Assessing confidence in inferring reactor type and fuel burnup: A Markov chain Monte Carlo approach

A technology to verify operator declarations of reactor type and fuel burnup has been introduced. The approach uses reactor modeling (the "forward model") to predict well-chosen isotope ratios as a function of reactor type and burnup. Mass spectrometry measurements of these ratios are then used to infer reactor type and burnup, (via a maximum likelihood solution to "the inverse problem"). We extend the previous maximum likelihood solution to this inverse problem by using Markov Chain Monte Carlo to simulate observations from the posterior distribution for reactor type and burnup. This provides an alternate estimation strategy, an estimate of the confidence in the inferred reactor type and burnup, and improves prediction performance. We also investigate the impact of increased error variance as a step toward including the effect of computer-code uncertainty. (c) 2005 Elsevier B.V. All rights reserved.