Neutron spectrum unfolding code based on iterative method combined with artificial neural networks for bonner sphere spectrometer

In this work, an improved neutron spectrum unfolding code 'Artificial Neural Network Iterative' (ANN_Iter), which utilizes the iterative method combined with artificial neural networks, is proposed for Bonner sphere spectrometer. 201 neutron energy spectra and the corresponding measured responses of Bonner spheres with 15 polyethylene shells of different diameter were provided by an IAEA report. In the process of unfolding, the pre-trained artificial neural network provides a guess spectrum according to the 'knowledge' it has learned. The guess spectrum was used as the default initial spectrum in the iterative method, which unfolds the spectrum by the response function and the measured counts. The results show that the ANN_Iter's unfolded spectra have a good agreement with the desired spectra. In addition, their accuracy has improved compared with ANN's guess spectra. (author)