北京航空航天大学学报 ›› 2015, Vol. 41 ›› Issue (7): 1177-1182.doi: 10.13700/j.bh.1001-5965.2014.0482

• 论文 • 上一篇    下一篇

基于PCA和WPSVM的航天器电特性识别方法

李可1, 刘祎1, 杜少毅2, 孙毅3, 王浚1   

  1. 1. 北京航空航天大学 航空科学与工程学院, 北京 100191;
    2. 西安交通大学 人工智能与机器人研究所, 西安 710049;
    3. 中国空间技术研究院, 北京 100194
  • 收稿日期:2014-07-31 修回日期:2014-09-26 出版日期:2015-07-20 发布日期:2015-07-30
  • 通讯作者: 李可(1980—),男,江苏徐州人,讲师,like@buaa.edu.cn,主要研究方向为智能控制、机器学习等. E-mail:like@buaa.edu.cn
  • 基金资助:
    航空科学基金(2012XX1043); 中央高校基本科研业务费专项资金(YWF-14-HKXY-017)

Spacecraft electrical characteristics identification method based on PCA feature extraction and WPSVM

LI Ke1, LIU Yi1, DU Shaoyi2, SUN Yi3, WANG Jun1   

  1. 1. School of Aeronautic Science and Engineering, Beijing University of Aeronautics and Astronautics, Beijing 100191, China;
    2. Institute of Artificial Intelligence and Robotics, Xi'an Jiaotong University, Xi'an 710049, China;
    3. China Academy of Space Technology, Beijing 100194, China
  • Received:2014-07-31 Revised:2014-09-26 Online:2015-07-20 Published:2015-07-30

摘要: 针对航天器电特性监测系统识别过程中存在测试数据量大、特征维数高、样本少、计算速度慢和识别率低等问题,提出基于主成分分析(PCA)的特征提取和加权近似支持向量机(WPSVM)的在线故障诊断方法.实现了对信号故障特征的主成分分析、选择和提取,并对高维特征数据实现了降维,提高了航天器电特性在线故障诊断的准确性和速度.针对PCA中的结果选取问题,提出运用数据贡献度阈值进行数据截取的方法,有效地保证了数据的有效性与一致性.结果表明:该方法充分利用了航天器电特性监测系统的有用数据特征,有效提高了识别的精度,且计算时间较短,效率较高.

关键词: 航天器, 主成分分析(PCA), 降维, 小样本, 支持向量机(SVM), 电特性识别

Abstract: To solve the problems of large amount of unlabeled test data, high dimension characteristics, slow computing speed and low recognition rate during the spacecraft electrical characteristics identification process of monitoring system, an on-line identification algorithm based on principal component analysis (PCA) feature extraction and weighted proximal support vector machine (WPSVM) was proposed. The principal component analysis is used for feature selection and extraction during complex signal analysis process, to reduce the characteristics dimension and improve the speed of the spacecraft electrical on-line identification. In order to resolve the PCA results selection problem, our team put forward data capture contribution method by using threshold to capture data, effectively guarantee the validity and consistency of the data. The experimental results indicate that this method we proposed can get better spacecraft electrical characteristics data feature, improve the accuracy of identification, and shorten the compute-time with high efficiency at the same time.

Key words: spacecraft, principal component analysis (PCA), dimensional-reduction, small sample, support vector machine (SVM), electrical characteristics identification

中图分类号: 


版权所有 © 《北京航空航天大学学报》编辑部
通讯地址:北京市海淀区学院路37号 北京航空航天大学学报编辑部 邮编:100191 E-mail:jbuaa@buaa.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发