摘要
核主元分析(KPCA)是一种有效的非线性特征提取方法,但其缺点是对样本中的野点比较敏感。为了消除野点对KPCA算法的影响,介绍一种鲁棒KPCA算法,通过修改特征空间中映射样本的最小重构误差表达式,并预先定义训练样本集中的野点数目,实现了在特征空间剔除野点的目的。将改进后的鲁棒KPCA算法应用于齿轮泵故障特征提取,试验结果表明:该算法的抗噪性比经典KPCA算法明显增强,能有效区分齿轮泵的不同故障模式。
Kernel principal component analysis is an effective nonlinear feature extraction method, but its drawback is more sensi- tive to outliers in the samples. In order to eliminate the effect of outliers on KPCA algorithm, a robust KPCA algorithm was presented for the purpose of eliminating oufliers in feature space, by modifying the minimum reconstruction error expression of mapping samples in feature space and predefining a few outliers in training sample set. The improved robust KPCA algorithm was applied to fault feature ex- traction for gear pump, The experimental results show that the algorithm not only can significantly enhance noise immunity than classi- cal KPCA algorithm, but also can effectively distinguish between different fault patterns of gear pump.
出处
《机床与液压》
北大核心
2013年第17期171-175,共5页
Machine Tool & Hydraulics
基金
总装备部重点国防预研项目(403040102)
国家自然科学基金青年科学基金项目(61201449)
关键词
鲁棒核主元分析
野点
齿轮泵
故障诊断
Robust kernel principal component analysis
Outlier
Gear pump
Fault diagnosis
作者简介
王涛（1977-）,男,博士研究生,副教授,研究方向为故障诊断、统计模式识别等.E-mail：taotaowang0927@sina.com.
