| Traditional statistical control chart methodology is based upon an assumption that the data are independent and identically distributed. However, oftentimes statistical process control data exhibit autocorrelation. It has previously been shown that smoothing algorithms can provide the basis for a method to detect nuclear material diversions and losses and can also provide a general approach to industrial statistical process control even with autocorrelated data. It is the research hypothesis of this dissertation that prior methods can be improved upon by using robust smoothers. Thus, the main objective of this dissertation is to develop a superior method of detecting when a process is no longer in statistical control. Achievement of this objective is demonstrated by an actual comparison of the methods presented in this dissertation with previously published methods. |
