Coursecode: wb2301
Coursename: Systeemidentification and Parameter Estimation

ECTS credit points: 7

Faculty of Mechanical Engineering and Marine Technology

Lecturer(s): Van der Helm, Prof.dr. F.C.T.

Tel.: 015-27 85616

Catalog data:
Non-parametric system identification based on estimators of spectral densities. Application to open-loop and closed-loop systems. Parameter estimation for linear and non-linear systems.

Course year: MSc 1st year
Semester: 2A / 2B
Hours p/w: 2
Other hours: -
Assessment: see remarks
Assessment period: on appointment

Prerequisites: wb2206

Follow-up:

Uitgebreide beschrijving van het onderwerp:

• Analysis of unknown dynamic systems in the time-domain and the frequency domain
• Application to relation between transferfunctions and spectral densities in open-loop and closed-loop systems.
• Modelling of systems: Choice of model structure and –parametrisation
• Linear and non-linear model structures
• Parameter estimation by optimization
• Optimization techniques: Gradient methods, random-search methods, genetic algorithms
• Experimental validation of models: Coherence, Variance-Accounted-For (VAF)
• Special non-linear model structures: Expert systems, neural networks, fuzzy models.

Course material (available through Blackboard website):
Dictaat Signaalanalyse, Van Lunteren / Dankelman (in Dutch)
Dictaat Systeemidentification A
Overheads
Demonstration programs in Matlab

Literature references:

• Jenkins G.M. and Watts D.G. (1968). Spectral analysis and its applications. Holden Day, San Francisco.
• Priestly M.B. (1981). Spectral analysis and time series. Academic Press, London. ISBN 0-12-564922-3.

Remarks:
In this course many small assignments will be required. The course will be finished with a larger assigment, in the data processing of experimental data. Oral exam will be taken based on a written report of the final assigment.

Goal:
The student must be able to:

1. design test signals to identify an unknown system

-design proper experimental measurement conditions

-understand the differences between stochastic and deterministic signals

-indicate the differences in application between transient and continuous signals

2. estimate a nonparametric model of the unknown system from recorded signals

-recognize and identify open-loop and closed-loop relations between measured signals

-employ proper techniques to identify models in the frequency and time domain

-validate the nonparametric models using different indicators

3. parameterize nonparametric models

-derive the best model structure based on a priori knowledge from physics

-parameterize the dynamic relation between the recorded signals using linear and non-linear parameter estimation techniques

-implement different optimization techniques

-assess the uniqueness of the parameters using correlation analysis

-evaluate the derived parameterized model through validation techniques

-recognize three non-linear model structures, and their applicability in a given situation

Computer use:
Practical assignments on a PC with a number of available programs in MATLAB/SIMULINK

Final Assignment:
At the end of the course, a choice can be made out of three final assignments, for which recorded signals are available. The available demonstration programs have to be adapted in order to estimate a proper transfer function, and proper model parameters.

Design content: n.a..

Percentage of Design: 0%