An Expert System for the Identification of Nonlinear Dynamical Systems

This paper describes an Expert System that can detect and quantify the nonlinearity present in a given dynamical system and, subsequently, determine and apply the most suitable nonlinear system identification method. The internal workings, algorithms and decision making processes of the Expert System are discussed. For demonstration purposes the Expert System is applied to a nonlinear experimental test-rig. The results show that the Expert System is an automatic tool that will detect nonlinearity, choose the best class of model for the system under investigation and perform optimal parameter estimation, so that the resulting identified models are parsimonious and accurate.