MOCABA is a combination of Monte Carlo sampling and Bayesian updating algorithms for the prediction of integral functions of nuclear data, such as reactor power distributions or neutron multiplication factors. Similarly to the established Generalized Linear Least Squares (GLLS) methodology, MOCABA offers the capability to utilize integral experimental data to reduce the prior uncertainty of integral observables. The MOCABA approach, however, does not involve any series expansions and, therefore, does not suffer from the breakdown of first-order perturbation theory for large nuclear data uncertainties. This is related to the fact that, in contrast to the GLLS method, the updating mechanism within MOCABA is applied directly to the integral observables without having to "adjust" any nuclear data. A central part of MOCABA is the nuclear data Monte Carlo program NUDUNA, which performs random sampling of nuclear data evaluations according to their covariance information and converts them into libraries for transport code systems like MCNP or SCALE. What is special about MOCABA is that it can be applied to any integral function of nuclear data, and any integral measurement can be taken into account to improve the prediction of an integral observable of interest. In this paper we present two example applications of the MOCABA framework: the prediction of the neutron multiplication factor of a water-moderated PWR fuel assembly based on 21 criticality safety benchmark experiments and the prediction of the power distribution within a toy model reactor containing 100 fuel assemblies.

Several methodologies using different levels of approximations have been developed for propagating nuclear data uncertainties in nuclear burn-up simulations. Most methods fall into the two broad classes of Monte Carlo approaches, which are exact apart from statistical uncertainties but require additional computation time, and first order perturbation theory approaches, which are efficient for not too large numbers of considered response functions but only applicable for sufficiently small nuclear data uncertainties. Some methods neglect isotopic composition uncertainties induced by the depletion steps of the simulations, others neglect neutron flux uncertainties, and the accuracy of a given approximation is often very hard to quantify. In order to get a better sense of the impact of different approximations, this work aims to compare results obtained based on different approximate methodologies with an exact method, namely the NUDUNA Monte Carlo based approach developed by AREVA GmbH. In addition, the impact of different covariance data is studied by comparing two of the presently most complete nuclear data covariance libraries (ENDF/B-VII.1 and SCALE 6.0), which reveals a high dependency of the uncertainty estimates on the source of covariance data. The burn-up benchmark Exercise I-1b proposed by the OECD expert group "Benchmarks for Uncertainty Analysis in Modeling (UAM) for the Design, Operation and Safety Analysis of LWRs" is studied as an example application. The burn-up simulations are performed with the SCALE 6.0 tool suite.

A Monte Carlo-based Bayesian inference model is applied to the prediction of reactor operation parameters of a PWR nuclear power plant. In this non-perturbative framework, high-dimensional covariance information describing the uncertainty of microscopic nuclear data is combined with measured reactor operation data in order to provide statistically sound, well founded uncertainty estimates of integral parameters, such as the boron letdown curve and the burnupdependent reactor power distribution. The performance of this methodology is assessed in a blind test approach, where we use measurements of a given reactor cycle to improve the prediction of the subsequent cycle. As it turns out, the resulting improvement of the prediction quality is impressive. In particular, the prediction uncertainty of the boron letdown curve, which is of utmost importance for the planning of the reactor cycle length, can be reduced by one order of magnitude by including the boron concentration measurement information of the previous cycle in the analysis. Additionally, we present first results of non-perturbative nuclear-data updating and show that predictions obtained with the updated libraries are consistent with those induced by Bayesian inference applied directly to the integral observables.

No abstract

Abstract. MYRRHA is a lead-bismuth cooled MOX-fueled accelerator driven system (ADS) currently in the design phase at SCK•CEN in Belgium. The correct evaluation of the decay heat and of its uncertainty level is very important for the safety demonstration of the reactor. In the first part of this work we assessed the decay heat released by the MYRRHA core using the ALEPH-2 burnup code.The second part of the study focused on the nuclear data uncertainty and covariance propagation to the MYRRHA decay heat. Radioactive decay data, independent fission yield and cross section uncertainties/covariances were propagated using two nuclear data sampling codes, namely NUDUNA and SANDY. According to the results, 238 U cross sections and fission yield data are the largest contributors to the MYRRHA decay heat uncertainty. The calculated uncertainty values are deemed acceptable from the safety point of view as they are well within the available regulatory limits.