摘要
通过简化多维正态分布计算各工业变量导致采样点异常的概率,依据此概率结合对工业过程的分析,实现故障源的有效识别.证明了待求多维正态分布概率落在等马氏距离边界内,利用这一结论及多元统计学原理将复杂的多维正态分布转化为F分布,解决了多维正态分布求解难的问题.仿真结果表明了这种基于多元统计过程控制(MSPC)的新方法的有效性和准确性.
By simplifying the multivariate normal distribution, the probability of each industrial variable making the sample abnormal was computed. According to the probability and the analysis of the industrial process, the fault origin was efficiently found out. It is proved that the probability for the multivariate normal distribution is in the area within the invariable Mahalanobis distance. From the conclusion and the multivariate statistical theory, the complicated multivariate normal distribution was dramatically converted into F distribution, so the difficulty of calculating the multivariate normal distribution probability was resolved. Simulation result proved the validity and accuracy of the new method based on multivariate statistical process control (MSPC).
出处
《浙江大学学报（工学版）》
EI
CAS
CSCD
北大核心
2005年第5期663-667,共5页
Journal of Zhejiang University：Engineering Science
基金
国家"十五"科技攻关资助项目(2001BA204B01)
国家"863"高技术研究发展计划资助项目(2001AA413020).
关键词
多维正态分布
故障源
马氏距离
F分布
多元统计过程控制(MSPC)
Computer simulation
Fault tolerant computer systems
Matrix algebra
Multivariable systems
Normal distribution
Statistical methods
