A Kullback-Leibler information control chart for profile monitoring

• Main Authors: Chang-MingChen, 陳長明
• Other Authors: Yu-Ching Chang
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/pej7p8

碩士 === 國立成功大學 === 工業與資訊管理學系 === 107 === In many statistical process control applications, we use control chart to monitor process where performance is measured by one or multiple quality characteristics. However, some processes can be characterized better by a linear function (called linear profile). The objective of this thesis is to monitor a linear functional relationship between a response variable and two explanatory variables. Namely, we focus on monitoring the regression coefficients and the variance of error term. We design a control chart to measure difference between process in control and out of control by Kullback-Leibler distance (called ITPM chart). The performances of ITPM chart, generalized likelihood ratio (GLR) chart, and exponential weight moving average (EWMA-type) chart are compared by average run length (ARL). The simulation results show that the performance of ITPM chart is much better than other charts in detecting a range of small shift sizes when we consider detecting the regression coefficients. The overall performance of ITPM chart is much better than other charts in detecting a wide range of shift sizes when we consider detecting the variance of error term. The ITPM chart also has the advantage that users only need to design control limit. Finally, we provide an equation to obtain control limit given an in-control ARL. It will be more simple and effective to obtain the control limit for practitioners.