), y(t)) = ai, j × gi, j (u (t), y (t)) ; i = 0, 1, K, n , (5) j =1 where gi, j are known (generally nonlinear) functions of the measured data and ai, j (generally unknown) constants. Example: f3(u, y) = a3, 1 × u1 × y - a3,2 × y3 + a3, 3 × cos(u1 + u2 - y)
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
