Input Requirements and Parametric Errors for System Identification Under Periodic Excitation

Abstract

This paper provides the conditions on periodic system excitation necessary for unique identification using a multiple input state model of a dynamic system. Results include the minimum number of input frequencies necessary to uniquely determine all state and input matrix elements of an n dimensional linear system. It is shown that this development encompasses earlier work done on single input-output systems. A technique is provided for predicting parametric errors to be expected from identification under periodic excitation, and several examples are used to illustrate these errors.