Research On The Statistical Process Control Theory And Algorithms Of Autocorrelated Process

The traditional control chart assumes that observations are independent and identically distributed. However, autocorrelated effects are often substantial in manufacturing processes and the assumption of independence is violated. There has been systematic study on the quality control theory and algorithms of autocorrelation in this paper.It contains two parts.Firstly, we analyse the effects of detection capability and its cause, when using traditional statistical process control charts monitors the autocorrelated process. The results show that the usual estimator for the standard deviation is biased in case of correlated observations. Consequently, the control limits are not correctly and the detection capability of traditional control charts is reduced. There appear such an increasing number of false alarm that lead to much of the cost of quality control.Secondly, we investigate three remedial process monitoring methods proposed for autocorrelated data:the residual control chart, the remedial exponentially weighted moving average control chart(REWMA), the autoregressive moving average control chart(ARMA).1.The approach of the residual control chart is to fit an appropriate time-series model to the observations and apply conventional control charts to the stream of residuals since they are independent and identically distributed random variables. a system change should increase the absolute value of the residuals. Thus the residual chart provides a mechanism for detecting a process change.2.Because ordinary EWMA chart methodology is based on a fundamental assumption that process data are statistically independent. So the approximate variance of the EWMA statistic is recalculated and the control limits of the REWMA chart are determinded by the process variance and autocorrelation. The REWMA chart is simple to implement, and no modeling effort is required.3.Intergrating the two former design concepts of the control methods, the ARMA charting technique bases on an autoregressive moving average statistic and developsan informal procedure to determine the appropriate parameter values of the proposed chart based on two signal-to-noise ratios. Because this new chart provides a more flexible choice of parameters to relate the autocorrelation structure of the statistic to the chart performance,it could improve the sentitivity of ARMA control charts . Studies show that it is better than the other charts for autocorrelated observations.