Using System Identification Process models for transfer functions with more than one zero

I am trying to use the system identification GUI, and more specifically, the process models GUI in order to identify the transfer function coefficients of a system that has more than one zero. Unfortunately, the process models GUI only allows for the possibility of one zero. I am looking for a way to use the system identification with my transfer function of at least 2 zeros and at least 3 poles. Any ideas?

Answer 1

[Rajiv Singh]

I am assuming you want to estimate a continuous-time model directly from data. There are some options:

• If you do not need to strictly control the number of zeros, estimate a state-space model followed by its conversion to ZPK form.
• If you have frequency domain data (FRD) with sample time of zero, you can use OE with nb = 3 and nk = 0. That would deliver a continuous time transfer function which you can easily convert into ZPK form.
• If you want time domain identification of an arbitrary ZPK model (with potentially constraints on pole-zero locations), the most flexible approach is the grey-box one. Use either IDGREY or IDNLGREY model form where the main work required to transforming your true parameters (poles and zeros in your case) into either state-space matrices (idgrey) or output and state derivative values (idnlgrey). See the following:

http://www.mathworks.com/help/toolbox/ident/ug/bqs6lf8.html

http://www.mathworks.com/matlabcentral/answers/12074-system-identification-using-process-models-with-more-than-3-poles-and-or-2-zeros-and-delays

http://www.mathworks.com/products/sysid/demos.html: look for tutorials on nonlinear grey box identification