DETECTION AND ANALYSIS OF SHIPPER-RECEIVER DIFFERENCES

Much of I he nuclear material transferred between two facilities in the nucle- ar industry is in containers that may be transferred in lots of several containers, but measured serially by both shipper and receiver. In general, the receiver makes its measurements without knowledge of the measurement values recorded by the ship- per. Unless the instrumentation at both facilities provides only gross data, dif- ferences between shipper and receiver measurements will be observed for individual containers, as shown in Figure 1, and for the lot as a whole. One must then decide either that these differences indicate an intentional or unintentional diversion or that they have arisen from errors in the measuring process. The primary goal of the research described here was to find the best technique for deciding whether the observed shipper-receiver differences were within reason- able ranges of random measurement errors or whether these differences indicated some pattern of diversion or bias shift in the measurement process. To achieve this goul it was necessary 1o extract as much information as possible from the avail able data, using not only the set of shipper-receiver differences observed and the total shipper-receiver difference for the lot, hut also the individual shipper meas- urements, receiver measurements, and the order in which the containers were meas- ured by the receiver.