부산대학교 도서관

Statistical methods are central to safeguards work. Measurements forming the basis of much materials accountancy are not perfect - ``perfect`` in the sense of being error free. Other sessions in this course address the destructive and nondestructive measurement of nuclear material, together with the inherent limitations in those measurements. The bottom line is that measurement errors are a fact of life and, since we can`t eliminate them, we have to find a rational way to deal with them. Which leads to the world of statistics. Beyond dealing with measurement errors, another area of statistical application involves the sampling of items for verification. Inspectors from the IAEA and domestic regulatory agencies periodically visit operating facilities and make measurements of selected items. By comparing their own measured values to those declared by the facilities, increased confidence is obtained. If verification measurements were not expensive, time consuming, and disruptive to operations, perhaps verification of 100% of the inventories would be desirable. In reality, many constraints lead to inspection of only a portion of those inventories. Drawing inferences about a larger ``population`` of declared items in a facility based on verification information obtained from a sample of those items is a statistical problem. There are few texts on statistics in safeguards. The lengthy exposition ``IAEA Safeguards: Statistical Concepts and Techniques`` and the US NRC book edited by Bowen and Bennet are two good sources of general information. In the next section, the subject of measurement quality is addressed. The third section covers the evaluation of MUFs, and discusses the related subjects of error propagation and sequential analysis. The final section covers verification, inspection sample size calculations, and the D statistic. The text is written at an elementary level, with references to the safeguards literature for more detailed treatment.