Design Method for Fault Diagnosis of Small-Satellites Based on Multi-Level Fuzzy Neural Network

摘要

A design method based on multi-level fuzzy neural network for fault diagnosis is presented in this article which uses a multi-level method to diagnose small-satellites. In the first level, fuzzy clustering method is used to estimate which subsystem works abnormally. In this level only key parameters rather than all parameters of each subsystem are collected. In the second level, radial basis function neural network is used and most of the parameters of the faulty subsystem identified by the first level diagnosis will be considered. Then the specific component or part with fault can be confirmed. In the first level all the subsystems are integrated into one system and the small-satellite is regarded as a system to be diagnosed. Therefore, the result of the diagnosis is more accurate than traditional methods. Furthermore, in the first level only key parameters are collected and analyzed, so the complexity of the calculation can be reduced significantly. A simulation experiment for fault diagnosis on a small-satellite has been undertaken. The result indicates that in a small-satellite fault diagnosis, this multi-level method can significantly reduce the complexity of calculation in the process of the diagnosis.