A comparison of classical, state-space and neural network control to avoid slip

A neural network controller has been implemented on a microcomputer in Pascal to control the slip in the drive of a film-copying machine. Two conventional controllers were also used; these were a simple integral controller and a state-space controller with an estimator. Control with varying loads showed a step response could be followed with no overshoot and with a setting time of less than 0.1 s by the neural controller. The neural network was 50% faster than the other controllers in responding to step velocity input preventing slip from occurring at very small axle load. The response to a sine wave velocity input was shown to be non-linear but of the same amplitude for the neural network. Experiments with four times faster sampling gave a 30% improvement in response times.