New Method of Position Estimation for Self-Sensing Active Magnetic Bearings Based on Artificial Neural Network

Artificial neural network (ANN) - as general tools for implementing nonlinear mapping between inputs and outputs - is proved to be feasible position estimation for self-sensing active magnetic bearings. A neural network with five neurons in hidden layer is constructed and well trained. With neural network act as position feedback, the active magnetic bearings system performed well. Simulation results show that neural networks can well extract the information of rotor position from the current wave form.