Unfolding of mono-energy neutron spectra using artificial neural network based on LMBP training algorithm

Abstract

In this work, neutron spectra are unfolded using artificial neural networks (ANNs). The neutron response of the NE213 scintillator detector (pulse height distribution) is calculated as the data required to unfold the energy spectrum using the MCNPX-PHOTRACK calculation code and to use the software to generate analytical response functions. A two-layer neural network was used in the artificial neural network method. The two-layer neural network has a hidden layer (the first layer) with the desired number of neurons depending on the problem and a layer with a neuron whose function must be the purelin function, or y = x. In this query, the Levenberg-Marquardt method (LMM), which has a high computational speed in the learning method, is used to train the network. It is good for the desired energies. The results also show that the noise in the unfolding spectrum obtained by the LMM method is lower than that obtained by the GRAVEL method.