Two-phase linear brushless machine control approach via recurrent fuzzy neural network theory

One of the advantages of using linear brushless motor for high speed linear drives is that the position and speed of these moving machines can be precisely controlled based on the feedback control technique. This paper represents a recurrent fuzzy neural network-based control (RFNNC) approach for a two-phase linear brushless machine built with Neodymium-Iron-Boron (NdFeB) permanent magnets. The proposed RFNNC possesses four layers of recurrent neural network structure to perform the Takagi-Sugeno (T-S) fuzzy inference. The recurrence is formed by using the lagged membership-grade signals as the internal feedbacks to the membership layer, and it is expected to give the potential to trace the unknown system dynamics. The interconnection weights of the network can be on-line tuned by the gradient descent method to achieve satisfactory control performance. Verification shows that the proposed RFNNC is found to achieve good tracking performances since the on-line learning scheme is applied.