A Machine Learning Approach to Multi-Directional Neutron Detection

The efficient detection of neutrons from fissile nuclear materials is important for security purposes. The neutron source direction is the direction calculated from an active detector to a neutron source placed somewhere outside the detector. A novel method of detecting the direction of fissile Californium 252 neutron sources was developed using TexAT, a time projection chamber built at the Texas A&M Cyclotron Institute. The neutron source direction is identified through a machine learning approach, using the distribution of proton streaks in the active volume generated from neutron-proton elastic scattering events. A Monte-Carlo simulation was developed to simulate TexAT neutron detection. Simulated neutron proton elastic scattering data was saved for the purpose of training the machine learning directional detection algorithm. In addition, an experimental setup using the TexAT detector was created to allow for a variety of neutron source locations. The data from this setup will ensure the accuracy of the simulations and directional detection algorithm. The machine learning method will allow for the efficient detection of sources placed in multiple directions by analyzing the distribution of neutron proton scattering events.