A satellite fault in-orbit real-time fault diagnosis method and system based on deep learning

Abstract

The invention discloses a satellite fault in-orbit real-time fault diagnosis system and method based on deep learning. The method comprises the following specific steps: (1) acquiring historical observation data of a satellite single machine and a subsystem; Wherein the historical observation data is remote measurement data; (2) constructing a training set by adopting historical observation data of the single satellite and the sub-system, and training the deep learning model to obtain a satellite fault deep learning model of the single satellite and the sub-system; (3) predicting on-orbit observation data collected in real time by adopting a satellite fault deep learning model to obtain a prediction result of a next frame; (4) obtaining actual measurement data of the next frame, and comparing the actual measurement data with a prediction result to obtain a prediction error; (5) judging whether the prediction error continuously exceeds a preset range for N times or not, and if yes, performing fault diagnosis according to a comparison result; Otherwise, repeating the steps (3)-(5). According to the invention, the satellite in-orbit fault autonomous diagnosis capability is improved.

Description

technical field [0001] The invention relates to a method and system for on-orbit real-time fault diagnosis of satellite faults based on deep learning, and is suitable for real-time fault diagnosis of high-orbit satellite on-orbit faults. Background technique [0002] With the development of aerospace technology and the increasing demand for satellites in various countries, the number of artificial satellites launched around the world has also increased dramatically, and the reliability of satellites has seriously affected the quality of satellite services. In order to ensure the safety and reliability of the satellite, it is extremely necessary for the satellite to have an autonomous fault diagnosis function. Problems faced by current satellite in-orbit fault diagnosis: [0003] 1. The number of satellites has increased sharply, and the traditional on-orbit operation and maintenance model is limited. For a long time, when the satellite is in orbit, the response to abnormal...

AI Technical Summary

Problems solved by technology

[0004] 2. The relationship between satellite failures is more complicated

[0006] 4. Satellite failure data is a small sample identification problem

Due to the difficulty in obtaining specific fault data and the huge cost, most of the existing fault identification methods are based on learning a small amount of labeled data, which is a small sample identification problem. These factors restrict the accuracy of the existing methods to a certain extent. generalization

Images