PREDICTIVE MODELING OF A PARADIGM SPENT FUEL DISSOLVER

摘要

This paper presents an application of "predictive modeling" to a spent fuel dissolver model of interest to non-proliferation and international safeguards. Predictive modeling includes the quantification of the uncertainties induced in responses due to uncertainties in the model parameters), model calibration following the assimilation of (uncertain) experimental information, and prediction of optimal values for responses and parameters, with reduced predicted uncertainties. The predictive modeling methodology combines responses sensitivities with uncertainties in the model parameters and in the computed and experimentally measured responses. The model response sensitivities are computed by applying the "ASAM for operator-type responses, " which reduces the computational effort for this illustrative application by a factor of over 20 by comparison to the customary adjoint methods, and by a factor of over 60 by comparison to "brute-force" finite-differencing using the re-computations. This work also shows that even though the measurements were performed in a single dissolver compartment, the measurements enable a significant reduction of the predicted uncertainties in the model parameters and responses not only in the compartment where the measurements were made, but also in all other compartments.