Neural Network Designs for Partially Known Dynamic Systems

Abstract

Algorithms are investigated for system identification and control using neural networks and validated using on-line hardware implementation. Such algorithms require very little knowledge about the system which, together with their capability of learning, make them attractive for the modeling and control of nonlinear partially known dynamic systems. An implementation architecture for neural dynamic back propagation suitable for application to other machine tools and manufacturing processes, and a network training scheme with more general features are proposed.