For many applications, ranging from controls engineering to natural
sciences and economics, precise dynamic models must be derived. In the
vast majority of applications, such precise models cannot be derived by
theoretical considerations only. The book discusses methods, which allow
the determination of dynamic models based on measurements taken at the
process, which is known as system identification or process
identification respectively.
After a short introduction into the required methodology of
continuous-time and discrete-time linear systems, the focus is first on
the identification of non-parametric models with continuous-time signals
employing methods such as Fourier transform, measurement of the
frequency response and correlation analysis. Then, the parameter
estimation for parametric models is presented with a focus on the method
of Least Squares, followed by some of its most prominent modifications.
Issues such as parameter estimation for time-variant processes,
parameter estimation in closed-loop, parameter estimation for
differential equations, continuous time processes and efficient
implementations of the algorithms are discussed. The different methods
are compared and an outlook is given on non-linear system identification
methods, such as neural networks and look-up tables.
Powerpoint slides for a 12-14 week graduate level course can be made
available to teachers
