【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.更多还原

【Abstract】 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.更多还原