Results of tests for explosives in luggage from fast-neutron time-of-flight transmission measurements

Energy-dependent neutron attenuations measured by time-of- flight techniques are fitted with measured cross sections to determine projected number densities of H, C, N, and O in each 3 by 3-cm2 pixel of a suitcase. Attenuations due to other elements are lumped together as element X. Number densities are normalize by the total to reduce effects of variable thickness. Normalized number densities are constrained to lie in a 4D space. That space is finely binned. A fifth dimension contains six bins of N + O density, as deduced from average suitcase thickness and measured projected number densities. Each bin is labeled with the probability of encountering an explosive, as determined from distributions of elements from simulated explosive and nonexplosive situations. Probabilities for individual pixels are combined to produce an overall probability of an explosive suitcase. About 130 measurements involving 50 suitcases have been made. For about half of those measurements, an explosive was placed in the suitcase. Ten different plastic explosives were used in varying amounts and configurations. When suitcases are ordered by explosive probability, and a threshold is set to minimize errors, the false alarm rate is about 2 percent and the explosive detection rate is about 88 percent. Potential improvements are discussed.