基于核函数主元分析的机械故障模式识别方法的研究  

Method of Machine Fault Recognition Based on Kernel Principal Component Analysis

作　　者：李志农[1] 王心怡[1] 付求涯[2] 张新广[1] 

机构地区：[1]郑州大学机械工程学院,郑州450001 [2]江西理工大学应用科学学院,江西赣州341000

出　　处：《噪声与振动控制》2008年第5期77-79,共3页Noise and Vibration Control

摘　　要：提出基于核函数主元分析的机械故障诊断方法,它保留主元分析的优点并具有处理非线性的能力。该方法通过核函数映射将非线性问题转换成高维的线性特征空间,然后对高维空间中的映射数据作主元分析,提取其非线性特征,对机械故障模式进行识别。并与主元分析方法进行对比分析,实验结果表明核函数主元分析法非常有效。A method of machine fault diagnosis was proposed using the kernel principle component analysis （ KPCA ）. This method reserves the merit of principle component analysis （ PCA） and possesses a strong ability to deal with nonlinear problem. It can shift the nonlinear problem into a high dimensional linear space problem by kernel function mapping. Then the PCA method was used to analyze the mapping data in the high dimensional space, and to extract the nonlinear features from it. The mechanical fault patterns could be recognized due to these nonlinear features. The effects of recognition of PCA and KPCA were compared each other. The experiment result showed that the KPCA method is very effective.

关 键 词：振动与波 核函数主元分析 模式识别 故障诊断 

分 类 号：TH165.3[机械工程—机械制造及自动化]