multivariable system model reduction by orthogonal base

• Main Authors: Min-Li Perng, 彭明麗
• Other Authors: Professor Chyi Hwang
• Format: Others
• Language: zh-TW
• Published: 1994
• Online Access: http://ndltd.ncl.edu.tw/handle/97119412039231430793

碩士 === 國立成功大學 === 化學工程研究所 === 82 === For the purposes of analysis and control,it is often required to simplify a system model.Recently,the topic of model reduction of multivariable systems by the balanced realization method has been received a great deal of attention .By the balanced truncation method,the states which have less contributions to the system response are truncated.This method share several advantages,e.g.the truncated model is stable if the origianl one is stable,and the error bound between the original and reduced models can be easily estimated. However, it suffers from the drawback that it does not assure the zero steady-state response error for a step input. In this thesis, we attempt to solve the problem by combining the block Schwarz and balanced realizations to derive stable reduced-order models for multivariable systems. The basic idea is to regard the Schwarz realization as a system decomposition scheme which decompose a stable system into subsytems of different orders while the balanced trunction rocedure is acturally performed on a decomposed subsystem.The approach adopted comprises of the following main steps: (i)The original right matrix fraction description (RMFD) is realized by the block Scharwz form,which defines a set of block orthogonal states. (ii) The decomposed subsystem is reduced by the balanced trunction method.(iii) The truncated subsystem is combined with other portion of the system to form the reduced order models for the original system. The properties of the reduced-order models thus obtained are also explored.