Evaluating Nuclear Fuel Cycles with Incomplete Knowledge: A Bayesian Approach to Pathway Analysis

Acquisition pathway analysis can be used to evaluate the feasibility of diversion pathways in a nuclear fuel cycle (NFC) that lead to proliferation. Safeguards can then be applied to the fuel cycle to increase the likelihood of detecting diversion along these pathways. Quantifying feasibility along different pathways can be difficult because it requires measuring subjective characteristics such as operational capabilities or detectability, both of which can change in the long term. We consider how pathway analysis can be conducted with incomplete knowledge of the fuel cycle of interest. Consider an NFC model to consist of three components: the operational parameters for nuclear facilities, the fuel cycle itself, and output observables. Several questions may be of interest from a nonproliferation perspective depending on which of these three components is unknown. The forward problem (unknown observables) can be solved with modeling and simulation, such as system dynamics or agent-based modeling. The inference problem (unknown inputs) can be addressed with a Bayesian analysis. For analyzing incomplete fuel cycles with potentially clandestine facilities discoverable by an analysis of material accountancy, we propose a Bayesian analysis to estimate the likeliness of specific pathways. Nuclear engineering can inform the plausible pathways between facilities, and acquisition pathway analysis can inform the plausible pathways for proliferation. Modeling and simulation then provide a quantitative method of evaluating these pathways in a way that considers uncertainty in operational parameters and observables. These three analyses encompass a wholesale approach for evaluating NFCs conditioned on known (potentially incomplete) knowledge of a fuel cycle.