【摘要】 主元分析(Principal component analysis,PCA)固有的模式复合效应使得多尺度主元分析(Multi-scaleprincipal component analysis,MSPCA)仍无法做故障模式辨识,且各尺度上和重构后数据分别建立PCA模型的计算量非常大。本文建立一种多尺度指定元分析(Multi-scale designated component analysis,MSDCA)方法,将具有明确物理意义的指定模式作为多尺度空间中观测数据和重构后观测数据的统一投影框架,旨在解决MSP-CA在多尺度空间中仍无法进行故障模式辨识的不足,并避免其在各尺度上需分别建立不同统计模型的繁琐性,同时可改进单尺度指定地分析(Designated component analysis,DCA)方法的诊断性能。仿真结果表明了该方法的有效性。更多还原

【Abstract】 Multi-scale principal component analysis(MSPCA) is not a good fault diagnosis method because of pattern compounding effect of principal component analysis(PCA).In addition,multi-scale filtering is required for multiple PCA modeling to increase the computational complexity.In this paper,a multi-scale designated component analysis(MSDCA) method is proposed by using the designated pattern as a uniform projection frame.MSDCA can well solve the problem of pattern compounding effect of MSPCA and reduce the complexity of establishing multiple PCA modeling during multi-scale filtering.Furthermore,MSDCA can also improve the performance of designated component analysis(DCA) fault diagnosis method.Simulation shows the effectiveness of this algorithm.更多还原