Unique historical data from UK criticality experiments with intermediate enriched uranium (IEU) have been collated, reviewed and evaluated as criticality safety benchmark
experiments. Four benchmark evaluations, detailing 136 configurations, have been accepted by the International Criticality Safety Benchmark Evaluation Project (ICSBEP)
and published in the International Handbook of Evaluated Criticality Safety Benchmark Experiments. This work has increased the number of published IEU benchmark configurations from 125 to the current total of 261. An overview of the experiments, a
description of the evaluations and a summary of the benchmark models are provided.
Sample calculational results obtained with the Monte Carlo codes MONK and MCNP,
using the JEF-2.2 and ENDF/B-VII.0 nuclear data-sets are presented.
In order to analyse the data, a novel data assimilation technique employing Kalman filtering is derived. This uses sensitivity data, relating changes in the neutron multiplication parameter keff , to perturbations in the underlying nuclear data. The sensitivity
module of MONK, used to obtain these data, is compared to similar codes as part of
an expert group study and is shown to give results in good agreement. Also utilised
are covariance matrices for uncertainties in the nuclear data, benchmark keff and the
sensitivity data itself-the sources of these data are explained. The assimilation scheme
is tested using a sample set of benchmark experiments to demonstrate its functionality.
The newly developed IEU benchmarks are used with the assimilation scheme coupled
to MONK, in order to calculate adjustments to the ENDF/B-VII.0 nuclear data. Similar, independent benchmarks are also selected for use in the analysis, eliminating those
which show significant statistical inconsistencies. Conclusions are drawn from this analysis; regions for consideration of the nuclear data are discussed and computational biases
and associated trends in the calculated keff data are provided. Finally, suggestions for
further work are described; these include further expansion of benchmark data, software
developments, alternative data assimilation methods and possible design properties for
future integral experiments.