An enhanced method of neural network algorithm with multi-coupled gamma and neutron characteristic information for identifying plutonium and uranium

Technologies of supervising special nuclear material (SNM) have developed for scenarios of both nuclear verification and nuclear security. The traditional methods have a potential risk of leaking sensitive information of SNM especially during the nuclear disarmament. A neural network algorithm based on characteristic information extracted from low-resolution gamma-ray spectra of SNMs has been proposed to solve the problem. To improve the performance of this approach, a 3He neutron detector is added and optimized afterwards. Various SNM items, including 239Pu metallic hemispherical shells, canned 238PuO2 powder and high enriched uranium hemispherical shells, are used to validate this approach. The results show that the method with multi-coupled low-resolution gamma-ray and neutron characteristics is more accurate and eligible than that with only gamma-ray characteristics.