NEURAL NETWORK-BASED FAULTS DETECTION AND SOLATION FOR ATTITUDE CONTROL SUBSYSTEM OF SATELLITES

摘要

The objective of this paper is to develop a scheme based on neural networks for fault detection and isolation in reaction wheels of a satellite. The goal is to decide whether a bus voltage, current loss, or temperature fault has occurred in a reaction wheel and also localize which wheel is faulty. In order to achieve these objectives, three neural networks are introduced to model the dynamics of the reaction wheels on all three axes separately. Due to the dynamic property of the wheels, the neural network architecture we apply is the Elman recurrent network with backpropagation training algorithm. The efficiency of this neural network observerbased fault detection and isolation scheme is carefully investigated, and a comparative study is conducted with the performance of a generalized Luenberger linear observerbased scheme. The simulation results demonstrate the advantages of the neural network-based method developed.