Study on unfolding method of neutron spectrum of BSS (Bonner Sphere Spectrometer) based on compressed sensing

Abstract

The Bonner Sphere Spectrometer (BSS) is widely used for neutron spectrum measurement for its broad energy sensitivity (from thermal to GeV). Information of neutron spectrum cannot be directly obtained and must be unfolded from the measurement readings. In this paper, a novel method of unfolding neutron spectrum from BSS measurement based on compressive sensing (CS) is proposed. Sparse representation based on learning dictionary is employed. The orthogonal matching pursuit (OMP) algorithm and the K-SVD algorithm are applied for sparse coding and dictionary update. The SL0 algorithm is used for reconstruction. Computational experiment is carried out, in which 5 types of the neutron spectrum are tested, including the D–T fusion neutron spectrum, Watt fission spectrum, Maxwell fission spectrum, the evaporation spectrum and 241Am-Be neutron spectrum taken from ISO standard 8529-1. The computational experiment reveals that: firstly, with appropriate sparse representation, neutron spectrum can be unfolded with high accuracy; secondly, the unfolding method based on CS is very suitable for solving the highly under determined problem. For instance, neutron spectrum with 300 energy groups is unfolded with 7 equations.