A Bayesian Model to Assess the Size of North Korea's Uranium Enrichment Program

Abstract

This article presents a model to estimate North Korea's uranium enrichment capacity and to identify probable bottlenecks for scaling up that capacity. Expert assessment is used to identify and estimate the size of key centrifuge materials and component stockpiles. Bayesian probability networks are used to characterize uncertainties in these stockpiles and a deterministic optimization model to estimate the capacity of North Korea's uranium enrichment program given the assumed components and materials constraints. A Monte Carlo simulation model is used to propagate uncertainties through the optimization model. An illustration of this approach, based on the opinions of three experts, suggests that North Korea was likely (about 80 percent chance) to have a larger uranium enrichment capacity than what was displayed to visitors to the Yongbyon nuclear complex in 2010. The three most important bottlenecks to increases in enrichment capacity are the availability of pivot bearings, maraging steel, and high-strength aluminum. The nature of the model allows it to be easily updated as new information becomes available about centrifuge materials and component stockpiles.