On-orbit spacecraft flywheel fault detection method based on kernel principal component analysis

Abstract

The invention discloses an on-orbit spacecraft flywheel fault detection method based on kernel principal component analysis. The method comprises the steps of correspondingly collecting the real-time state information of the rotating speed, torque and the like of a flywheel through a sensor according to different control modes and detection modes of the flywheel, carrying out noise reduction through a wavelet threshold method, and carrying out the homologous dimension expansion; carrying out high-dimensional mapping on the data through a kernel principal component analysis method, obtaining T2 and SPE control limits reflecting the degree of deviation from a training model according to historical normal data, and then obtaining a comprehensive index fault control limit; and calculating T2 and SPE statistics of the real-time data, then obtaining a comprehensive index, and judging whether a fault exists or not according to whether the comprehensive index exceeds a limit or not. According to the invention, high accuracy can be achieved for the single-machine fault and the multi-machine fault of the component-level flywheel, and the fault flywheel can be positioned; the method is more visual in detection index display, and engineering practice is utilized; and the method is high in expandability and has a relatively great application prospect.

Description

【Technical field】 [0001] The present invention provides a flywheel fault detection method based on nuclear principal component analysis for low-earth orbit spacecraft when performing inertial stabilization tasks, and the invention belongs to the field of fault detection of spacecraft actuators. 【Background technique】 [0002] Spacecraft accidents have been common since humans entered space. According to statistics, from 1990 to 2001, 121 of the 764 spacecraft launched by countries such as the United States, Europe, Japan, and Canada failed; during the period from 1990 to 2006, the economic losses caused by the failure of commercial satellites and scientific satellites were as high as 4.4 billion. Dollar. In recent years, control subsystem failures account for about 30% of the total number of failures. Typical cases include the U.S. GPS BII-7 satellite in 1996 due to attitude reaction flywheel failure, the satellite was completely out of control; in 2001, the BeppoSAX satel...

AI Technical Summary

Problems solved by technology

[0005] In summary, there are four deficiencies in the existing fault diagnosis methods. First, the ability to independently study and judge unknown faults is insufficient.

Second, the online analysis ability of massive data is insufficient

Third, the ability to deeply refine fault data is insufficient

Fourth, the ability to accurately identify complex faults is insufficient

Images