Reduction of discrete multivariable systems using pade type modal methods

A method of obtaining reduced order models is described for discrete multivariable systems. Irrespective of whether the original system is given in state space form (A, B, C) or in transfer matrix form [G (z)] the reduced order models are always in state space form (Aγ, Bγ, Cγ). The reduced order models retain the important characteristics such as stability steady state values etc. of original system and approximate it in pade sense. In the proposed method the common denominator of reduced order model is obtained by selecting the dominant poles in z and/or w domain. This fixes the structure of Aγ matrix. The Bγ and Cγ matrices are also chosen accordingly and then some of the elements of Bγ/ Cγ matrices are specified in such a way that the pade equations are rendered linear. The solution of these equations leads to unknown elements of Bγ/Cγ matrices. Examples illustrate the method.