自相关过程的改进型EWMA控制图  

The Improved EWMA Chart for Autocorrelated Processes

作　　者：张志雷[1] 

机构地区：[1]河南财经学院统计系,郑州450002

出　　处：《数理统计与管理》2008年第3期466-472,共7页Journal of Applied Statistics and Management

摘　　要：常规指数加权移动平均(EWMA)控制图的假设前提是观测数据相互独立,但在实际生产过程中,数据相关违背假设条件。本文首先讨论了序列自相关对常规EWMA控制图的影响,结果表明其检测效能降低。因此,重新估计了平稳过程的σz并在此基础上建立了改进型EWMA控制图。然后运用平均链长比较了改进型EWMA控制图与休哈特图和残差控制图,模拟研究说明当过程非强相关且过程均值发生中小偏移条件下。改进型EWMA控制图的检测效果要优于其他两种控制图。最后,通过一个实例验证了该方法的有效性。Traditional exponentially weighted moving average （EWMA） chart is based on a fundamental assumption that process data are statistically independent. Process data, however, are not always independent from each other in practice. First we discuss the effect of autocorrelation on the performance of the traditional EWMA chart and find that traditional EWMA procedures are not effective and appropriate for monitoring and improving process quality. So Z is reestimated based on a general stationary process and the improved EWMA chart is established. We compare the improved EWMA chart with the residual charts and Shewhart charts via the average run length. Simulation studies show that the improved EWMA chart is better than the others when the process autocorrelation is not very strong and the mean shifts are small to medium. Finally an example is given to indicate the effectiveness of the proposed method.

关 键 词：自相关过程 EWMA控制图 平均链长(ARL) 模拟研究 

分 类 号：O213.1[理学—概率论与数理统计]