Aiming at the features of multivariable system, i. e. , strong relevancy and strong coupling among each variable, and difficult to locate fault when system fails, the fault detection and diagnostic method based on multivariable statistical process is proposed. By adopting the method of principal component analysis ( PCA ) , the statistical characteristics of original complex data space are extracted; and the useful information of principal component data can be characterized greatly by mapping projection reconstruction, thus the fault information of the system can be detected and analyzed. Through practical application in multiple variable level control system, it is shown that the PCA method is able to effectively conduct fault detection and diagnostics for productive process, and reduce the influence of exterior noises, thus provides guarantee for fault-tolerant control of complex systems.%针对多变量系统中各变量之间的强关联、强耦合和系统故障时难以定位等特点，提出了一种基于多元统计过程的故障检测与诊断方法。采用主元分析法( PCA)提取原始复杂数据空间的统计特征，经映射投影重构能最大程度表征有用信息的主元数据，以检测和分析系统中的故障信息。多变量液位控制系统的实际运行表明，主元分析法不仅能够对生产过程进行有效故障检测与诊断，而且减小了外界噪声的影响，为实现复杂系统的容错控制提供了保障。
展开▼
