MODEL REDUCTION OF LINEAR DYNAMIC SYSTEMS BY SCHWARZ APPROXIMATION

Abstract

Modelling of linear dynamic systems from physical laws or identification of the same using experimental frequency / time domain characteristics and subsequent curve fitting , often leads to a large no. of differential equations or a high degree transfer function. Analysis and design of such large scale systems becomes more time consuming, cumbersome and costly in terms of computer time usage. In addition, control strategies for such models are costly to develop and some time unreliable too .Thus, there is often the need for reducing the order of such systems and a large number of research papers, books and monographs have appeared in this topic. In the existing literature ,methods for system order-reduction have been developed both in time and frequency domain .Normally the time domain methods start with a state space description of the orig`nal system while the frequency domain methods rely on the transfer function model. One major drawback in most of the existing techniques for system reduction is of getting unstable reduced models for stable systems In the present dissertation, model reduction of linear dynamic systems and their performance analysis have been discussed The method discussed is simple and computationally efficient. Computer programs have been developed in FORTRAN for carrying out the computational work .Overall, the viability ,validity and use of the model order reduction techniques developed in this dissertation have been conclusively established through several numerical experimentations .