摘要
将多变量统计过程控制应用于过程监控与诊断,在学术研究中已经较为普遍,但在工业实践方面还未充分施行。本文用一个化工过程的实例,讨论具体实施方案。首先,用一般的主元分析模型,分别使用了Q和T^2统计量,发现实际问题中有些情况下,两者提供的控制图信息不完全一致,造成工程人员难以分析判断。为提供更好的解决方案,采用两个新的统计量来代替Q统计量,应用结果表明,当Q统计量进一步分解后,可以对过程运行的状态作出更细致的解释,有助于找出过程运行中的故障。
Multivariate statistical process control for process monitoring and diagnosis are becoming more common in academic research, but are still underutilized in industrial practice. This paper discusses a practical case study on a chemical process. Conventional principal component model is firstly used to analysis the practical process with the control procedures built on Q and T^2 statistic. In the real situations of process, the information described by the Q statistic sometimes does not well match with those of T^2, it will bring confusions to the engineer. To provide better detection, an improved Q statistic is adopted, which introduces two new indices instead of Q statistic. Operation results show that the two indices stemmed from Q could give better explanations to the process behavior and do help to diagnose the faults.
出处
《计算机与应用化学》
CAS
CSCD
北大核心
2006年第7期631-634,共4页
Computers and Applied Chemistry
基金
江苏省高校高新技术产业发展项目(JH02-98)
关键词
多变量统计过程控制
化工过程
主元分析
Q统计量
multivariate statistical process control, chemical process, principal component analysis, Q statistic