Performance assessment and further optimization of the activation based criticality dosimetry system at the Belgian nuclear research centre SCK CEN

Abstract

In places where fissile materials are being handled, the risk of a criticality accident can never be absolutely eliminated. Criticality accidents require specialized dosimetry techniques, which differ markedly from those used in routine radiological protection. This is caused by the fact that one should be able to measure very high doses up to 10 Gy in mixed gamma and neutron fields with very high instantaneous dose rates up to 105 Gy/s and complex energy spectra. The criticality dosimetry system at the Belgian Nuclear Research Centre SCK CEN is accredited according to ISO 17025 by the Belgian accreditation body BELAC and is used to monitor about 4000 nuclear workers. The system is based on a dedicated criticality dosemeter with four activation detectors (gold foils with and without cadmium shielding, an indium foil and sulphur pellets) and an iterative unfolding algorithm starting from a guess neutron fluence energy spectrum. This work describes the design of the SCK CEN criticality dosemeters and the iterative unfolding algorithm. Furthermore, it also discusses two new developments. Firstly, effective neutron reaction cross sections were calculated by MCNP6.2 Monte Carlo radiation transport simulations to be used in the unfolding algorithm. These effective cross sections take into account the effects of the dosemeter, the potential presence of the body and the neutron angle of incidence. These effects were demonstrated to be important, especially for the gold foils and if the dosemeter is worn against the body. Secondly, a Bayesian unfolding algorithm based on a parametric description of the neutron fluence energy spectrum was developed and tested. The results of this new algorithm were in good agreement with the iterative algorithm. The uncertainty assessment was more realistic with the Bayesian algorithm because it includes also the uncertainty associated with the unfolding process. However, for spectra that cannot be well described by the parametric function, the Bayesian algorithm can lead to significant under- or overestimations. In case sufficient information is available, it is better to make use of the iterative algorithm with a realistic guess spectrum to start from. Finally, this work also demonstrated the compliance of the SCK CEN criticality dosimetry system with the IAEA criteria for criticality dosimetry systems.