Neural networks applied to the control of an induction motor

This paper deals with the control of an induction motor using a rotor field oriented method. Special attention is paid to the three independent neural networks applied to the two current control loops and the speed loop. These controllers can be adaptive and they can be tuned on-line following a gradient algorithm. The initial weights of neural networks are set by means of a fuzzy approach. Experimental results of a speed inversion under no load and under load are presented. The transient responses are commented upon and the results show the effectiveness of these kinds of ways.