A METHODOLOGY FOR NEURAL-NETWORK TRAINING FOR CONTROL OF DRIVES WITH NONLINEARITIES

The learning process of a multilayered feedforward neural network involves extracting a desired function from the training data presented through an appropriate training algorithm. To achieve the desired function, the generation of good training data is an important issue which needs to be addressed. This paper presents a closed-loop methodology for neural network training for control of drives with nonlinearities. In the paper, problems associated with the more common open-loop training scheme, and how these are addressed by the proposed closed-loop method, are discussed. An inverse nonlinear control using neural network (INC/NN), a control strategy which incorporates the neural network for control of nonlinear systems, is described and used to demonstrate the effectiveness of the closed-loop training scheme for neural network. Simulation studies and experimental results are presented to verify the improvement achieved by the closed-loop training methodology.