【摘要】 研究了将多指标在一个图上标示的多指标控制图。提出了三种质量控制限并作了比较。控制上限(UCL)取Hotelling的T2统计量的α/2上侧分位数。预控上限(UPCL)取Hotelling的T2统计量的λα/2上侧分位数。当第k个观测点的T2k值超过UPCL时,则接下去的抽样区间将缩短,以便及早发现失控状态。对超出控制限的情况下如何确定哪一个质量指标发生异常的问题,提出了一种可靠的分析方法。用这种方法解决了某卷烟厂的烟丝生产过程中的多指标质量控制和管理问题。更多还原

【Abstract】 Several correlated process characteristics are usually measured simultaneously in manufacturing for monitor the quality.Multivariate control charts which combine multiple characteristics into a single chart are used to monitor industrial process.Three kinds of control limits are given and compared with each other.The upper control limit（UCL） is obtained by determining the α/2 quantile of the T² distribution to give a false alarm rate of α/2.The upper prediction control limit（UPCL） is obtained by determining the 1 - λα/ 2 quantile of the T₂ distribution.When a value of T²_kexceeds an UPCL,the next sampling interval is reduced.Once an out-of-control signal is given,it is difficult to determine which component of process is out of control.We propose a diagnostic method to distinguish the component which is out of control.An application of this procedure can be illustrated using a cigarette silk quality monitoring example that 76 observations were measured on five variables.更多还原