Design of optimal experiments for dynamic MIMO identification

A4 Conference proceedings


Internal Authors/Editors


Publication Details

List of Authors: Kurt E. Häggblom
Editors: Antonio Espuña, Moisès Graells, Luis Puigjaner
Place: Amsterdam
Publication year: 2017
Publisher: Elsevier
Book title: 27th European Symposium on Computer Aided Process Engineering
Title of series: Computer Aided Chemical Engineering
Volume number: 40
Start page: 319
End page: 324
ISBN: 978-0-444-63965-3
eISBN: 9780444639707
ISSN: 1570-7946


Abstract

A new experiment design procedure for identification of dynamic multiple-input multiple-output (MIMO) systems is presented. The implementation of the design is similar to previously proposed rotated input designs based on an estimate of the static gain matrix of the system. However, in the new design procedure, dynamics and input constraints can be explicitly taken into account if an approximate dynamic model is available.

The design problem is formulated as a convex optimization problem, where constraints are handled by linear matrix inequalities (LMIs). The objective is to produce uncorrelated outputs with maximum variance, subject to constraints, in order to maximize identifiability. Standard input designs, where the inputs are perturbed simultaneously in an uncorrelated way, or rotated input designs based on a static gain matrix, do not generally produce outputs with the desired distribution.

Other design procedures to include dynamics and constraints have been suggested, but the proposed procedure is significantly simpler than the previous ones. Because standard software can be used for the design, it is well suited for practical applications. The method is illustrated by application to a distillation column model.


Keywords

Convex optimization, Experiment design, Identification for control, Multivariable systems, System identification

Last updated on 2019-19-09 at 05:55