The Spacecraft Health Management System Based On Bayesian Networks

With the development of space technology, the spacecraft plays a more and more important role in the scientific research and space exploration. According to the spacecraft functionality, complexity, and the improving of spacecraft’s manufacture and launch cost, the cost of spacecraft maintenance also increases. Due to the unknown operation environment of the spacecrafts, the complicated structure and high technical requirements, the spacecraft fault s occur occasionally. A health management system based on Bayesian networks is put forward to keep spacecraft steady in orbit.According to the current health managements and fault diagnosis algorithms, health management system is set up with appropriate algorithm by studying, analyzing, and comparing. And the three most significant modules of the spacecraft health management system, fault diagnosis module, fault prediction module and health assessment module are discussed.Firstly, a three-layer satellite hierarchical fault diagnosis model is established, with a structure of system –subsystem –component. Fault characteristics is taken as the primary elements of Bayesian network by the classification of the satellite function and tasks, which is defined as fault nodes and cause nodes. Spacecraft health management structure is set up by controlling and managing the states of system.Secondly, all layers of the Bayesian network structure is established with the confirmation of the relationship between fault nodes and cause nodes. And on the basis of Bayesian network, fault diagnosis is processed. Fault prediction of the dynamic Bayesian network model is realized by choosing appropriate propagation algorithm. The output of Bayesian networks are used to build evaluation for the system health value, which can realize the quantization of system health states.Finally, in each layer, appropriate simulation systems are chosen under different health management tasks to validate a complete health management system, which consist of fault diagnosis, fault prediction, and health assessment.