Lecture 11. Identification of Linear Dynamical Systems - part 3

Learning Goals: 

You should understand what are important for a successful identification experiment. You should be able to use the system identification toolbox for identifying arx, oe, armax and bj models.  You should know how to use the BIC(MDL) criterion for model order selection. You should know how to use residual analysis to guide you in model (in)validation. You should know that numerical issues might arise unless special care is taken, when identifying a system at a too high sample rate.

 

Material  Ch 12.7-8 and Ch 17 in Ljung+Glad+Hansson. (or  Ch 11.4-5 + Ch. 13 in old book)

Ch 14 (Nonlinear Models, Ch 12 in old book) is not part of the course, but is a recommended extra read for the interested student.

Lecture Slides 2022: lecture11_Sysidpart3.pdf Download lecture11_Sysidpart3.pdf

Lecture Video (from 2021): Lecture11_FULL.mp4 Download Lecture11_FULL.mp4Play media comment.

also available on Youtube 

 

A couple of typos from 2021 have been corrected in the Lecture Slides 2022

 

Code:

arxorder.m Download arxorder.m

Download arxtest.m

Download toofast.m

Download residuals.m

System Identification Toolbox:

  • Use "help ident" to investigate functionality in the system identification toolbox. 
  • Start the GUI by  the command "systemIdentification"
  • Study "doc X" where X is
    • arxstruc, selstruc, struc
    • arx, oe, armax, bj
    • present, polydata, getpvec, bodeplot
    • idpoly, iddata, idinput, idmodel/sim
    • compare, resid