A BP neural network controller for Magnetic Suspended Flywheel System

A BP neural network controller is proposed for direct suspending control for Magnetic Suspended Flywheel System (MSFS) that is supported by Active Magnetic Bearings (AMB). A one hidden layer configuration is adopted in the BP neural network, and the back propagated algorithm for network weights updating is derived based on AMB´s linear model. The discussed controller is implemented in the MSFS with random initial network weights, and it is trained online as the whole system operated. Simulations show the proposed BP neural network controller is apt to succeed in suspending the flywheel, and better performances such as precise position control, disturbance rejection, vibration suppression and quiet control are achieved under power consumption limitation. The results validate the feasibility and effectiveness of the presented BP neural network controller.