Title :
The research of spacecraft electrical characteristics identification and diagnosis using PCA feature extraction
Author :
Yi Liu ; Ke Li ; Shimin Song ; Yi Sun ; Yong Huang ; Jun Wang
Author_Institution :
Beijing Univ. of Aeronaut. & Astronaut., Beijing, China
Abstract :
As most electronic system structure is complex and uncertain, this paper presents a new efficiency method for spacecraft electrical characteristics identification. PCA (Principal Component Analysis) feature extraction, offline FCM (Fuzzy C-means) clustering and online SVM (Support Vector Machine) classifier is introduced into the registration model. At first step of the algorithm, get an expert training set by FCM clustering method, then using PCA feature extraction to get principal component. After get expert training set and using PCA feature extraction for SVM classifier make this method fast and effective. These are the foundation of online spacecraft electrical characteristics identification. A series of spacecraft electrical characteristics data demonstrate that the proposed method is more accuracy than the traditional way.
Keywords :
aerospace computing; electronic engineering computing; feature extraction; fuzzy set theory; pattern classification; pattern clustering; principal component analysis; space vehicle electronics; support vector machines; PCA feature extraction; electronic system structure; expert training set; fuzzy C-means clustering; offline FCM clustering method; online SVM classifier; online spacecraft electrical characteristics identification; principal component analysis; spacecraft electrical characteristics diagnosis; support vector machine classifier; Clustering algorithms; Electric variables; Feature extraction; Principal component analysis; Space vehicles; Support vector machines; Training; FCM clustering; PCA feature extraction; SVM classifier; spacecraft electrical characteristic;
Conference_Titel :
Signal Processing (ICSP), 2014 12th International Conference on
Conference_Location :
Hangzhou
Print_ISBN :
978-1-4799-2188-1
DOI :
10.1109/ICOSP.2014.7015232