Experiment designs to minimize input peak and crest factor in MIMO system identification

A4 Konferenspublikationer


Interna författare/redaktörer


Publikationens författare: Kurt E. Häggblom
Förlagsort: Piscataway, New Jersey, USA
Publiceringsår: 2019
Förläggare: IEEE
Moderpublikationens namn: Proc. 2019 IEEE Conference on Control Technology and Applications (CCTA)
Artikelns första sida, sidnummer: 499
Artikelns sista sida, sidnummer: 504
ISBN: 978-1-7281-2767-5/19/


Abstrakt

The quality of the data for system identification is of utmost of
importance. Ideally, the output data should satisfy requirements
regarding variance, and in the case of multiple-input multiple-output
(MIMO) systems, correlation between the outputs. Usually, it is desired
to achieve this by input perturbations with peak values as small as
possible. A related measure is the crest factor, which is a measure of
the (inverse) power of an input perturbation with given peak value. In
this paper, the effect of minimizing the input peak and the crest
factor, subject to desired output variances/covariances, is studied for
various types of perturbation signals. The design procedure is
completely data based; data are obtained by one or more preliminary
experiments with the system to be identified. A model of an
ill-conditioned distillation column is used for illustration.


Nyckelord

data-based design, experiment design, ill-conditioned systems, multivariable systems, system identification

Senast uppdaterad 2020-20-02 vid 06:16