摘要
提出了一种基于滑动中值滤波的多尺度主元分析(MSPCA)方法,该方法利用中值滤波对主元分析(PCA)前的原始数据进行预处理,以去除异常点,并用多尺度主元分析方法把小波变换和主元分析有机结合起来,通过对过程数据的多尺度建模,来消除系统中的次要主元和小的小波系数,这样既提高了对数据中细微、重要变化的检测灵敏度,又解决了在测量数据中含有异常点的情况下,现有多尺度主元分析难以去除因异常点的存在而产生的虚警问题。仿真验证了该方法的有效性和可行性。
A new multi-scale principal component analysis （MSPCA） method based on moving median filtering is presented in the paper. The method fiistly preprocesses the original data before principal component analysis （PCA） by median filtering to eliminate the outliers, and then, effectively combines the wavelet transform with PCA to eliminate the non-principal components and small wavelet coefficients by modeling the process data at multiple scales. This method can not only improve the ability for detecting small but important changes in data, but also resolve the false-alarms problem caused by the outliers in measured data, which is difficult to be dealt with by the existing MSPCA. The simulation demonstrates the effectiveness and feasibility of the proposed method.
出处
《高技术通讯》
EI
CAS
CSCD
北大核心
2008年第3期271-276,共6页
Chinese High Technology Letters
基金
国家自然科学基金(60572051)
浙江省科学重点科研国际合作项目(2006C24G2040012)
浙江省教育厅科技计划(20050530)资助
关键词
滑动中值滤波
小波变换
多尺度主元分析
MSPCA
故障检测
故障诊断
moving median filtering, wavelet transform, multi-scale principal component analysis （MSPCA）, fault detection, fault diagnosis
