I. Halvic and S. Prasad, “Genetic Algorithm for Spectrum Reconstruction of Neutron Sources”, 60th Annual Meeting of the Institute for Nuclear Materials Management (INMM), Palm Desert, California, 14-18 July 2019.
In this work a genetic algorithm is used to determine the neutron energy flux spectra of a nuclide source from a scintillation detector output. The neutron reconstructed spectrum can be used as a tool to identify and characterize neutron sources. Many existing spectrum reconstruction codes today require an a priori approximation of the neutron flux to operate. For the method developed here various existing measurements are incorporated into the algorithm in order to lessen the required knowledge of the source in question and guide the reconstructed spectra towards one that is physically realizable. The detector is modeled using a response matrix generated by MCNPX-PoliMi, which is then verified experimentally. The algorithm is then applied to a a variety of known sources to reconstruct their flux spectra. The reconstructed spectra is compared against reference spectra, and errors quantified. The genetic algorithm alone is capable of producing a rough approximation of the true neutron current in very few genetic generations. Subsequent generations produce refinement in the population average, though do little to change the best found neutron current approximation.