Development of Passive Neutron Emission Tomography and Its Applicability to Nuclear Safeguards

More sensitive and less intrusive Non-Destructive Assay (NDA) instruments are one of the technological development challenges in the long-term R&D plan of the Department of Safeguards of IAEA to perform partial defect verifications on a spent fuel assembly before transferred to storage facilities difficult to access. The Passive Neutron Emission Tomography (PNET) was initiated as a candidate technology to respond to the demands. This study is to propose and validate the principle of PNET, and to verify its applicability to nuclear safeguards as an NDA technology for partial defect verification. PNET is a tomography technology utilizing passive neutron measurement to obtainthe projection profile of the intensity distribution of neutron sources at different detection angles, and the projection profile can be reconstructed by the Maximum Likelihood-Expectation Maximization (ML-EM) methodology. Numerical calculations were carried out to validate the principle of PNET. The neutron source information of BWR spent nuclear fuel assembly was calculated using the SCALE6.2, and the neutron transport to the detectors was calculated by the MCNP6.2. Finally, ML-EM was applied to the neutron counts at each detector to achieve the reconstruction image. For the principle validation, the reconstruction images by PNET were examined with original fuel rod positions in air and underwater measurement conditions. As results, the reconstruction images obtained in air measurements match the original ones, but intensive noise appears in the reconstruction images and the fuel rod positions are not identified in underwater measurements. Detecting neutrons above 1 MeV would be an effective technical solution to utilize PNET underwater. Applicability of PNET to nuclear safeguards was evaluated for partial defect verification. The discriminant analysis was performed to identify the replacement of fuel rods by dummy rods statistically. Twelve sets of partial defects were constructed as learning data and discriminant conditions were derived to identify the defects 97.5% reliability. Applying this to several unknown partial defect models, probability of defect detection and fuel rod false detection were evaluated as 100%, and 2% or less. This performance satisfies the requirement for partial defect verification in IAEA safeguards.