Use of neural networks for generating on-off commands for flexible systems

Methods are presented that generate command profiles for flexible systems equipped with on-off actuators. Optimization methods are used to obtain the command profile for a limited number of possible slew distances. Using these data, neural networks are then trained to generate the command profiles for any desired slew distance. Several different types of networks are evaluated for use in this application. Using these results, methods are developed to produce new types of on-off commands that are more easily represented by neural networks.