Identification of Dynamic Systems by Digital Computer Modeling in State Space

Abstract

A dynamic system identification technique is described in which a digital computer model is constructed and adjusted until the model state transient follows the system state transient to a prescribed accuracy. This identification technique does not require testing signals and the observation and computation times are short relative to the longest system time constant. The principle of this identification technique is demonstrated by an example consisting of a combination jet pipe electrohydraulic servo and analog computer system. The system’s state-space dimension was larger than the model and the system had measurement noise. Favorable results of the experiment indicate that the method will be applicable to such problems as optimal control, adaptive control, learning control, and on-line component parameter identification.