Abstract
This paper deals with the problem of reducing the order of periodic linear time-invariant control systems which are described by input-output data with no noise.
One goal of the procedure given in this paper is to derive from input-output data the frequency transfer function matrix of the model by expanding it to the finite Kautz orthonormal function system. The other is to compose a completely controllable and observable model of the transfer function matrix by using the Gilbert's realization method. The order of expansion and the model dimension have a simple relation depending on the data, where they are identical in the case of a single input and/or a single output.
A uniform approximation criterion being is introduced, the relation between the dimension of the reduced model and the error between model and system is made definite.
As a practical example a dog systemic circulation is reduced to a two-dimensional model by the method above. This example will be useful in designing the adaptive control system of the cardiac assist pump being developed by the authors.
