Model reduction of linear multivariable control systems via frequency matching

A computer-aided method for model reduction of linear multivariable continuous control systems via frequency matching is presented in the paper. The objective is to obtain a simplified transfer function matrix from a given large-order transfer function matrix, so that the vital characteristics of the given system are retained in the simpler system. This is accomplished by matching the frequency responses of the given and the reduced systems by means of the weighted mean square error. As a result of this minimisation, formulas for computing the optimal coefficients for the simplified transfer function matrix are obtained. The technique is illustrated by a numerical example and the results are compared with those in the literature.