未知多故障诊断的扩展指定元分析方法  

Extended DCA method for unknown multiple faults diagnosis

作　　者：周福娜[1,2] 文成林[3] 汤天浩[2] 陈志国[1] 

机构地区：[1]河南大学计算机与信息工程学院,河南开封475004 [2]上海海事大学电气自动化系,上海200135 [3]杭州电子科技大学自动化学院,浙江杭州310018

出　　处：《华中科技大学学报（自然科学版）》2009年第S1期83-85,89,共4页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(60804026);河南省国际合作项目(094300510043);河南省科技攻关项目(092102210207);河南省科学自然科学基金资助项目(2009A510001);浙江省科学重点科研国际合作项目(2006C24G2040012)

摘　　要：为了克服已有指定元分析(DCA)分析方法无法进行未知类型故障诊断的不足,给出一种扩展DCA(EDCA)分析方法.首先将观测数据关于已知指定模式做DCA分析,并移除相应变化模式的影响,根据残差的显著性进行未知故障的检测;在有新故障发生的情况下,利用贡献图法确定新故障模式;再将残差数据关于新故障模式做DCA分析;重复上述过程,直到残差不再显著为止.包含4种共存故障的观测数据的仿真研究表明,EDCA可以很好地进行未知故障的检测和新故障方向的确定.Existed DCA(designated component analysis) method is only validated in diagnosing faults which has been defined in advance.In this paper,an extended DCA method is developed to solve this problem: implement DCA to the observation in the first,and remove the effect of known designated patterns to obtain the residual;then,determine whether unknown faults have occurred in the system via significance of the residual;if new faults are involved,implement PCA(principal component analysis) to the residual,and define the new fault pattern via contribution plot of the residual;implement another turn of DCA to the new residual for the new fault pattern;repeating this process until the residual is small enough.Simulation for data involved 4 faults shows the efficiency of this EDCA fault diagnosis method for unknown multiple faults diagnosis.

关 键 词：未知故障 扩展指定元分析 故障模式 主元分析 故障诊断 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]