论文部分内容阅读
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.
Aiming at the problems of large amount of test data, high feature dimension, low sample size, slow calculation speed and low recognition rate in the identification of spacecraft electrical property monitoring system, this paper proposes a feature extraction and weighted approximate support vector based on principal component analysis (PCA) (WPSVM) .This method realizes the principal component analysis, selection and extraction of signal fault features and reduces the dimensionality of high-dimensional feature data to improve the accuracy and speed of on-line fault diagnosis of spacecraft electrical characteristics Aiming at the selection of results in PCA, this paper proposes a method for data capture using data contribution threshold, which effectively ensures the validity and consistency of the data.The results show that this method makes full use of the useful data of the spacecraft electrical characteristics monitoring system Features, effectively improve the recognition accuracy, and the calculation time is shorter, more efficient.