作　　者： ; ; ;

机构地区： 华中科技大学机械科学与工程学院

出　　处： 《华中科技大学学报（自然科学版）》 2004年第12期54-56,共3页

摘　　要： 从特征提取角度出发 ,分析了曲元分析法 (CCA方法 )的非线性映射算法原理 ,研究了曲元分析在高维数据本征维数提取中的应用 ,在此基础上提出了基于非线性映射曲元分析法的设备状态特征提取技术 .对仿真和齿轮故障数据的研究表明 ,该方法能提取出对模式识别敏感的特征集 ,可有效用于机械设备的故障模式分类识别 . The principle of curvilinear component analysis (CCA) and its algorithm to realize dimension reduction and nonlinear mapping was investigated from the angle of feature extraction; how to estimate the intrinsic dimension in high-dimensional data space by using CCA method was discussed. The intrinsic dimension corresponded to the minimum number of variables necessary to describe the input data without significant loss of information. An approach to mechanical fault feature extraction was presented based on the CCA method and applied to real gear fault feature extraction. Research results indicated that the method was useful and effective in sensitive feature extraction for pattern recognition in mechanical fault diagnosis.

关 键 词： 模式识别 特征提取 非线性映射 曲元分析

领　　域： [机械工程]