A Study of Dealing With theAutocorrelated Data by Integrating EPC and SPC

• Main Authors: Shu-Lan Lu, 盧淑蘭
• Other Authors: Chau-Chen Torng
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/92426406850012769091

碩士 === 國立雲林科技大學 === 工業工程與管理研究所碩士班 === 91 === Autocorrelation commonly exists in the process industry. With reduced ARL, autocorrelation leads to increase the false alarm rate of control charts. It is required to deal with autocorrelation, because it can cause significant deterioration of control charting performance. Time series approach is generally used to describe the structure of autocorrelation. AR(1) and AR(2) models are investigated, and disturbances considered are stepped or linear types. Recent researches on process control find that integration of EPC and SPC has good performance. Purpose of this research is to study integrated EPC and SPC method in dealing with autocorrelated data. Three control charts are considered: Shwhart、CUSUM and EWMA chart. From this study,results show that when the level of autocorrelation is low, the result of integrated EPC and SPC is similar to time series model approach. However, when data with high autocorrelation, integrated EPC and SPC is better than time series model approach. CUSUM and EWMA charts are better than Shewhart chart in these situations:(1) in stepped disturbance, low level of autocorrelation and range of process shifts is 0.5 to 2 time standard deviation; (2) in stepped disturbance, high level of autocorrelation and range of process shifts is 1 to 3 time standard deviation; (3)in linear disturbance, the slope of process shifts is between 0.05 and 0.5. Therefore, it is a good choice using CUSUM or EWMA charts for process monitoring under these situations.