Stochastic Model Simulation For Evaluation Of Spent Fuel Pond Inventory Verification Sampling Plans

The IAEA employs statistical methods to calculate the probability of detecting diversion from each nuclear material stratum in a facility. These detection probabilities (DPs) are used for both inspection planning and effectiveness evaluation. The DPs can be calculated using a deterministic or a stochastic approach. This paper describes the benchmark calculation performed using the stochastic model for spent fuel physical inventory verification (PIV). The IAEA performs PIV inspections for spent fuel pools using several different nondestructive measurement techniques. Each instrument has a varying precision and measurement time. The inspection plan, i.e., the number and type of measurements, is designed to achieve a desired probability of detection by taking into account the strength and efficiency of each measurement type. For example, a diverter’s goal amount of one significant quantity can be achieved by taking many pins from few assemblies or by taking few pins from many assemblies, or by variations between these extremes. Varying the number of pins taken from each assembly impacts the achieved DP. The sampling plan must take these variations into account when determining the number and type of measurements. When the material taken from each tampered assembly is equal, the deterministic solution is a hypergeometric distribution function, but when different numbers of pins are taken from various assemblies the solution is a multivariate hypergeometric function. The stochastic approach simply simulates an inspection given the parameters of the diversion and measurements and averages over many trials. The paper will present the results from the stochastic and deterministic approach. The stochastic model results further showed that the order in which measurements are made (high precision to low precision, etc.) does not impact the achieved DPs in the sampling plan.