Esarda nda working group - Results of the Monte Carlo
"simple case" benchmark exercise

P M J Chard UKAEA, Dounreay, UK, S Croft Canberra Industries,
Meriden, US
O Mafra ABACC, Brazil

The ESARDA Non-Destructive Assay (NDA) Working Group (WG) has previously organised several intercomparison exercises, aimed at establishing the performance of NDA techniques currently employed in safeguards. These include round - robin exercises where laboratories make comparative NDA measurements on a set of samples, and intercomparisons for data analysis codes.

Passive and Active Neutron Coincidence Counting is widely used in safeguards for the verification of fuel pins and assemblies, the measurement of the fissile content of scrap residues from reprocessing activities, and the assay of individual fuel pellets for process control.

The use of Monte Carlo modelling is becoming increasingly widespread as a tool for reducing the reliance upon experiment (which often requires the use of costly standards) for calibration of neutron coincidence counting systems. Increasing availability of powerful computers means that the complexity with which physical systems can be modelled is increasing. However, the accuracy of the results obtained is also influenced by the nuclear data constants which are used by the program, as well as the interpretational models used to convert calculated quantities into measurement parameters. In this context, there is increasing interest in the safeguards community in establishing nuclear data sets and methodologies which can be used reliably for these applications.

The NDA WG recently published, the results of an Intercomparison Exercise, the "Reals benchmark exercise", in which a number of participants used MCNP TM, an established Monte Carlo code in safeguards, to predict the coincidence counting rates for a standard Euratom Active Neutron Coincidence Collar. The emphasis of this exercise was placed on studying the methods used when applying MCNP TM, that is, the interpretational models which are used to convert the raw calculated quantities, into quantities which are relevant to measurement, namely counting rates. To this end, participants used the same nuclear data for the actual MCNP TM runs, with a fixed, predefined geometry model, but different interpretational models. The results of comparisons with experiment demonstrated that predictions could generally be made to an accuracy of 5 - 10 %. However, due to uncertainties in the accuracy of the nuclear data constants used (neutron cross-sections, neutron source spectra, thermal neutron scattering treatments), it is not clear whether, nor by how much, it is possible to further improve on this figure, nor what are the factors which determine the fundamental limits. Furthermore, the relative contributions to the differences in results from i) differences in the nuclear data used in the interpretational models, and ii) differences in the physics of the interpretational models themselves, were not clear. Although the MCNP TM modelling was based on as accurate a geometry model as possible, certain physical effects were not taken into account, such as the true effective active length of the detectors, and the fractional wall effect losses. This means that there is an additional, as yet unquantified, source of uncertainty which undoubtedly affects the level of agreement which can generally be expected between experiment and calculation.

A new "Simple Case" benchmark Intercomparison Exercise was launched, intended to study the importance of the fundamental nuclear data constants, physics treatments and geometry model approximations, employed by Monte Carlo codes in common use. The exercise was also directed at determining the level of agreement which can be expected between measured and calculated quantities, using current state of the art modelling codes and techniques. To this end, measurements and Monte Carlo calculations of the Totals (or Gross) neutron count rates have been performed using a simple moderated 3He filled cylindrical proportional counter array or "slab monitor" counting geometry. It was decided to select a very simple geometry for this exercise. This was to ensure that there is little opportunity to introduce uncertainties into the results as a consequence of errors in the geometry modelling due to the geometry being not well defined. Furthermore the use of a standard, well characterised detector system minimises the risk of introducing additional uncertainties due to errors in modelling details such as moderator density, detector fill pressures, etc. so that there is minimum potential for uncertainty due to unquantifiable variables in the geometry. The comparison between measurement and calculation was directed at the simplest possible measurable quantity, namely the Totals counting rate, in order to direct the analysis towards the influence of nuclear data, physics treatment and geometry approximations, rather than the details of a potentially complex interpretational model (the Reals Prediction benchmark focussed on this).

It was agreed that Monte Carlo modelling would be carried out by participants from as wide a range of organisations as possible. By requesting that the participants each use their preferred codes, the exercise facilitated a comparison of all the codes in common use for NDA applications in safeguards. Furthermore, the intention was for each participating group to develop their own independent geometry model, based on detailed drawings of the as – built detector system, supplied by the project co-ordinator as obtained from the manufacturers of the equipment. This gives a good overall understanding of the range of possible geometry modelling approximations and their effects. The simple geometry treated in this exercise, gives a high degree of control over the geometry variables. It was anticipated that each group would use their own preferred nuclear data sets and physics treatments, such that at the end of the exercise, analysis would allow sensitivity studies to be performed for the various factors.

By benchmarking against the experiments, it was hoped that a consensus could be reached for a preferred set of data to be used for neutron assay systems as well as providing insight into the fundamental accuracy limitations of the Monte Carlo modelling.

This report describes the scope of the measurements and calculations, and gives a summary of the results obtained (the results were presented earlier). The results are compared, and sensitivity studies shown, to determine the influence of the various parameters. It is considered that this new "Simple Case Benchmark" intercomparison exercise can potentially offer interesting and useful results to the safeguards community. The results will benefit both the NDA specialist interested in the accuracy with which Monte Carlo modelling can be used for design / calibration work and the inspector, who needs to keep up to date with the expected performance of NDA instruments and predictive tools which are increasingly being used to assist calibrations of NDA equipment.

See more

full version

ABACC Laboratories Quality Assurance Though Secondary Standars Exchange Program

    
Considerations on Safeguards Approach for Small Centrifuge Enrichment Facilities

Development of Low-Level Environmental Sampling Capabilities For Uranium at Brazilian and Argentine Laboratories by Abacc

Esarda nda working group - Results of the Monte Carlo "simple case" benchmark exercise