改进PCA方法在化工过程中的故障诊断研究  

Research on Fault Diagnosis of Chemical Process Based on Improved PCA Method

作　　者：杜海莲[1] 苗诗瑜 杜文霞[1] 刘小亮 

机构地区：[1]河北师范大学职业技术学院河北石家庄050024 [2]北京交通大学电气工程学院北京100044 [3]石家庄铁道大学四方学院河北石家庄051132

出　　处：《山东科技大学学报（自然科学版）》2017年第5期16-22,共7页Journal of Shandong University of Science and Technology(Natural Science)

基　　金：国家自然科学基金项目(61673160,60974063,61175059);河北省自然科学基金项目(F2014205115);河北省教育厅课题(ZD2016053)

摘　　要：为了使化工生产中复杂系统的故障判断更加精准、更加有说服力,采集系统正常工作和故障状态时的数据,运用改进的主元分析(PCA)算法判断系统是否有故障产生。改进的主元分析算法是在传统主元分析的基础上将平方预测误差SPE统计量分化成与主元显著关联的检测变量残差(PVR)统计量和其余一般变量残差(CVR)统计量,再与Hotelling’s T2统计量相配合进行系统故障的判断,使检测到的结果更加精准,生产过程更加安全。将此改进的主元分析方法运用到田纳西—伊斯曼过程中,仿真结果验证了该方法可以有效识别系统处于正常工况状态还是故障状态,是一种系统故障分析和诊断的有效方法。In order to further improve the accuracy and persuasiveness of the fault diagnosis of complex system inchemical production, data of the system at normal working state and failure state were collected, aprincipal component analysis （PCA） algorithm was used to determine whether there was fault in the system. Based on the traditional PCA, the improved PCA algorithm decomposed the squared prediction error （SPE） statistic intoprincipal-component-related variable residual （PVR） statistic and common variable residual （CVR） statistic,and it was then cooperated with Hotelling？ sT2 statistics to diagnose fault of the system so that the diagnostic results weremore accurate and the production process was more safe. Finally , the improved PCA method was applied m Tennes-see Eastman process. The results show that this method , being able to dentify the normal mode state and famore accurately , s an effective method of fault analysis and diagnosis.

关 键 词：主元分析 故障诊断 生产安全 

分 类 号：TP206[自动化与计算机技术—检测技术与自动化装置]