Nonparametric preprocessing in system identification: a powerful tool.
Event details
| Date | 05.11.2010 |
| Hour | 10:15 |
| Speaker | Prof. J. Schoukens, Dept. of Fundamental Electricity and Instrumentation, Vrije Universiteit Brussel, Belgium |
| Location |
ME C2405
|
| Category | Conferences - Seminars |
My dream is to develop robust identification methods that provide good models to the user without requesting advanced user interactions. High quality nonparametric frequency response function measurements together with nonparametric measurements of the disturbing noise power spectra turn out to be important steps to make this dream comes through. Nonparametric models can not only be used to initialize more advanced identification methods, the results can also be used directly in modern control design approaches.
In this presentation we study first the properties of the existing nonparametric methods for estimating the plant and noise transfer functions of a linear dynamic system. The analysis is based on the recent insight that leakage errors in the frequency domain have a smooth nature that is completely similar to the initial transients in the time domain. This not only allows us to understand better the existing classic methods, but also opens the road to new better performing algorithms. These will be presented in the second part of the presentation. The presentation includes the output error setup, the errors-in-variables setup, and measurements under feedback conditions. Eventually, some of the methods are illustrated on experimental data.
Links
Practical information
- General public
- Free