55 / 2025-03-29 20:40:23

Spacecraft Fault Diagnosis based on Coupled Feature Extraction from High-Dimensional Data

spacecraft,high-dimensional data,coupled features,generalization,fault diagnosis

In recent years, with the rapid development trend of satellite constellations in the field of space, in order to ensure the safe, reliable and stable operation of constellations, each on-orbit spacecraft is required to have intelligent autonomous fault diagnosis capability to reduce the pressure of ground measurement and control. Considering the characteristics of different types of spacecraft system configurations, closed-loop propagation of spacecraft control system faults, and high-dimensional coupling of the whole satellite system data, this paper proposes a generalized fault diagnosis method based on the autonomous extraction of high-dimensional data coupling features, which adopts a Stacked Autoencoder to intelligently extract the fixed-format generalized two-dimensional fault pattern, and then combines with a deep convolutional neural network to realize high-precision fault diagnosis and localization. Combined with the ground test data of spacecraft attitude and orbit control system with arbitrary configuration, the effectiveness of the proposed method is verified by comparing with various intelligent diagnosis algorithms.