The Multiple Classification Method of Signal Recognition for Spacecraft Based on SAE Network

  • Wei Lan
  • , Yixin Liu
  • , Zhang Qi
  • , Shimin Song
  • , Chun He
  • , Lijing Wang
  • , Ke Li*
  • *Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Based on deep learning, a multi-classification algorithm network is designed for the large amount of data generated in spacecraft test. In the algorithm, the initial offsets and weights of a multi-layer neural network are initialized using an auto-encoder method. The initialized parameters are monitored by the gradient descent method to make the dimension data more separable. Many shortcomings of traditional algorithms can be effectively overcome using this algorithm. For example, the storage space can be reduced and the calculation time can be saved when the data is large or complex. Expert knowledge of the spacecraft health management platform can be provided through the study of measured data. Experimental data shows that the depth learning algorithm which is based on SAE has higher accuracy in spacecraft multi-class signal testing.

Original languageEnglish
Title of host publicationMan-Machine-Environment System Engineering - Proceedings of the 18th International Conference on MMESE
EditorsBalbir S. Dhillon, Shengzhao Long
PublisherSpringer Verlag
Pages679-689
Number of pages11
ISBN (Print)9789811324802
DOIs
StatePublished - 2019
Event18th International Conference on Man-Machine- Environment System Engineering, MMESE 2018 - Nanjing, China
Duration: 20 Oct 201822 Oct 2018

Publication series

NameLecture Notes in Electrical Engineering
Volume527
ISSN (Print)1876-1100
ISSN (Electronic)1876-1119

Conference

Conference18th International Conference on Man-Machine- Environment System Engineering, MMESE 2018
Country/TerritoryChina
CityNanjing
Period20/10/1822/10/18

Keywords

  • Auto-encoder
  • Data compression
  • Deep belief network
  • Deep learning
  • PHM
  • Pattern recognition

Fingerprint

Dive into the research topics of 'The Multiple Classification Method of Signal Recognition for Spacecraft Based on SAE Network'. Together they form a unique fingerprint.

Cite this