Jun Yue

System Realization and Model Reduction with Maintainable Structures

Date: December 7, 2007

Abstract:

This thesis considers the connection between system realization and model reduction with maintainable structures. In general, system realization problem is to find a realization with desired characteristics. Here the discussion on system realization is aimed to Minimum round-off noise and coefficient sensitivity. Sectional optimization, block optimization, balanced realization, and constrained noise algorithm are developed. The latter can achieve acceptable noise gain and coefficient sensitivity with desirable structure.

The problem of model reduction for various classes of systems is also analyzed in this thesis. Model reduction is preformed under structure maintainability. The goal is to apply model reduction techniques such that the reduced order systems preserve the structures of the original models. Among the special structures, the Metzlerian structure for continuous time systems is found to be interesting in the sense that the reduction methodology can also be generalized to its uncertain interval structure.

Thesis Committee:
Prof. B. Shafai (advisor)
Prof. M. Salehi
Prof. R. Sipahi