基于MAF-ICA的工业过程故障诊断  

Fault Detection Based on Industrial Process of MAF-ICA

作　　者：刘春菊[1] 刘春玲 李召[3] 

机构地区：[1]石家庄铁道大学四方学院,河北石家庄050000 [2]石药集团中奇制药技术有限公司,河北石家庄051132 [3]东北大学信息科学与工程学院,辽宁沈阳110819

出　　处：《仪表技术与传感器》2017年第1期177-180,183,共5页Instrument Technique and Sensor

摘　　要：针对复杂工业数据存在自相关的特点,提出一种基于最小/最大自相关因子分析(MAF)结合独立元分析(ICA)的故障诊断方法。首先利用自相关因子分析得到监测量的最大自相关矩阵,该矩阵包含数据信息的空间特征,并且去除了噪声;然后对其进行独立元分析,根据高阶统计信息,提取出独立元;并依据I2和SPE统计量值来判断系统的工作状态;最后采用变量贡献图对故障进行定位。将该方法应用到TE过程,验证了其可行性以及与ICA和MAF相比时的优越性。There are some autocorrelation characteristics in the complex industrial data variables, a fault diagnosis method based on rain/max autocorrelation factors （MAF）and independent component analysis （ICA）is proposed.Firstly, a max autocorre- lation matrix is obtained by MAF, which contains space characteristics of data information, and efficiently eliminates noise.Sec- ondly ICA is used to extract the independent components based on the high order statistics.According to the 12 and SPE statistics, the system state is determined.Finally, the fault is located by using the variable contribution plot.ThiS method is applied to TE process, and results demonstrate it is feasibility, effectiveness and superiority when compared with single ICA and MAF.

关 键 词：自相关因子分析 独立元分析 贡献图 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]