摘要
核函数主元分析KPCA(kernel princ ipal component analysis)能够提取机械故障信号的非线性特征,可以应用于机械故障状态识别。但是KPCA是一种无监督的特征提取方法,不能利用故障信号中的类别信息。本文介绍了一种核最优K-L变换,它可以充分利用类别信息,它能够提取类平均向量和方差向量中的判别信息,使提取的特征分类效果更好。在齿轮故障诊断实验中,采用核最优K-L变换提取故障信号的非线性特征,实验结果表明核最优K-L变换相比KPCA故障识别结果更为理想。
Kernel principal component analysis （KPCA） is capable of extracting nonlinear features from fault signals, thus applied to fault condition identification. But KPCA is an unsupervised feature extraction method and cannot make use of class information of fault signals. A kernel K-L transformation method proposed in the paper can extract discrepant information on each class mean vector and variance vector, and the classification effects of the features extracted are better. In a gear fault diagnosis experiment, kernel K-L Transformation was applied to extracting nonlinear features from fault features. An experiment shows kernel K-L Transformation is more effective for fault identification than KPCA.
出处
《机械科学与技术》
CSCD
北大核心
2006年第3期288-291,共4页
Mechanical Science and Technology for Aerospace Engineering
