Control and Identification of DC Machine by Neural Networks

This paper shows the performance of a Backpropagation neural network (BPNN) for off-line identification and control of separately excited DC motor (SEDCM). The choices of process excitation signal, data sampling time, and neural network model structure are investigated by time-domain simulation. Validity of neural networks models are carried out by cross-validation test. Direct inverse control scheme is used. The performance of BPNN controller is investigated based on step response, sharp changes in speed trajectory, sudden loads, and changes in motor parameters. The simulation results shows that BPNN controller has excellent step response, adapt well to sharp instantaneous changes in speed trajectory and load connected to the motor, and adapt to the changes of motor parameters.