Research On Neural Network Fault Diagnosis Technology For Small Satellite Group

With the continuous development of space technology,space exploration is gradually entering the age of cluster.Small satellites have become an important tool for space experimentations and space exploration because of their small size,light weight,short development cycle,low cost and easy networking.At the same time,complex space missions also put forward higher demands on the reliability of small satellite groups.Therefore,it is especially important for satellites to have autonomous fault diagnosis and troubleshooting capabilities.In this paper,small satellite group is taken as the research object,and the fault diagnosis technology of the group system is studied.Mainly cascaded feed-forward neural network is used as a tool to design the fault diagnosis strategy of attitude control system of leader satellite and follower satellites in a small satellite group,and the fault diagnosis strategy in satellite group formation maneuver transformation process.The main tasks are as follows:Aiming at the faults of attitude control system of the leader satellite in small satellite group,an observer based on cascaded feed-forward neural network is designed and the Lyapunov function is introduced to prove its stability.On this basis,a fault diagnosis system based on neural network observer is established.Taking the actuator fault in attitude control system as an example,the neural network observer is used to estimate the state variables of the system to determine whether there is a fault,so as to accomplish the fault diagnosis of the leader satellite in the satellite group.Aiming at the faults of attitude control system of the follower satellites,a neural network fault diagnosis strategy using the leader and followers relative attitude information is proposed.The actuator of follower satellite,PPT,is modeled and analyzed,and its two common faults are discussed.A new algorithm is used to construct the cascade feed-forward neural network,and the weight training algorithm is optimized.This neural network is used to simulate the relative attitude information between the leader and followers,a strategy with ‘one leader and two followers’ structure is designed.Aiming at the problem of small satellite groups during formation maneuver transformation process,a fault diagnosis method based on neural network observer is designed.Based on the the relative dynamics model of satellite group system and the strategy of satellite maneuver,by using the new algorithm to build a cascade feed-forward neural network.On this basis,the design of the neural network observer is completed.Taking the actuator fault as an example,the estimated output value of the observer is compared with the actual measured value of the sensor to generate the residual so as to detect the fault during the satellite group maneuvering.The effectiveness and reliability of the strategy proposed by this paper is verified through STK and MATLAB interconnect simulation and visual demonstration.The proposed neural network fault diagnosis strategy for small satellite group can lay foundation for the smooth development of space missions.