Automatic classification of multi-channel pulse shape discrimination for fast neutron radiography acquisition

Abstract

The fast neutron radiography system is a type of non-destructive testing used in various fields such as homeland security, material classification, and neutron source location detection. There are two methods for measuring fast neutrons: a thermal neutron capture detection system and a recoil proton-based neutron scattering detection system. In particular, the recoil proton-based neutron detection system can be utilized as a fast neutron radiography system. In this thesis, we proposed a machine learning-based algorithm. The proposed algorithm automatically classifies neutron signals in the entire energy range from pulse shape discrimination results consisting of the multi-channel organic plastic scintillators and silicon photomultiplier (SiPM) for fast neutron radiography acquisition. Through the characteristics of the data distribution in the high-energy region, the PSD results in the low-energy region were effectively classified. As a result of applying the algorithm to the neutron and gamma-ray signals obtained by the time of flight measurement, the AUC of the ROC curve was 0.9987, which showed excellent classification performance.