Nonlinear speed control of a dc motor drive system with online trained recurrent neural network

A nonlinear speed drive system with online trained recurrent neural network is proposed. The recurrent neural network is trained first off-line to learn the inverse dynamics of the dc motor drive system. The direct adaptive control approach with a model following controller is performed next to adjust the network weights using the decoupled extended Kalman filter algorithm. The experimental results obtained verify the effectiveness of the proposed control method and its robustness to external disturbances