Identification of nonlinear dynamic systems with recurrent neural networks and Kalman filter methods

摘要

Identification of nonlinear dynamic systems has been the topic of many research projects in recent years. At this moment no uniform method to solve this problem exists. In this paper a new approach of identifying nonlinear dynamic systems is presented. It is based on the use of a General Regression Neural Network (GRNN), the parameters of which are trained by the Extended Kalman Filter Method. This strategy can be used in systems, in which not all states are accessible, and was analysed for the nonlinear behaviour of the roll-bite in rolling mills.