Neutron spectroscopy with TENIS using an artificial neural network.

In this research, a ThErmal Neutron Imaging System (TENIS) consisting of two perpendicular sets of plastic scintillator arrays for boron neutron capture therapy (BNCT) application has been investigated in a completely different approach for neutron energy spectrum unfolding. TENIS provides a thermal neutron map based on the detection of 2.22 MeV gamma-rays resulting from H(n, γ)D reactions, but in the present study, the 70-pixel thermal neutron images have been used as input data for unfolding the energy spectrum of incident neutrons. Having generated the thermal neutron images for 10 incident mono-energetic neutrons, a 70 × 109 response matrix has been generated using the MCNPX2.6 code for feeding into the artificial neural network tools of MATLAB. The errors of the final results for mono-energetic neutron sources are less than 10% and the root mean square error (RMSE) for the unfolded neutron spectrum of Cf is about 0.01. The agreement of the unfolding results for mono-energetic and Cf neutron sources confirms the performance of the TENIS system as a neutron spectrometer.