Consequences of non-zero item variability on the IAEA’s inspection sampling plans

Some of the IAEAs verification activities are based on verifying operator’s declarations. Verification activitiesare usually based on a sampling plan that is designed to achieve a specified probability to detect falsification ofoperator records.To develop a practical approach, the IAEA’s methods for establishing a sampling plan rely on the assumptionthat – before a potential diversion – all items in a stratum have the same nuclear material mass, i.e. the itemvariability is assumed to be zero. This mass is estimated using the average nuclear material mass from theoperator’s declarations. In reality, however, the nuclear mass of items in a stratum may vary due to processvariation. Thus, this paper investigates the consequences of the zero item variability assumption (constant massassumption) on IAEA inspection sampling plans.To quantify the impact of deviations from the zero item variability assumption the following model is considered:1) the nuclear material mass of each item is modeled as a random variable whose values are influenced by theaverage nuclear material mass and the item variability, and 2) the operator diverts a significant quantity fromthe stratum by taking all nuclear material from each of as few as possible items by gross defect falsification.Thus, the number of falsified items needed to acquire a significant quantity as well as the sample size needed toachieve a required selection (detection) probability are random variables.To quantify the impact of deviation from the zero item variability assumption, the expected sample size and theexpected achieved selection probability are evaluated. The former one is a measure of cost/time and the latterone is a measure of effectiveness.