Material Classification by Analysis of Prompt Photon Spectra Induced by 14-Mev Neutrons

Abstract

Neutron based technologies are widely used in the field of bulk material analysis. These methods employ characteristic prompt gamma rays induced by a neutron probe for classification of the interrogated object using the elemental parameters extracted from the spectral data. Automatic data analysis and material classification algorithms are required for applications where access to nuclear spectroscopy expertise is limited and/or the autonomous robotic operation is necessary. Data obtained with neutron based systems differ from elemental composition evaluations based on chemical formulae due to statistical nature of nuclear reactions, presence of shielding and cladding, and other environmental conditions. Experimental data that are produced by the spectral decomposition can be expressed graphically as sets of overlapping classes in a multidimensional space of measured elemental intensities. To discriminate between classes of various materials, decision-tree and pattern recognition algorithms were studied. Results of application of these methods to data sets obtained for a pulsed 14-MeV neutron generator based active interrogation system are discussed.