[Zurück]

G. Fuchs, A. Steindl, S. Jakubek:
"Order Reduction for a Realtime Engine Model Using Flat and Nonlinear Galerkin Methods";
Vortrag: European Conference of Systems (ECS '10), Puerto de la Cruz; 30.11.2010 - 02.12.2010; in: "Advances in Communications, Computers, Systems, Circuits and Devices", V. Mladenov, K. Psarris, N. Mastorakis, A. Caballero, G. Vachtsevanos (Hrg.); WSEAS Press, (2010), ISBN: 978-960-474-250-9; S. 255 - 260.

In this paper a methodology for the order reduction of
large scale nonlinear dynamic systems is proposed and discussed.
It is based on the known concepts of linear and nonlinear Galerkin
projection. For the actual mode of operation the projection space is
determined by snapshot decomposition. Nonlinear Galerkin methods
require the existence of an attracting manifold of the system. As an
approximation an iterative scheme is proposed, which is augmented
and stabilized by making use of the local Jacobian of the system.
The main idea pursued in this context is based on the notion that
many technical systems can be easily decomposed into operating
regimes which are characterized by locally constant Jacobians. The
effectiveness of the proposed methodology is demonstrated by a large
scale dynamic model of a turbocharged heavy-duty diesel engine.

Model order reduction, Singular value decomposition, Snapshot method, Galerkin methods, Local model network