基于贝叶斯检验动态预测置信区间的统计过程控制方法  

A Statistical Process Control Method Based on Bayesian Test for Dynamic Prediction Confidence Interval

作　　者：朱永忠[1] 夏世源 Zhu Yongzhong;Xia Shiyuan(School of Mathematics,Hohai University,Nanjing 211100,China)

机构地区：[1]河海大学数学学院,南京211100

出　　处：《统计与决策》2025年第8期47-52,共6页Statistics & Decision

基　　金：国家自然科学基金重点项目(71831006)。

摘　　要：时间序列数据处理是多元统计过程控制方法研究的一个重要方向。PCI是一种通过预测置信区间检测时间序列数据异常的方法,文章通过Bayes检验对预测结果进行修正并进行适应性处理,提出了一种基于Bayes-PCI的统计过程控制方法。通过数值模拟,比较在不同置信度p和窗口尺寸W下的平均运行长度,分别验证了所提方法在处理一维和高维数据时的性能;通过田纳西-伊斯曼工艺数据证明了所提方法在实际应用中的可行性。Time series data processing is an important direction of multivariate statistical process control method research.PCI is a method for detecting abnormalities in time series data by predicting confidence intervals.This paper uses the Bayes test to correct the prediction results and do adaptive processing,proposing a statistical process control model based on Bayes-PCI,and then through numerical simulation,makes a comparison on the average running length under different confidence p and window size W,verifying the performance of the proposed method in processing one-dimensional and high-dimensional data,respectively.Finally,the paper employs the process data of Tennessee-Eastman to prove the feasibility of the proposed method in practical application.

关 键 词：统计过程控制 Bayes检验 置信区间 平均运行长度 

分 类 号：O212[理学—概率论与数理统计]