Simulation of Unexploded Ordnance Identification Using Tagged-neutron Interrogation and Neural Networks
MetadataShow full item record
This research is a collaboration between the Department of Mechanical Engineering at Stony Brook University and the Department of Environmental Sciences at Brookhaven National Laboratory. It is proposed to use 14 MeV neutrons tagged by the associated particle neutron time-of-flight technique (APnTOF) to identify the fillers of unexploded ordnances (UXO). The ultimate goal is to construct a prototype portable neutron interrogation probe that will search for UXO in a target volume, locate targets in three dimensions, and accurately identify the major elemental constituents. To facilitate the design of a prototype system, a preliminary simulation model was developed, using the Geant4 toolkit. This work established the toolkit environment for generating tagged neutrons, their transport and interactions within a sample to induce emission, and detection of characteristic gamma-rays. Continuous coincident neutron and alpha fluxes were based on the Deuterium-Tritium fusion reaction principle which produces back-to-back emissions of neutrons and alpha particles of 14.1 and 3.5 MeV respectively. An algorithm has been developed for correlating the positions of alpha particles hitting the alpha detector with the tagged neutron and gamma-ray time-of-flight information, thereby making the system capable of 2D and 3D-image reconstruction of the interrogated object. The thesis demonstrates the novelty of the tagged-neutron approach for extracting useful signals of an object-of-interest with high signal-to-background ratio. Simulations indicated that an UXO filled with the RDX explosive, hexogen (C3H6O6N6), can be identified to a depth of 20 cm when buried in soil. The energy deposited in the detectors by gamma-rays from several elements and materials was recorded, and spectra were plotted. Neural networks were constructed with Matlab for spectra pattern identification. The results showed that the networks can effectively differentiate hexogen from other innocuous materials like nylon and PAN.