Photon-Based Active Interrogation of Shielded Nuclear Materials Using Trans-Stilbene Organic Scintillators and an Artificial Neural Network System.

Active interrogation methods based on the measurement of prompt fission neutrons show promise in the detection of shielded and hidden special nuclear materials (SNM). The detection of prompt fission neutrons in active interrogation scenarios, however, can be challenging in high radiation environments. Traditional He-3 detectors often used for fast neutron detection require moderation and do not provide any spectroscopic information. Organic scintillators, such as trans-stilbene detectors, can directly detect fast neutrons, are spectroscopy-capable, and have a faster timing characteristic than He-3 detectors. Trans-stilbene detectors are sensitive to photons in addition to fast neutrons, which requires an analysis method to discriminate neutron and photon pulses. Pulse shape discrimination (PSD) analysis is typically used to classify photons and neutrons; however, in high radiation environments PSD is significantly deteriorated due to the presence of piled-up pulses. If piled-up pulses are not flagged, they can be misclassified as neutrons using traditional PSD methods. Previous work developed and demonstrated the ability of a system of artificial neural networks (ANNs) to identify single and recoverable piled-up pulses as neutrons or photons. We extend our work by actively interrogating depleted uranium (DU) in three configurations: unshielded, iron-shielded, and polyethylene-shielded. Preliminary results obtained using the trained ANN system indicates the presence of prompt fission neutrons produced from photofission reactions in the DU target. Because bare SNM targets are unlikely to be encountered in inspection systems, this work marks an important step in establishing the capabilities of the ANN system in real-world scenarios, thereby expanding the possibilities for safeguards in photon-flux active interrogation of shielded SNM.