摘要
主元分析方法 (PCA)是基于多元统计分析的过程监测和故障诊断手段。在假设过程只存在传感器故障的情况下 ,系统地分析了PCA方法在传感器典型故障下的检测行为。首先导出了HotellingT2 和Q两个检测统计量在传感器不同故障下的变化关系和规律 ,然后从理论上给出了每个传感器故障的可检测性条件。
Principle component analysis is an effective multivariate statistical method for process monitoring and fault diagnosis. In this paper, a systematical analysis of detection behaviors of PCA for various sensor faults is presented. The characters of two basic detection statistics, Hotelling T 2 and Q are analyzed under different sensor fault types. Then fault detectablity for every individual sensor is given. The validity of analytical results is demonstrated by a real-world application to a boiler system.
出处
《传感技术学报》
CAS
CSCD
2003年第4期419-423,共5页
Chinese Journal of Sensors and Actuators
