| Summary: | 碩士 === 南台科技大學 === 工業管理研究所 === 91 === The main purpose of applying statistical process control schemes（SPC）is to detect the process variation when process is out of statistical control. By doing this, one can detect the process variation and improve the process before nonconforming products are made.
In general, the assumption of conventional statistical process control methods is that process data are normally and independently distributed. One first collects measurements from the process, then sets the control limits and center line of the control chart, and finally plots the data on the control chart to analyze if the process is in statistical control. However, measurements from industrial processes are often serially correlated. It has been shown many false alarms occurred if the correlated structure of the observations is not taken into account while the conventional control charts are applied. Therefore, developing a good method to detect the process variation with correlated measurements becomes very important.
The research investigates the impact of correlated measurements on the performance of the zone control chart（ZCC） based on the values of average run length (ARL). In addition to this, experimental design method is used to evaluate the performance of ZCC on the process with correlated observations. Finally modify the ZCC to improve its performance on detecting process variation. The research considers time series data of ARIMA model and various step shifts on the process mean. The research applies SAS programming language to simulate correlated observations of ARIMA models and takes into account two methods for estimating process standard deviation. ARL values are calculated under various step shifts in the process mean in order to compare the performance among different control schemes. This research develops a good
modified control scheme based on ZCC suitable for the process with ARIMA-correlated data.
|
|---|
