Neural Networks for the Neutron Spectrum Determination Based on the Foil Activation Method

Abstract

The multi-foil activation method using (n,p) and (n,α) of threshold reactions coupled with the neural network technique is considered to determine the neutron spectra of the 252Cf fission neutron source of the Research Institute for Radiation Biology and Medicine (RIRBM), Hiroshima University and a reactor at Kinki University (UTR-KINKI). Training data for the neural network were prepared from computational data sets of Monte Carlo calculation. The reproducibility of the calculated activities was verified to be within 15% of the experimentally determined activities. As it has been shown that the incident neutron spectrum of the 252Cf fission neutron source will reproduce the Maxwellian function, the reactor neutron spectrum of UTR-KINKI is successfully determined. A comparison with a previous evaluation of the reactor spectrum using an unfolding NEUPAC code is also given.