考虑轮廓间一阶自相关的二项响应轮廓控制图  

Control charts for profiles with binary data in the presence of between-profile first-order autocorrelation

作　　者：商艳芬[1] 李振 何曙光[1] Shang Yanfen;Li Zhen;He Shuguang(College of Management and Economics,Tianjin University,Tianjin 300072,China)

机构地区：[1]天津大学管理与经济学部,天津300072

出　　处：《系统工程学报》2020年第1期24-32,共9页Journal of Systems Engineering

基　　金：国家自然科学基金资助项目(71672122,71532008,71872123).

摘　　要：轮廓数据是一类广泛存在于复杂制造过程中的质量数据类型.针对轮廓间存在一阶自相关的情形,本文通过引入广义线性混合模型用来描述轮廓间的相关性,进而通过转化得到了独立轮廓数据的模型,并设计了详细的参数估计方法及步骤.在此基础上,提出了一种基于似然比统计量的简单的休哈特类型控制图.同时,针对休哈特类型控制图对于小偏移不敏感的问题,构建了基于标准化似然比统计量的累积和控制图.仿真结果表明,无论是本文所提出的休哈特类型控制图还是累积和控制图都具有较好的性能,而累积和控制图整体上优于休哈特类型控制图.Profile data exists in a lot of complicated manufacturing processes.Considering the first-order autocorrelation between profiles,the generalized linear mixed model(GLMM)is used to express the autocorrelations between profiles.Furthermore,the profiles are transformed into independent ones based on the GLMM model and related parameter estimation methods with step-by-step procedures are designed.Then,a Shewhart-type control chart based on likelihood ratio statistics is proposed to monitor the shifts in processes with first-order auto-correlated profile data.Considering the insensitiveness to small shifts in Shewhart-type control charts,a cumulative sum(CUSUM)control chart is also proposed based on the standardized likelihood ratio statistics.Simulation results show that both the Shewhart-type control chart and the CUSUM chart proposed in this paper can be used to detect shifts in a process.Moreover,the CUSUM chart generally outperforms the Shewhart-type control chart.

关 键 词：二项响应 广义线性混合模型 轮廓间自相关 似然比统计量 累积和控制图 轮廓监控 

分 类 号：TP206.3[自动化与计算机技术—检测技术与自动化装置]