摘要
提出一种基于核函数的Fisher鉴别分析的故障特征提取方法。该方法通过内积核函数将原始特征空间映射到高维特征空间,然后在高维特征空间作线性Fisher判别分析,从而得到原始特征空间的非线性特征,最后应用滚动轴承的故障数据对该方法进行了检验。结果表明,与线性Fisher鉴别分析和核主元分析方法相比,基于核函数的Fisher鉴别分析更适合提取机械故障的非线性特征,它所提取的故障特征对故障具有更好的识别分类能力,并且对分类器具有较强的鲁棒性。
A feature extraction approach based on kernel Fisher discriminant analysis (KFDA) was proposed. In this approach, the integral operator kernel functions were used to realize the nonlinear map from the raw feature space to high dimensional feature space.By performing Fisher discriminant analysis (FDA) on the high dimensional feature sets, the nonlinear features of raw feature space were obtained. Experimental fault data of roll bearings were used to test the performance of this method. Practical results show that the method is more suitable for nonlinear feature extraction from fault signals, the extracted features based on KPCA exhibit better ability in fault recognition and they are robust for various classifiers.
出处
《振动．测试与诊断》
EI
CSCD
2008年第4期322-326,共5页
Journal of Vibration,Measurement & Diagnosis
基金
国家自然科学基金资助项目(编号:60672179)
关键词
故障诊断
特征提取
核函数Fisher鉴别分析
模式分类
fault diagnosis feature extraction kernel-based Fisher discriminant analysis pattern classification
