Model Simplification and Digital Design of Multivariable Sampled-Data Control Systems Via a Dominant-Data Matching Method

Abstract

A dominant-data matching method is developed for fitting the matrix coefficients of a matrix pulse-transfer function or a pulse-transfer function matrix using the frequency-response data of a multivariable sampled-data control system. If the experimental frequency response data are noise free and the degree of the original system is known, the proposed method can be used to identify the pulse-transfer function matrix of the system. When the frequency response data of a high-degree multivariable system can be evaluated and the degree of the identified pulse-transfer function matrix is less than that of the original system, the obtained pulse-transfer function matrix is the reduced degree model of the original high degree system. On the other hand, if the available dominant frequency response data are the data of the desired multivariable digital controller, the determined pulse-transfer function matrix is the desired digital controller. A mixed method combining dominant-data matching and dominant-pole technique is also derived for determining a stable reduced-degree multi-variable digital controller.