Identification of a Nonlinear Function in a Dynamical System

Abstract

Using an adaptive method introduced in (Messner et al., 1991), a standard identification technique for linear systems can be extended to identify nonlinear functions in dynamical systems under a mild condition. Specifically, the assumption is that the nonlinear function can be represented as a integral equation of the first kind. The method identifies the nonlinear function indirectly by estimating the influence function of the integral equation. By analogy to linear methods the kernel of the integral equation serves as the “regressor,” while the influence function is the “parameter” to be identified. This paper focuses on an application to disk drive servos.