USING POLYNOMIAL SMOOTHING AND DATA BOUNDING FOR THE DETECTION OF NUCLEAR MATERIAL DIVERSIONS AND LOSSES

Successive nuclear material balances form a time series of often autoc orrelated observations. Significiant deviations from the underlying in -control process model or time series pattern, i.e., outliers, indicat e an adverse process change or out-of-control situation relative to th e model. This paper uses these ideas to demonstrate a method for detec ting nuclear material losses earlier than currently used models and th us helps to alleviate a problem which is growing to world crisis propo rtions, The process control capabilities of these methods were success fully tested on two autocorrelated data sets of nuclear material balan ces with known removals. These algorithms should be of assistance to t he nuclear engineer involved in process control.