Singular perturbation approximation of large, dense linear systems

Author

Benner, Peter ; Quintana-Ortí, Enrique S. ; Quintana-Orti, Gregorio

Author_Institution

Fachbereich Math. und Inf., Bremen Univ., Germany

fYear

2000

fDate

2000

Firstpage

255

Lastpage

260

Abstract

We discuss the parallel implementation of singular perturbation approximation (SPA) model reduction methods for large-scale linear control systems. The methods are based on balanced truncation methods using full-rank factors of the system Gramians. These factors are computed via an efficient iterative scheme based on the sign function method for solving linear matrix equations. The implementation of these iterations and of the SPA formulae is shown to be highly efficient on parallel computers with distributed memory

Keywords

approximation theory; control system CAD; iterative methods; large-scale systems; linear systems; reduced order systems; singular value decomposition; transfer function matrices; Gramians; balanced truncation methods; distributed memory; full-rank factors; large dense linear systems; large-scale linear control systems; linear matrix equations; sign function method; singular perturbation approximation; Concurrent computing; Control system synthesis; Distributed computing; Equations; Frequency; Iterative methods; Large-scale systems; Linear systems; Reduced order systems; Steady-state;

fLanguage

English

Publisher

ieee

Conference_Titel

Computer-Aided Control System Design, 2000. CACSD 2000. IEEE International Symposium on

Conference_Location

Anchorage, AK

Print_ISBN

0-7803-6566-6

Type

conf

DOI

10.1109/CACSD.2000.900220

Filename

900220